machine learning &

knowledge extraction

Systematic Review
Alzheimer’s Disease Detection Using Deep Learning on
Neuroimaging: A Systematic Review
Mohammed G. Alsubaie 1,2, * , Suhuai Luo 1 and Kamran Shaukat 1,3,4, *

1 School of Information and Physical Sciences, The University of Newcastle, Newcastle, NSW 2308, Australia
2 Department of Computer Science, College of Khurma University College, Taif University,
Taif 21944, Saudi Arabia
3 Centre for Artificial Intelligence Research and Optimisation, Design and Creative Technology Vertical,
Torrens University, Ultimo, Sydney, NSW 2007, Australia
4 Department of Data Science, University of the Punjab, Lahore 54890, Pakistan
* Correspondence: [email protected] (M.G.A.); [email protected] (K.S.)

Abstract: Alzheimer’s disease (AD) is a pressing global issue, demanding effective diagnostic
approaches. This systematic review surveys the recent literature (2018 onwards) to illuminate the
current landscape of AD detection via deep learning. Focusing on neuroimaging, this study explores
single- and multi-modality investigations, delving into biomarkers, features, and preprocessing
techniques. Various deep models, including convolutional neural networks (CNNs), recurrent
neural networks (RNNs), and generative models, are evaluated for their AD detection performance.
Challenges such as limited datasets and training procedures persist. Emphasis is placed on the need
to differentiate AD from similar brain patterns, necessitating discriminative feature representations.
This review highlights deep learning’s potential and limitations in AD detection, underscoring
dataset importance. Future directions involve benchmark platform development for streamlined
comparisons. In conclusion, while deep learning holds promise for accurate AD detection, refining
models and methods is crucial to tackle challenges and enhance diagnostic precision.

Keywords: Alzheimer’s disease; AD detection; convolutional neural network; recurrent neural

network; graph neural network; autoencoders
Citation: Alsubaie, M.G.; Luo, S.;
Shaukat, K. Alzheimer’s Disease
Detection Using Deep Learning on
Neuroimaging: A Systematic Review.
Mach. Learn. Knowl. Extr. 2024, 6, 1. Introduction
464–505. https://doi.org/10.3390/ Advancements in medical sciences and healthcare have led to improved health indica-
make6010024 tors and increased life expectancy, contributing to a global population projected to reach
Academic Editor: Andreas
around 11.2 billion by 2100 [1]. With a substantial rise in the elderly population, projections
Holzinger suggest that by 2050, approximately 21% of the population will be over 60, resulting in
a significant elderly demographic of two billion [2]. As the elderly population grows,
Received: 29 December 2023 age-related diseases, including Alzheimer’s disease (AD), have become more prevalent.
Revised: 2 February 2024
AD, the most common form of dementia, is a progressive and incurable neurodegenerative
Accepted: 6 February 2024
disorder characterized by memory loss, cognitive decline, and difficulties in daily activ-
Published: 21 February 2024
ities [3]. While the exact cause of AD remains unknown, genetic factors are believed to
play a significant role [4]. Pathologically, the spread of neurofibrillary tangles and amyloid
plaques in the brain disrupts neuronal communication and leads to the death of nerve cells,
Copyright: © 2024 by the authors.
resulting in a smaller cerebral cortex and enlarged brain ventricles [5].
Licensee MDPI, Basel, Switzerland. AD is an irreversible and progressive neurodegenerative disorder that gradually
This article is an open access article impairs memory, communication, and daily activities like speech and mobility [6]. It is the
distributed under the terms and most prevalent form of dementia, accounting for approximately 60–80% of all dementia
conditions of the Creative Commons cases [7]. Mild cognitive impairment (MCI) represents an early stage of AD, characterized
Attribution (CC BY) license (https:// by mild cognitive changes that are noticeable to the affected individual and their loved
creativecommons.org/licenses/by/ ones while still allowing for the performance of daily tasks. However, not all individuals
4.0/). with MCI will progress to AD. Approximately 15–20% of individuals aged 65 or older

Mach. Learn. Knowl. Extr. 2024, 6, 464–505. https://doi.org/10.3390/make6010024 https://www.mdpi.com/journal/make

Mach. Learn. Knowl. Extr. 2024, 6 465

have MCI, and within five years, around 30–40% of those with MCI will develop AD [8].
The conversion period from MCI to AD can vary between 6 to 36 months but typically
lasts around 18 months. MCI patients are then classified as MCI converters (MCIc) or non-
converters (MCInc) based on whether they transition to AD within 18 months. Additionally,
there are other less commonly mentioned subtypes of MCI, such as early or late MCI [9].
The primary risk factors for AD include family history and the presence of specific
genes in an individual’s genome [10]. The detection of AD relies on a comprehensive
evaluation that includes clinical examinations and interviews with the patient and their
family members. However, a definitive detection can only be confirmed through autopsy,
which limits its clinical applicability. Autopsy-confirmed cases of AD have been utilized in
some studies to establish reliable detection. In the absence of definitive diagnostic data,
additional criteria are required to confirm the presence of AD and enable its detection in
living patients. The National Institute of Neurological Disorders and Stroke (NINCDS)
and the Alzheimer’s Disease and Related Disorders Association (ADRDA) established
clinical diagnostic criteria for AD in 1984 [11], which were revised in 2007 to include
memory impairment and additional features such as abnormal neuroimaging (MRI and
PET) or abnormal cerebrospinal fluid biomarkers [12]. The National Institute on Aging-
Alzheimer’s Association (NIA-AA) has also revised the diagnostic criteria, incorporating
brain amyloid, neuronal damage, and degeneration measures. Regular updates to the
criteria, approximately every 3–4 years, are suggested to incorporate new knowledge about
the pathophysiology and progression of the disease [13].
Commonly used assessment tools for AD include the Mini-Mental State Examination
(MMSE) [14–16] and the Clinical Dementia Rating (CDR) [17–19]. However, it is important
to note that utilizing these tests as definitive benchmarks for AD may not provide complete
accuracy. The accuracy of clinical detection compared to postmortem detection ranges
between 70% and 90% [20–23]. Despite its limitations, clinical detection remains the
best available reference standard, although the accessibility of recognized biomarkers is
often limited.
Globally, dementia affects 35.6 million people over the age of 60 as of 2010, with
projections indicating a doubling every 20 years, reaching 115 million by 2050 [24]. In
Australia, dementia has become the second leading cause of death, leading to significant
economic implications due to the rising nursing care costs for AD patients [25,26]. Despite
various treatment strategies being explored, their success has been limited, underscoring
the importance of early and accurate detection for appropriate interventions [27].
To address the need for unbiased clinical decision making and the ability to differ-
entiate AD and its stages from normal controls (NCs), a multi-class classification system
is necessary [28–30]. While predicting conversion to mild cognitive impairment (MCI) is
more valuable than solely classifying AD patients from normal controls, research often
focuses on distinguishing AD from normal controls, providing insights into the early signs
of AD [31–38]. The key challenge lies in accurately determining MCI and predicting disease
progression [34,39]. Although computer-aided systems cannot replace medical expertise,
they can offer supplementary information to enhance the accuracy of clinical decisions.
Furthermore, studies have also considered other stages of the disease, including early or
late MCI [40,41].
Detecting AD using artificial intelligence presents several challenges for researchers.
Firstly, there is often a limitation in the quality of medical image acquisition and errors in
preprocessing and brain segmentation [42]. The quality of medical images can be compro-
mised by noise, artefacts, and technical limitations [43], which can affect the accuracy of
AD detection algorithms. Additionally, pre-processing and segmentation technique errors
further hinder the reliable analysis of these images.
Another challenge lies in the unavailability of comprehensive datasets encompassing
a wide range of subjects and biomarkers. Building robust AD detection models requires
access to diverse datasets that cover different stages of the disease and include various
Mach. Learn. Knowl. Extr. 2024, 6 466

biomarkers [44]. However, obtaining such comprehensive datasets with a large number of
subjects can be difficult, limiting the ability to train and evaluate AI models effectively.
In AD detection, there is a low between-class variance in different stages of the disease.
Distinguishing between these stages can be challenging due to limited variation in imaging
characteristics. For example, certain signs associated with AD, such as brain shrinkage, can
also be observed in the brains of normal, healthy older individuals. This similarity can lead
to ambiguity in classification and make it harder to differentiate between AD and normal
aging [28].
The ambiguity of boundaries between AD/MCI (mild cognitive impairment) and
MCI/NC (normal control) based on AD diagnostic criteria is another obstacle [45]. The
diagnostic criteria for AD and its transitional stage, MCI, can be subjective and open to inter-
pretation. Determining the boundaries between AD/MCI and MCI/NC can be challenging,
as there may be overlap and inconsistency in classification based on these criteria.
Moreover, the lack of expert knowledge, particularly in identifying regions of interest
(ROIs) in the brain, poses a challenge [46,47]. Accurate identification of specific brain
regions relevant to AD requires expertise, but expertise in identifying these ROIs can be
limited. This limitation hampers the development of precise AI algorithms for AD detection.
Lastly, medical images used in AD detection are more complex compared to natural
images [48,49]. Magnetic resonance imaging (MRI) and positron emission tomography
(PET) scans often exhibit intricate structures, subtle variations, and imaging artefacts.
Analyzing and interpreting these complex medical images requires specialized algorithms
and techniques tailored to the unique characteristics of these imaging modalities.
Overcoming these challenges is crucial for advancing AI-based AD detection systems,
as they hold significant potential for early and accurate disease detection. Addressing issues
related to image quality, dataset availability, classification ambiguity, expert knowledge,
and the complexity of medical images will contribute to the development of more reliable
and effective AI algorithms for AD detection.
Numerous studies have been dedicated to detecting Alzheimer’s disease (AD) using ma-
chine learning techniques. These studies have extensively covered various aspects, including
different classifiers [50–55], monomodal and multimodal models [56–60], feature extraction
algorithms [61–63], feature selection methods [64–66], validation approaches, and dataset
properties [67,68]. The findings from these studies have highlighted the effectiveness of
machine learning approaches in analyzing AD and have been further complemented by com-
petitions such as CADDementia (https://caddementia.grand-challenge.org/) (accessed on
7 July 2023), TADPOLE (https://tadpole.grand-challenge.org/) (accessed on 9 July 2023), and
The Alzheimer’s Disease Big Data DREAM Challenge (https://www.synapse.org/#!Synapse:
syn2290704/wiki/64632) (accessed on 9 July 2023). These competitions provide a valuable
platform for unbiased comparisons of algorithms and tools using standardized data, engaging
participants worldwide.
However, traditional machine learning approaches have faced limitations in dealing
with the intricacies of AD detection [69–71]. Distinguishing specific features within similar
brain image patterns is crucial but challenging. In recent years, significant advancements
in deep learning algorithms, fueled by the enhanced processing capabilities of graphics
processing units (GPUs), have brought about a paradigm shift in performance across
various domains, including object recognition [72–74], detection [75–77], tracking [78–80],
image segmentation [81–83], and audio classification [84,85]. Deep learning, a subfield
of artificial intelligence that emulates the human brain’s data processing and pattern
recognition mechanisms, holds great promise in medical image analysis.
This paper aims to comprehensively review the current landscape of Alzheimer’s
disease (AD) detection using deep learning techniques. Specifically, our goal is to explore
the application of deep learning in both supervised and unsupervised modes to gain
deeper insights into AD. By examining the latest findings and emerging trends, we examine
Alzheimer’s disease detection using deep learning.
 the application of deep learning in both supervised and unsupervised modes to gain
deeper insights into AD. By examining the latest findings and emerging trends, we
Mach. Learn. Knowl. Extr. 2024, 6 467
examine Alzheimer’s disease detection using deep learning.
This paper looks at the different methodologies and approaches employed in Alz-
heimer’s disease detection using deep learning. By analyzing recent research, we aim to
This paper looks
comprehensively at the different
understand methodologies
the progress madeand approaches
in this field. Weemployed in Alzheimer’s
investigate the use of
disease detection using deep learning. By analyzing recent
deep learning models to discover valuable information about Alzheimer’sresearch, we aim to comprehen-
disease,
sively understand
shedding light on the
the progress madeofinknowledge.
current state this field. We investigate the use of deep learning
models to discover
Through valuableliterature
an extensive information aboutwe
review, Alzheimer’s
collect and disease, shedding
synthesize light on
the most the
recent
current state of knowledge.
results regarding detecting Alzheimer’s disease using deep learning. Our analysis encom-
Through an extensive literature review, we collect and synthesize the most recent
passes a range of supervised and unsupervised deep learning techniques, exploring their
results regarding detecting Alzheimer’s disease using deep learning. Our analysis encom-
effectiveness and potential for improving the accuracy of Alzheimer’s disease detection.
passes a range of supervised and unsupervised deep learning techniques, exploring their
In addition, we examine current trends in Alzheimer’s disease detection using deep
effectiveness and potential for improving the accuracy of Alzheimer’s disease detection.
learning, identifying key areas of interest and innovation. By understanding the current
In addition, we examine current trends in Alzheimer’s disease detection using deep
landscape, we aim to provide valuable insights into the direction of research and devel-
learning, identifying key areas of interest and innovation. By understanding the cur-
opment in this rapidly evolving field.
rent landscape, we aim to provide valuable insights into the direction of research and
The rest of this paper is organized as follows: Section 2 delves into the Alzheimer’s
development in this rapidly evolving field.
disease
Thedetection system.
rest of this paperSection 3 discusses
is organized the review
as follows: protocol,
Section while
2 delves intoSection 4 explores
the Alzheimer’s
input
diseasemodalities, input types,
detection system. datasets,
Section and prediction
3 discusses the review tasks: exploring
protocol, variations
while Section in AD
4 explores
detection. Section 5 focuses on deep learning for Alzheimer’s disease detection,
input modalities, input types, datasets, and prediction tasks: exploring variations in AD followed
by Section Section
detection. 6, which highlights
5 focuses trending
on deep technologies
learning in AD Studies.
for Alzheimer’s Section 7 discusses
disease detection, followed
the heterogeneous nature of AD. Section 8 addresses the challenges encountered
by Section 6, which highlights trending technologies in AD Studies. Section 7 discusses in this
domain, and Section 9 provides insights into future perspectives and recommendations.
the heterogeneous nature of AD. Section 8 addresses the challenges encountered in this
Finally,
domain,inandSection 10, 9we
Section draw our
provides conclusion.
insights into future perspectives and recommendations.
Finally, in Section 10, we draw our conclusion.
2. Alzheimer’s Disease Detection System
2. Alzheimer’s Disease Detection
Figure 1 illustrates the AD System
detection system, an intricate and comprehensive
frameFigure 1 illustrates
work designed the AD detection
to facilitate system,
the efficient an intricate
detection of AD.and
Thiscomprehensive
system harnessesframe-
the
work designed
synergistic to facilitate
integration the efficient
of essential detection ofincluding
components, AD. This brain
systemscans,
harnesses the syner-
preprocessing
gistic integration
techniques, dataof essential components,
management including
strategies, brain scans,
deep learning preprocessing
models, techniques,
and evaluation. To-
gether, these elements establish a robust foundation for the system, ensuringthese
data management strategies, deep learning models, and evaluation. Together, ele-
its effec-
ments establish
tiveness, a robust
reliability, and foundation
precision. for the system, ensuring its effectiveness, reliability,
and precision.

Figure
Figure 1.
1. Illustration
Illustration depicting the interconnected
depicting the interconnectedelements
elementsofofthe
theAD
ADdetection
detection system.
system.

2.1. Brain
2.1. Brain Scans
Scans
Brain scans
Brain scansplay
playa afundamental
fundamentalrolerole
in in
thethe
ADAD detection
detection system,
system, as they
as they provide
provide crit-
critical information about structural and functional changes associated
ical information about structural and functional changes associated with AD [86]. with AD [86].
Various imaging
Various imaging techniques
techniques areare used
used to obtain detailed
to obtain detailed images
images ofof the
the brain,
brain, including
including
magnetic resonance imaging (MRI), positron emission tomography (PET), and
magnetic resonance imaging (MRI), positron emission tomography (PET), and diffusion
diffusion
tensor imaging
tensor imaging (DTI)
(DTI) [87].
[87]. MRI
MRI uses
uses magnetic
magnetic fields
fields and
and radio
radio waves
waves to
to generate
generate high-
high-
resolution images, revealing anatomical features of the brain [88]. PET involves
resolution images, revealing anatomical features of the brain [88]. PET involves injecting injecting a
radioactive tracer into the body, which highlights specific areas of the brain associated with
a radioactive tracer into the body, which highlights specific areas of the brain associated
AD pathology [89]. DTI measures the diffusion of water molecules in brain tissue, which
with AD pathology [89]. DTI measures the diffusion of water molecules in brain tissue,
allows for the visualization of white matter pathways and assessment of the integrity of
which allows for the visualization of white matter pathways and assessment of the integ-
neuronal connections [90].
rity of neuronal connections [90].
Brain scans provide valuable information about structural changes, neurochemical
abnormalities, and functional alterations in people with AD [91]. These scans can detect
the presence of amyloid plaques and neurofibrillary tangles, the characteristic pathologies
of AD, and reveal patterns of brain atrophy and synaptic dysfunction [12].
The data acquired by the brain scan serve as the basis for further analysis and interpre-
tation [92]. However, it is important to note that interpreting brain scans requires expertise
Mach. Learn. Knowl. Extr. 2024, 6 468

and knowledge of neuroimaging. Radiologists and neurologists often collaborate to ensure

accurate and reliable interpretation of scans [93].
In the context of the AD detection system, brain scans serve as the primary input data,
capturing the unique characteristics of each individual’s brain. These scans undergo further
preprocessing and analysis to extract meaningful features and patterns that can contribute
to the detection and classification of Alzheimer’s disease [94].

2.2. Preprocessing
Preprocessing plays a critical role in the AD detection system by applying essential
steps to enhance the quality and reliability of data obtained from brain scans. This subsec-
tion focuses on the key preprocessing techniques used to prepare acquired imaging data
before further analysis and interpretation.
One of the initial preprocessing steps is image registration, which involves aligning
brain scans to a common reference space. This alignment compensates for variations in
positioning and orientation, ensuring consistent analyses across different individuals and
time points [95]. Commonly used techniques for image registration include affine and
non-linear transformations.
Following image registration, intensity normalization techniques are applied to ad-
dress variations in signal intensity between scans. These techniques aim to normalize
intensity levels, facilitating more accurate and reliable comparisons among different brain
regions and subjects [96]. Common normalization methods include z-score normalization
and histogram matching.
Another important preprocessing step is noise reduction, which aims to minimize
unwanted artefacts and noise that can interfere with subsequent analyses. Techniques such
as Gaussian filtering and wavelet denoising are commonly employed to reduce noise while
preserving important features in brain images [97].
Spatial smoothing is an additional preprocessing technique that involves applying a
smoothing filter to the data. This process reduces local variations and improves the signal-
to-noise ratio, facilitating the identification of relevant patterns and structures in brain
scans [98]. Furthermore, motion correction is performed to address motion-related artefacts
that may occur during brain scan acquisition. Motion correction algorithms can detect and
correct head movements, ensuring that the data accurately represent the structural and
functional characteristics of the brain [99].
It is important to note that preprocessing techniques may vary depending on the imag-
ing modality used, such as MRI or PET. Each modality may require specific preprocessing
steps tailored to its characteristics and challenges.

2.3. Data Management

Data management is a crucial component of the Alzheimer’s disease detection sys-
tem, as it involves the efficient organization, storage, and handling of large quantities of
imaging and clinical data. This subsection focuses on the key aspects of data manage-
ment in Alzheimer’s disease research, including data acquisition, storage, integration, and
quality control.
Data acquisition involves the collection of imaging data from various modalities such
as MRI, PET, or CT scans, as well as clinical data, including demographic information, cog-
nitive assessments, and medical history. Standardized protocols and validated assessment
tools are used to ensure consistent data collection procedures [92].
Once data has been acquired, the storage of large-scale imaging and clinical datasets
requires efficient and scalable storage solutions. Various database management systems,
such as relational databases or NO Structured Query Language (NoSQL) databases, can
be used to organize and store data securely and provide efficient retrieval and query
capabilities [100].
The integration of heterogeneous data from different sources is crucial to enable
comprehensive analysis and interpretation. Data integration techniques, such as data
Mach. Learn. Knowl. Extr. 2024, 6 469

fusion or data harmonization, aim to combine data from multiple modalities or studies into
a unified format to ensure compatibility and enable holistic analysis [101].
Data quality control is an essential step in guaranteeing the reliability and validity
of the data collected. It involves identifying and correcting anomalies, missing values,
outliers, or artefacts that could affect the accuracy and integrity of subsequent analyses.
Quality control procedures, including data cleaning and validation checks, are applied to
maintain data consistency and accuracy [102].
Effective data management also involves adherence to ethical and privacy guidelines
to protect participant confidentiality and ensure data security. Compliance with regulatory
requirements, such as obtaining informed consent and anonymizing data, is essential to
protect participants’ rights and maintain data integrity.

2.4. Deep Learning Model

Deep learning models have emerged as powerful tools in Alzheimer’s disease detec-
tion, leveraging their ability to learn complex patterns and representations from large-scale
imaging datasets. This subsection explores the application of deep learning models in
Alzheimer’s disease detection, highlighting their architectures, training strategies, and
performance evaluation.
Convolutional neural networks (CNNs) have been widely adopted in Alzheimer’s
disease research due to their effectiveness in analyzing spatial relationships within brain
images. CNNs consist of multiple convolutional and pooling operations layers, followed by
fully connected layers for classification [103]. These architectures enable automatic feature
extraction and hierarchical learning, capturing local and global brain scan patterns.
To train deep learning models, large annotated datasets are required. The Alzheimer’s
Disease Neuroimaging Initiative (ADNI) and other publicly available datasets, such as the
Open Access Series of Imaging Studies (OASIS) and the Australian Imaging, Biomarkers
and Lifestyle (AIBL) study, have played crucial roles in facilitating the development and
evaluation of deep learning models for Alzheimer’s disease detection [104].
Training deep learning models involves optimizing their parameters using labelled
data. Stochastic gradient descent (SGD) and its variants, such as Adam and RMSprop, are
commonly used optimization algorithms for deep learning [105]. Additionally, regulariza-
tion techniques like dropout or batch normalization are employed to prevent overfitting
and improve generalization performance [106].
The performance of deep learning models in Alzheimer’s disease detection is typically
evaluated using metrics such as accuracy, sensitivity, specificity, and area under the receiver
operating characteristic curve (AUC-ROC). Cross-validation or independent test sets are
used to assess the generalization ability of the models [107].
Moreover, transfer learning has shown promise in Alzheimer’s disease detection by
leveraging pre-trained deep learning models on large-scale natural image datasets, such as
ImageNet. By fine-tuning the pre-trained models on brain image data, transfer learning
allows for effective knowledge transfer and improved performance, even with limited
labelled training samples [108].

2.5. Evaluation
Evaluation plays a crucial role in assessing the performance and effectiveness of
Alzheimer’s disease detection systems. This subsection focuses on the evaluation metrics
and methodologies commonly employed in the assessment of these systems, providing
insights into the accuracy and reliability of the detection results.
Evaluation metrics in Alzheimer’s disease detection often include accuracy, sensi-
tivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC).
Accuracy measures the overall correctness of the system’s predictions, while sensitivity
and specificity assess the system’s ability to correctly identify positive and negative cases,
respectively [109]. AUC-ROC provides a comprehensive measure of the system’s discrimi-
nation ability, capturing the trade-off between true positive rate and false positive rate [110].
Mach. Learn. Knowl. Extr. 2024, 6 470

Cross-validation is a widely used evaluation methodology to assess the generalization

performance of Alzheimer’s disease detection systems. In k-fold cross-validation, the
dataset is divided into k subsets, and the system is trained and tested k times, with each
subgroup serving as the testing set once. This approach provides a more robust estimation
of the system’s performance by utilizing the entire dataset for evaluation [111].
Independent test sets are also utilized for evaluation, where a separate dataset, not
seen during training, is used to assess the system’s performance. These test sets provide
an objective assessment of the system’s ability to generalize to unseen data, reflecting its
real-world performance [29].
Furthermore, performance comparison against baseline methods and existing state-
of-the-art algorithms is important to demonstrate the advancement and effectiveness of
Alzheimer’s disease detection systems. These comparisons help researchers identify the
strengths and limitations of their proposed approaches and highlight the progress made in
the field [112].

3. The Review Protocol

This review aims to systematically analyze and synthesize the recent advancements
in AD detection using CNNs, RNNs, and generative modeling techniques. This review
will focus on papers published between 2018 and 2023, aiming to provide a comprehensive
overview of the state-of-the-art methods, their performance, and potential contributions to
AD detection.

3.1. Inclusion Criteria

The following inclusion criteria were be applied when selecting papers for this review:
• Papers published between 2018 and 2023.
• Papers that specifically address AD detection using CNNs, RNNs, or generative
modeling techniques.
• Papers that report on original research, including novel methodologies, experimental
studies, or significant advancements in the field.
• Papers published in peer-reviewed journals or presented at reputable conferences.

3.2. Search Strategy

A systematic search was conducted to identify relevant papers for inclusion in this
review. The search was performed in major scientific databases, such as Scopus (https:
//scopus.com/), PubMed (https://pubmed.ncbi.nlm.nih.gov/), IEEE Xplore (https://
ieeexplore.ieee.org/Xplore/home.jsp), ACM Digital Library (https://dl.acm.org/), and
Google Scholar (https://scholar.google.com/). The search terms included variations of
“Alzheimer’s disease”, “AD detection”, “CNN”, “Convolutional Neural Network”, “RNN”,
“Recurrent Neural Network”, “Generative Modeling”, and their combinations. The search
was limited to papers published between January 2018 and December 2023.

3.3. Selection Process

The selection process consisted of two stages: screening and eligibility assessment.
• Screening: Titles and abstracts of the retrieved papers were screened independently
by two reviewers to determine their relevance to the review topic. Papers that clearly
did not meet the inclusion criteria were excluded at this stage.
• Eligibility Assessment: The full texts of the remaining papers were obtained and
independently assessed by two reviewers. Any discrepancies in eligibility assessment
were resolved through discussion or consultation with a third reviewer if necessary.

3.4. Data Extraction and Synthesis

Data were extracted using a standardized form to collect relevant information from
the selected papers. The extracted data included authors, publication year, study objec-
tives, dataset characteristics, CNN, RNN, or generative modeling architectures, employed
Mach. Learn. Knowl. Extr. 2024, 6 471

evaluation metrics, and key findings. The extracted data were synthesized to provide
a comprehensive summary of the methodologies, performance, and advancements in
AD detection using CNNs, RNNs, and generative modeling.

3.5. Data Analysis

The synthesized data were analyzed qualitatively to identify common trends, chal-
lenges, and advancements in AD detection using the specified techniques. Key findings,
strengths, and limitations of the approaches were highlighted. If feasible, a quantitative
analysis, such as meta-analysis or statistical comparisons, was conducted to assess the
overall performance of the methods across the included papers.

3.6. Reporting
The results of this review were reported following the Preferred Reporting Items
for Systematic Reviews and Meta-Analyses (PRISMA) [113] guidelines. The findings are
organized and presented in a coherent manner, providing a clear overview of the state-of
the-art in Alzheimer’s disease detection using CNNs, RNNs, and generative modeling.

3.7. Limitations
This review may have certain limitations. Firstly, it relied on the availability and
quality of published papers within the specified time frame. Secondly, the search strategy
may not have captured all relevant papers, although efforts were made to include major
databases and employ appropriate search terms. Lastly, this review focuses on AD detection
using specific neural network architectures and may not cover other relevant approaches
or techniques. By following this review protocol, we aimed to minimize bias and ensure
a systematic and comprehensive analysis of the selected papers. We strived to address
these limitations by conducting a thorough search, employing standardized screening and
eligibility assessment processes, and reporting our findings transparently.

4. Input Modalities, Input Types, Datasets, and Prediction Tasks: Exploring Variations
in AD Detection
In the realm of deep learning for AD detection, various modalities are employed to cap-
ture different aspects of the disease. Structural magnetic resonance imaging (sMRI) provides
detailed anatomical information, aiding in brain atrophy detection through cross-sectional
or longitudinal scans [114]. PET offers functional insights, with fluorodeoxyglucose-
positron emission tomography (FDG-PET) detecting glucose hypometabolism and amyloid-
PET identifying amyloid deposits associated with AD [114,115]. Resting-state fMRI mea-
sures functional connectivity [116], while EEG records brain electrical activity linked to AD
degeneration [117–120]. Some studies employ diffusion tensor imaging (DTI) [121].
Cognitive assessments, e.g., MMSE and Alzheimer’s Disease Assessment Scale-Cognitive
Subscale (ADAS-Cog), evaluate cognitive abilities. Genetic factors, notably the APOE gene
(e2, e3, e4 forms), influence AD risk, with the e4 form increasing susceptibility [122]. Com-
bining physiological, chemical, and cognitive data, the APOE genotype, and demographics
provides a comprehensive AD detection approach. Deep learning architectures primarily use
3D MRI and PET scans [122].

4.1. Input Data Selection

In deep learning for AD detection, selecting input data is crucial. This section explores
methods to convert medical images into suitable formats for training deep neural networks
(DNNs), including 2D slicing, patch extraction, and feature selection/construction. To handle
the computational demands, researchers split 3D neuroimages into 2D slices using established
architectures designed for 2D input data [48,123–125]. Another approach extracts patches
from 3D images, employing RoI-based or data-driven methods [47,126,127]. Feature selec-
tion/extraction techniques, such as PCA, mitigate computational costs, although they have
limitations for images [128–133].
Mach. Learn. Knowl. Extr. 2024, 6 472

Despite these approaches, concerns persist about losing valuable relationships in

neuroimages, and selection processes often rely on prior knowledge. Numerical features,
demographic information, and cognitive measures are used alongside image-based features
for AD detection [134,135]. Addressing the temporal nature of fMRI data, researchers use
various approaches, including 3D whole-brain images, 2D slices along the temporal axis,
and correlation matrices of pre-determined brain regions, [136] to enhance fMRI-based
AD detection.

4.2. Datasets
In AD detection, data availability is crucial for deep learning model development.
Collaborative efforts and publicly accessible datasets have enriched AD causes, symptoms,
and early detection research.
One prominent example is the Alzheimer’s Disease Neuroimage Initiative (ADNI),
which combines various data types to study cognitive impairment progression [137]. The
Open Access Series of Imaging Studies (OASIS) offers freely accessible neuroimaging
datasets for neuroscience advancements [138]. The Minimal Interval Resonance Imaging in
Alzheimer’s Disease (MIRIAD) dataset provides longitudinal T1 MRI scans for AD clinical
trial optimization [139]. The Australian Imaging, Biomarker & Lifestyle Flagship Study of
Aging (AIBL) dataset aids biomarker and lifestyle research related to AD onset [140].
In addition to these, universities and research centers, such as Chosun University Na-
tional Dementia Research Center, Davis Alzheimer’s Disease Center, and Dong-A Univer-
sity Korea, have their datasets [141,142]. These diverse resources empower AD researchers
and enhance our understanding of the disease.

4.3. Prediction Tasks

In healthcare, diagnostic predictions encompass binary classification, multi-class
classification, and regression-based predictions.
Binary classification is common in distinguishing healthy and diseased subjects, espe-
cially in diseases with limited data for each class or complex progression stages.
In AD detection, a 3-class classification separates subjects into AD, MCI, and NC
categories. Within MCI, distinguishing progressive MCI (pMCI) from stable MCI (sMCI)
adds complexity, resulting in a 4-class classification.
Regression-based predictions in AD research involve real-value output measures,
such as cognitive tests [143], progression likelihood [144], and time-to-event prediction
tasks [145], providing insights into quantifiable aspects of AD progression.
By considering these approaches, researchers gain valuable insights into different
facets of AD detection and prognosis, enhancing our ability to effectively detect and
monitor the disease.

5. Deep Learning for Alzheimer’s Disease Detection

Deep learning has become a prominent approach in AD detection using medical image
data, incorporating modalities like PET and MRI [71]. Initially, unsupervised pre-training
and network architectures like stacked autoencoders and Restricted Boltzmann machines
were prevalent, but since 2013, there has been a surge in deep learning applications for AD
detection, especially in neuroimaging [54,71,146].
Clinical research has identified critical AD-related changes: beta-amyloid plaque
accumulation, tau protein tangles, and brain atrophy [147]. Imaging modalities like MRI,
PET, and fMRI detect these changes, while CSF analysis quantifies specific proteins [148].
Brain tissue volume (grey matter, white matter, CSF) changes correlate with AD
severity [149]. Some studies use GM volume, while others consider all tissue volumes
for deep learning models [150–153]. Hippocampal atrophy, a risk factor for dementia
progression, is also used, with various representations (whole volume, patches, slices,
numerical features) employed for AD detection [28,146,154–156].
Mach. Learn. Knowl. Extr. 2024, 6 473

These advances in deep learning and multimodal imaging have improved AD de-
tection accuracy and effectiveness, leveraging CNNs, RNNs, and generative modelling
techniques. The following sections will explore specific methodologies and findings in
deep learning approaches for AD detection.

5.1. Convolutional Neural Networks for AD Detection

Convolutional neural networks (CNNs) have gained prominence in medical imaging
for AD detection, as they excel at learning hierarchical features from raw image data, en-
abling accurate predictions [157]. CNNs’ promising results in AD detection have garnered
attention from researchers and clinicians.
This systematic review focuses on CNNs in AD detection, providing a comprehensive
overview of the literature and summarizing key findings. Examining various approaches,
input types/modalities, techniques, evaluation metrics, and additional insights, we aim to
highlight CNNs’ potential as a diagnostic tool for AD [157].

5.1.1. Neuroimaging and CNNs for AD Analysis

CNNs have been employed for AD diagnosis, classification, prediction, and image
generation. Neuroimaging data, including T1-weighted MRI scans and PET images, serve
as foundational inputs [158]. CNNs adeptly extract features from these images, enabling
precise AD classification and prediction.

5.1.2. CNN-3D Architecture for AD Classification

CNN-3D architectures, designed to harness the 3D nature of neuroimaging data,
excel in capturing spatial relationships and fine details. Across various datasets, CNN-3D
models exhibit robust performance in AD classification, capitalizing on their ability to
extract discriminative features from 3D brain images.

5.1.3. GANs for Data Augmentation and Enhancement

Generative adversarial networks (GANs) generate synthetic brain images resembling
real ones, alleviating limited labelled data availability. By enhancing downstream tasks like
AD classification through realistic image synthesis, GANs contribute to improved model
generalization and performance.

5.1.4. Transfer Learning and Multimodal Fusion

Transfer learning fine-tunes pre-trained CNN models from general image datasets to
AD tasks, compensating for limited AD-specific data. Multimodal approaches, merging
data from diverse imaging modalities, enhance classification by capturing complementary
AD pathology facets.

5.1.5. Temporal Convolutional Networks (TCNs)

TCNs are another type of neural network architecture that can capture temporal
dependencies in sequential data. Unlike LSTMs, TCNs utilize 1D convolutional layers
with dilated convolutions to extract features from the temporal sequences. TCNs have
been applied to various AD-related tasks, including disease classification, progression
prediction, and anomaly detection. They offer computational efficiency and can capture
both short-term and long-term temporal dependencies in the data.

5.1.6. Dataset Quality and Interpretable Models

Dataset size and diversity critically impact a CNN’s performance. Standardized, broad
AD datasets are pivotal for robust models. Addressing interpretability, efforts should focus
on unveiling learned features and decision-making processes to enhance trustworthiness
and clinical applicability.
Mach. Learn. Knowl. Extr. 2024, 6 474

5.1.7. Overview of Convolutional Neural Network (CNN) Studies for AD Detection

Table 1 offers an overview of major studies that have utilized CNNs for AD detection
from 2018 to 2023. The table includes relevant information such as the study name, date,
input type/modalities, technique, evaluation metric, and additional notes. The aim of
this table is to provide a concise reference for researchers and practitioners interested in
CNN-based approaches for AD detection.

Table 1. Overview of CNN studies for AD detection.

Reference Input Data Technique Evaluation Notes

CNNs, including both 2D Combination of transfer learning
Accuracy, sensitivity, and specificity of the
[48] MRI and 3D models, as well as and 3D voxel data led to improved
different models.
RNNs. AD classification accuracy.
Pre-trained (CNN) model, CNN model with Softmax, SVM, and RF Deep learning with pre-trained
specifically ResNet50, used classifiers achieved high accuracy, ranging CNN models improved AD
[61] MRI
for automatic feature from 85.7% to 99%, outperforming other diagnosis to enhance patient
extraction. state-of-the-art models. survival rates.
CNN-based AD
classification algorithm
using coronal slices from
CNN-based AD classification algorithm
T1-weighted MRI images. It CNN-based AD classification
achieved AUCs of 0.91–0.94
T1-weighted was evaluated on data from demonstrated promising accuracy
[158] (within-dataset) and 0.88–0.89
MRI two populations (SNUBH and generalization for AD
(between-dataset). Processing time was
and ADNI) using classification with high AUC values.
23–24 s per person.
within-dataset and
between-dataset validations
with AUC.
Comparison with single modality CNN-based fusion of sMRI and DTI
CNN integrating sMRI and approach, analysis of data augmentation modalities on hippocampal variable
[159] sMRI, DTI
DTI modalities. for class balancing, and investigation of the ROI showing promising results on
impact of ROI size on classification results. ADNI dataset.
3D inception-based CNN using
3D inception-based CNN Comparison with a conventional
multi-modal fusion to outperform
[160] sMRI, DTI with the fusion of sMRI and AlexNet-based network using ADNI
conventional AlexNet-based
DTI modalities. dataset.
networks on the ADNI dataset.
Feature extraction using Patient classification using 3D-MRI,
Performance comparison of classification
AlexNet and classification i.e., extraction of 2D features and
[161] 3D MRI based on extracted features versus
using machine learning (ML) dimensionality reduction leading to
Softmax-based probability scores.
algorithms. improved accuracy.
A deep learning approach for early
3D CNN (HadNet) using Classification to segregate AD, MCI, and Alzheimer’s diagnosis using 3D
[162] MRI
stacked convolutions. healthy individuals. CNN with a reported accuracy of
88.31%.
CNN model for AD detection in
MRI images achieved 80% accuracy
Convolutional neural
[163] MRI Accuracy. on the OASIS dataset using
network (CNN) model.
Python’s Keras library, but needs a
performance improvement.
Combines content-based
image retrieval (CBIR) with Validation of an ensemble approach
a 3D capsule network, a The performance of the proposed model 3D capsule networks, CNNs, and a
3D-convolutional neural was evaluated using accuracy as the metric pre-trained 3D autoencoder for early
[164] 3D MRI
network (CNN), and for AD classification, and it achieved up to AD detection, showing CapsNet’s
pre-trained 3D-autoencoder 98.42% accuracy in AD classification. potential towards future
technology for early AD improvements.
detection.
CNN integrates the
multimodality information The proposed network was evaluated on
The proposed method reported
from T1MR and FDG-PET the ADNI dataset with T1-MR and
T1-weighted better performance, i.e., integration
images to diagnose AD. The FDG-PET images. Accuracy results were
[165] MRI, of T1-MR and FDG-PET data
CNN learns features directly 90.10% for CN vs. AD, 87.46% for CN vs.
FDGPET improved CNN results, showcasing
from the 3D images without pMCI, and 76.90% for sMCI vs. pMCI
AI’s potential in AD diagnosis.
the need for manually classification.
extracted features.
Mach. Learn. Knowl. Extr. 2024, 6 475

Table 1. Cont.

Reference Input Data Technique Evaluation Notes

The paper proposes
PFSECTL, a mathematical
PFSECTL model employed transfer
model using transfer
The proposed method achieved 95.73% learning with VGG-16 for feature
learning with VGG-16, a
accuracy for 3-way classification on the extraction and high accuracy
[166] MRI CNN architecture. The
ADNI database, but the specific classes demonstrates the potential of the
pretrained VGG-16 model
were not specified. proposed method towards AD
from ImageNet is used as a
detection.
feature extractor for
classification.
ADNet model for AD biomarker
extraction and classification
Convolutional neural
Accuracy of classification into AD, MCI, reported 52.3% accuracy in the
[167] sMRI networks (CNN) for feature
and CN groups. CADDementia challenge,
extraction and classification.
demonstrating its potential for
efficient early AD detection.
An ensemble approach using CNN,
Amalgamation of deep
RNN, LSTM models, and bagging
learning models (CNN,
for precise dementia level
RNN, long short-term Accuracy, sensitivity, specificity, and
[168] MRI determination in AD reported
memory (LSTM)) using precision.
92.22% accuracy, i.e., notable
ensemble and bagging
enhanced diagnostic accuracy on
approaches.
the OASIS Brain dataset.
The proposed technique in The performance of the proposed model
A 12-layer CNN model achieved
this paper is a 12-layer CNN was evaluated using various metrics,
97.75% accuracy, outperforming
[169] MRI model for the binary including accuracy, precision, recall,
existing models on the OASIS
classification and detection F1-score, and the receiver operating
dataset for brain MRI data.
of AD. characteristic (ROC) curve analysis.
CNN framework for AD diagnosis
Unified CNN framework
using sMRI data was proposed with
combining 3D CNN and 3D
[170] sMRI Accuracy for AD detection. a reported accuracy of an
convolutional long
impressive 94.19% for AD detection
short-term memory.
on the ADNI dataset.
CNNs were employed to
classify AD. The authors
CNN models were evaluated using An open-source framework for AD
compared different CNN
T1-weighted accuracy, sensitivity, specificity, and AUC. classification ensuring
[171] architectures, including 2D
MRI Rigorous validation and data integrity reproducibility, transparency, and
slice-level, 3D patch-level,
were ensured. improved evaluation procedures.
ROI-based, and 3D
subject-level approaches.
The study improved 3D
CNNs for early AD
detection. Techniques
Model provided insights for
explored include instance Improved CNN models were evaluated on
improving 3D CNN models in AD
normalization instead of the ADNI dataset, showing a 14% accuracy
detection, i.e., effectiveness of
[172] sMRI batch normalization, increase over existing models. Similar
normalization, early downsampling,
avoiding early spatial performance was observed on an
and model widening were
downsampling, widening independent dataset.
investigated.
the model, and
incorporating age
information.
Multi-modal deep learning
framework for joint
hippocampal segmentation The proposed method achieved 87.0% dice
A multi-modal deep learning (DL)
and AD classification using similarity for hippocampal segmentation
framework for early-stage AD
structural MRI data. It and 88.9% accuracy, 92.5% AUC for AD vs.
[173] sMRI diagnosis, outperforming
includes a multi-task CNN NC classification, and 76.2% accuracy,
single-model methods and
model for segmentation and 77.5% AUC for MCI vs. NC classification.
competitors in AD.
classification, along with a It outperformed other methods.
3D DenseNet model for
disease classification.
Mach. Learn. Knowl. Extr. 2024, 6 476

Table 1. Cont.

Reference Input Data Technique Evaluation Notes

CNN-EL approach combines
CNNs and ensemble The ensemble’s performance was Proposed CNN-EL method resulted
learning for AD evaluated using fivefold cross-validation in improved AD and MCI
classification using MRI for AD vs. HC, MCIc vs. HC, and MCIc vs. classification using MRI data.
[174] MRI
slices, identifying brain MCInc classification. Brain regions Moreover, this method also
regions contributing to contributing to classification were identified key brain regions
classification based on identified based on intersection points. associated with early AD.
intersection points.
Proposed methodology:
An efficient method for
2D-ACF for noise reduction, The proposed method outperformed
segmentation of AD-related regions
[175] MRI EP-CI for image existing approaches in segmentation
in MR brain images, leading to
enhancement, and EFCMAT quality.
improved diagnostic performance.
for AD region segmentation.
The study improved
Alzheimer’s disease
detection accuracy using Proposed methodology evaluated using Effective detection of Alzheimer’s
[176] 3D MRI dimensionality reduction accuracy 93% and other metrics: confusion using random forest and CNN with
methods (PCA, RP, FA) and matrix, precision, recall, and F1-score. proposed RF model’s 93% accuracy.
applied RF and CNN with
reduced features as inputs.
The proposed framework achieved high CNN-based framework reported
CNN-based framework for
classification accuracies of 99.6%, 99.8%, excellent accuracy in AD
AD classification utilized
[177] MRI and 97.8% for binary AD vs. CN classification using brain MRI scans,
deep learning’s advantages
classification and 97.5% for showing potential for early AD
over traditional methods.
multi-classification on the ADNI dataset. diagnosis.
Accuracy of the enhanced network AlzNet, a CNN trained on 2D MRI
Convolutional neural (Alzheimer Network—AlzNet) for slices from the OASIS dataset,
[178] 2D MRI
network (CNN). discriminating between Alzheimer’s reported 99.30% accuracy in AD
patients and healthy patients. recognition.
Converting 3D images to 2D The experimental results show 99.95%
Enhanced AD diagnosis through
and using VGG-16 CNN for accuracy for fMRI classification and 73.46%
preprocessing, CNN models, and
[179] fMRI, PET feature extraction. Various for PET. Compared to existing methods, it
diverse classifiers surpasses prior
classifiers were employed exhibited superior performance in various
methods.
for image classification. parameters.
Ensemble model
architecture using 2D CNNs
selects the top 11 coronal
slices, trains VGG16, Ensemble learning architecture
ResNet50, GAN Proposed approach evaluated for AD vs. reported high accuracy for AD
[180] sMRI discriminator models, CN, AD vs. MCI, and MCI vs. CN classification for limited data, which
majority voting for accuracy. has been a problem for conventional
multi-slice decisions, deep learning models.
ensemble model, and
transfer learning for domain
adaptation.
The proposed method
combines CNN and DNN
models for hippocampal
localization and
classification. Proposed approach achieves high accuracy Proposed hybrid method reported
Three-dimensional patches for left and right hippocampi: 94.82% and high accuracy in Alzheimer’s
[181] sMRI
were extracted, and 94.02%, respectively, with AUC values of diagnosis by combining CNN and
two-dimensional slices were 92.54% and 90.62%. DNN localized positions.
obtained from them.
Volumetric features were
extracted using DVE-CNN
for classification.
CNN for Alzheimer’s
Evaluated on accuracy, sensitivity, Proposed CNN system improved
prediction from brain MRI
specificity, and AUC for Alzheimer’s early Alzheimer’s detection,
[182] MRI scans. Extracts
prediction. Outperforms existing methods ensured timely interventions, and
disease-related features for
in diagnostic accuracy. reduced false negatives.
accurate diagnosis.
Mach. Learn. Knowl. Extr. 2024, 6 477

Table 1. Cont.

Reference Input Data Technique Evaluation Notes

CNN with building
components for AD CNN model achieved 97.8%
The proposed CNN-based method was
classification. It extracted accuracy in Alzheimer’s disease
[183] MRI evaluated using accuracy for disease
essential features from MRI detection from brain MRI images
classification.
images, aiding disease using automated feature extraction.
classification.
Mp-CNN utilized three 2D
Method achieved 93% accuracy for MP-CNN outperformed the
CNNs to analyze
multiclass AD-MCI-NC classification, with single-plane approach, offering an
[184] MRI discriminatory information
precision rates of 93% for AD, 91% for MCI, effective method for early AD
from multiple planes in
and 95% for NC subjects detection in 3D images.
3D-MRI.
Pre-trained CNNs
(DenseNet196, VGG16, and
Model employed pre-trained CNNs
ResNet50) were used for
for feature extraction and a stacking
feature extraction from MRI The proposed model achieved 89%
[185] MRI ensemble to classify Alzheimer’s
images. The stacking accuracy on brain MRI data.
disease stages with impressive 89%
ensemble method was
accuracy.
employed for multi-class
AD stage classification.
Evaluated using accuracy, sensitivity, and
specificity metrics. For ADNI-2 MRI A 3D CNN with a focus on the
volumes, it achieved an accuracy of 88.06%, temporal lobe achieved high
[186] MRI 3D CNN.
sensitivity of 94.03%, and specificity of performance in AD classification
82.09% in classifying AD from normal from 3D MRI volumes.
controls.
CNN, specifically MobileNet
pre-trained model, for early Model achieved an accuracy of 96.6% for MobileNet-based framework
AD prediction and multi-class AD stage classifications. enabled precise AD progression, i.e.,
[187] MRI classification. Transfer Comparison with VGG16 and ResNet50 stage-classification, which greatly
learning was applied to models was performed on the same contributed towards early detection
leverage pre-trained models dataset. and classification.
for health data classification.
Combined CNN and KNN The performance of the CNNKNN
CNN-KNN integrated framework
for AD detection. CNN framework was evaluated using accuracy,
with 99.58% accuracy in AD
[188] MRI extracted features from MRI precision, recall, F1-score, MCC, CKC, ROC
detection surpassed existing deep
images, used to train and curves, and stratified K-fold
CNN models for clinical diagnosis.
validate the KNN model. cross-validation.
FFNN and various feature
Methodologies were evaluated using
extraction methods, such as Combination of DL model with
accuracy, sensitivity, AUC, precision, and
GoogLeNet, DenseNet-121, exclusive feature extraction
[189] MRI specificity to measure their effectiveness in
PCA, DWT, LBP, and GLCM, improved AD detection to
detecting AD and predicting disease
for classifying MRI images promising 99.7% accuracy.
progression stages.
as AD or non-AD.
CNN and GAN for AD and
MCI diagnosis. The GAN Combining CNN and cGAN, the
generated additional hybrid model efficiently diagnosed
The classification accuracy was evaluated
[190] MRI training instances, AD and MCI and reported
using Keras.
improving accuracy. CNN improved accuracy on the ADNI
extracted brain features dataset.
from 2D images.
The proposed method used Proposed CNN analyzed MRI scans
a 12-layer CNN for early AD for early AD detection with a
The model was evaluated based on its
identification from brain reported accuracy rate of 97.80%,
[191] MRI accuracy in detecting AD. The accuracy of
MRI scans, taking advantage emphasizing the importance of
the model was 97.80%.
of CNNs’ effectiveness in timely diagnosis for both mental
image processing tasks. and physical health in AD patients.
Deep learning framework
with multi-task learning for
MTDL model evaluation: accuracy of
hippocampus segmentation The proposed method improved
97.1% and Dice coefficient of 93.5%. For
and AD classification. AD detection accuracy using
binary classification (AD vs. non-AD),
[192] sMRI Capsule network CNN hippocampus segmentation and
there was an accuracy of 96%, and for
model with optimized AD categorization for
multi-class classification (AD stages), there
hyperparameters using deer ADNI datasets.
was an accuracy of 93%.
hunting optimization
(DHO).
Mach. Learn. Knowl. Extr. 2024, 6 478

Convolutional neural networks (CNNs) have demonstrated remarkable achievements

in tasks such as organ segmentation [193–195] and disease detection [196–198] within the
field of medical imaging. By leveraging neuroimaging data, these models can uncover hid-
den representations, establish connections between different components of an image, and
identify patterns related to diseases [199]. They have been successfully applied to diverse
medical imaging modalities, encompassing structural MRI [200], functional MRI [201,202]
(fMRI), PET [203,204], and diffusion tensor imaging (DTI) [205–207]. Consequently, re-
searchers have begun exploring the potential of deep learning models in detecting AD
using medical images [61,208–210]. In recent years, CNNs in Alzheimer’s disease (AD)
research have garnered substantial interest. This discussion provides a condensed overview
of the diverse applications of CNNs in AD analysis, offering insights for future exploration.

5.1.8. Performance Comparison

When assessing the performance of different CNN-based approaches in Alzheimer’s
disease (AD) research, their remarkable potential becomes evident. Traditional CNN
architectures, such as LeNet-5 and AlexNet, offer robust feature extraction capabilities,
enabling them to capture intricate AD-related patterns. However, it is essential to note that
LeNet-5′ s relatively shallow architecture may limit its ability to discern complex features,
while the high parameter count in AlexNet can lead to overfitting concerns.
Transfer learning, a strategy where pre-trained models like VGG16 are fine-tuned
for AD detection, has emerged as a highly effective approach. By leveraging the insights
gained from extensive image datasets, transfer learning significantly enhances AD detec-
tion accuracy.
The introduction of 3D CNNs has further expanded the capabilities of CNN-based
methods, particularly in the analysis of volumetric data, such as MRI scans. These models
excel at learning nuanced features, a critical advantage given the temporal progression
of AD.
In terms of performance evaluation, CNN-based methods are typically assessed using
various metrics, including accuracy, sensitivity, specificity, precision, and the F1-score.
While these metrics effectively gauge performance, interpretability remains a challenge.
Nevertheless, ongoing efforts, such as attention mechanisms and visualization tools, aim to
address this issue.
Despite their promise, CNNs face limitations, primarily related to data availability. To
ensure the generalization of CNN-based AD detection models across diverse populations,
acquiring and curating large, representative datasets remains a priority for future research.
In summary, CNN-based methodologies have demonstrated their mettle in AD research,
showcasing strengths across traditional and 3D architectures, transfer learning, and ongoing
interpretability enhancements. To realize their full potential for real-world clinical applications,
addressing data limitations and improving generalization are critical objectives.

5.1.9. Meaningful Insights

The application of convolutional neural networks (CNNs) in Alzheimer’s disease
(AD) detection has unveiled several meaningful insights. CNNs, particularly 3D architec-
tures, have showcased their prowess in deciphering complex patterns within volumetric
neuroimaging data.
One remarkable insight is the ability of CNNs to extract hierarchical features from
brain images. Traditional CNN architectures, like LeNet-5 and AlexNet, excel in capturing
intricate structural information but may struggle with deeper, more abstract features. In
contrast, transfer learning, where pre-trained models are fine-tuned for AD detection, has
proven highly effective. This approach capitalizes on the wealth of knowledge acquired
from diverse image datasets, offering a robust foundation for AD-related feature extraction.
The introduction of 3D CNNs has further illuminated the importance of spatial context in
AD diagnosis. These models excel in capturing nuanced patterns across multiple image
slices, aligning with the progressive nature of AD.
Mach. Learn. Knowl. Extr. 2024, 6 479

Performance metrics, including accuracy, sensitivity, specificity, and precision, have

substantiated CNN’s effectiveness. These metrics provide quantitative evidence of CNNs’
diagnostic capabilities. Additionally, ongoing efforts in developing attention mechanisms
and visualization tools aim to enhance model interpretability.
However, the ultimate insight gleaned from CNN-based AD detection is the need for
substantial data. Generalizability across diverse populations demands large, representative
datasets. This challenge underscores the importance of data acquisition and curation efforts.
In conclusion, CNNs have illuminated the path towards more accurate, data-driven
AD detection. Leveraging hierarchical feature extraction, embracing 3D architectures, and
ensuring interpretability are pivotal in harnessing CNNs’ potential for earlier and more
reliable AD diagnosis.

5.2. Recurrent Neural Networks (RNN) for AD Detection

Recurrent neural networks (RNNs) have acquired considerable attention in medical
imaging for AD detection. These deep learning models are well-suited for capturing
temporal dependencies and sequential patterns in data, making them particularly useful
for analyzing time series or sequential data in AD detection tasks. In recent years, RNNs
have shown promising results in capturing complex relationships within longitudinal
neuroimaging data and aiding in the early diagnosis of AD.
In this section, we present a systematic review focused on the application of recurrent
neural networks for AD detection from 2018 to 2023. Our objective is to provide a com-
prehensive overview of the existing literature and summarize the key findings regarding
the use of RNNs in this domain. By examining the different studies’ approaches, input
types/modalities, techniques, evaluation metrics, and additional notes, we aim to highlight
the potential of recurrent neural networks as a powerful tool for AD detection.

5.2.1. Long Short-Term Memory (LSTM) Networks

LSTM is a type of RNN that can effectively capture long-term dependencies in sequen-
tial data. Several studies have employed LSTM networks for AD diagnosis and prediction.
These models typically take sequential data, such as time-series measurements from brain
imaging or cognitive assessments, as input, and learn temporal patterns to classify or
predict AD progression. LSTM-based models have demonstrated promising results in
accurately diagnosing AD and predicting cognitive decline.

5.2.2. Encoder–Decoder Architectures

Encoder–decoder architectures, often combined with attention mechanisms, have been
used in AD research to address tasks such as predicting disease progression or generating
informative features. These models encode input sequences into latent representations and
decode them to generate predictions or reconstructed sequences. Encoder–decoder archi-
tectures with attention mechanisms allow for the network to focus on relevant temporal
information, improving prediction accuracy and interpretability.

5.2.3. Hybrid Models

Some studies have combined RNNs with other deep learning architectures, such as
convolutional neural networks (CNNs) or generative adversarial networks (GANs), to
leverage their respective strengths. These hybrid models aim to capture both spatial and
temporal information from brain imaging data, leading to improved performance in AD
diagnosis, progression prediction, or generating synthetic data for augmentation.

5.2.4. Overview of Recurrent Neural Network (RNN) Studies for AD Detection

Table 2 summarizes major studies that have utilized RNNs for AD detection. It
provides relevant information such as the study name, date, input type/modalities, tech-
nique, evaluation metric, and additional notes. This table serves as a quick reference for
researchers and practitioners interested in RNN-based approaches for AD detection.
Mach. Learn. Knowl. Extr. 2024, 6 480

Table 2. Overview of RNN studies for AD detection.

Reference Input Data Technique Evaluation Notes

The proposed method reported
Combination MLP and RNN
89.7% accuracy for AD classification
T1-weighted for spatial and longitudinal
[211] Classification accuracy. by utilizing T1-weighted sMR
sMRI feature extraction,
images demonstrating the potential
respectively.
for longitudinal AD diagnosis.
Integration of 3D-CNN and
Combining 3D CNN and SBiRNN was reported for AD
Average accuracy for AD vs. normal
stacked bidirectional diagnosis. Accordingly, MRI and
[212] MRI, PET classification (NC), pMCI vs. NC, and
recurrent neural network PET modalities demonstrated
MCI vs. NC.
(SBi-RNN). improvements over the
ADNI dataset.
AD progression using RNN-LSTM
Multivariate time and fully connected neural
series data from networks with reported 88.24%
Performance comparison with baseline
the Alzheimer’s RNN with strategies to accuracy on the ADNI dataset. The
[213] models in multi-step prediction of
Disease handle missing data. proposed method improved AD
Alzheimer’s disease progression.
Neuroimaging progression, i.e., stage prediction
Initiative (ADNI) using RNN with strategies for
handling missing data.
Heterogeneous
medical data of Enhanced RNN tracks AD
Long short-term memory
5432 patients with Accuracy comparison with classic progression with >99% accuracy,
[214] recurrent neural networks
probable baseline methods. showcasing potential for chronic
(RNN).
Alzheimer’s disease progression prediction.
disease (AD)
Proposed 2D CNN-RNN framework
achieved high AUC values without
Combination of 2D CNN AUC for AD vs. NC classification and
[215] FDG-PET requiring image registration or
and RNNs. MCI vs. NC classification.
segmentation in the
preprocessing stage.
Combination of CNN and The proposed method attained a
RNN for spatial and classification accuracy of 91.33% for
[216] sMRI Classification accuracy.
longitudinal feature AD vs. NC and 71.71% for pMCI vs.
extraction, respectively. sMCI, indicating promise.
Hybrid convolutional and Combination of CNN and RNN for
recurrent neural network AD diagnosis using MR images of
[217] sMRI using DenseNets and Area under ROC curve (AUC). the hippocampus reported
bidirectional gated recurrent promising results (AUCs: 91.0%,
units (BGRU). 75.8%, and 74.6%).
The proposed RNN model to
Classification accuracy, identifying identify AD risk using diffusion
Diffusion tensor Recurrent neural network
[218] individuals with early mild cognitive tensor imaging (DTI) data achieved
imaging (DTI) (RNN) model.
impairment (EMCI). promising results and high accuracy
in predictions.
DenseNet and LSTM integration for
Combination of pre-trained Performance comparison with
precise AD classification resulted in
[219] MRI DenseNet with long state-of-the-art deep learning methods
improvement over state-of-the-art
short-term memory (LSTM). using 5-fold cross-validation.
methods on the OASIS dataset.
LSTM-based model was proposed
Long short-term memory
Comparison of the predictive to predict AD progression, which
(LSTM) network with fully
[220] MRI, PET, DTI performance of the proposed LSTM demonstrated impressive
connected and
model with existing models. performance towards AD prediction
activation layers.
in MRI/PET data.
The study reported average accuracies
DL framework was proposed for
Combining 3D CNNs and for AD vs. NC, pMCI vs. NC, and MCI
AD diagnosis using 3D-CNN and
fully stacked bidirectional vs. NC classification tasks. The
[221] MRI, PET FSBi-LSTM, which reported a higher
long short-term memory proposed method was compared with
classification rate w.r.t conventional
(FSBi-LSTM). existing algorithms to assess its
algorithms.
performance.
Bidirectional LSTM-based AD
Bidirectional long
Prediction of AD development and prediction was proposed using
[222] MRI short-term memory (LSTM)
classification into NL, MCI, and AD. neuro-psychological, genetic, and
with attention mechanism.
tomographic data.
Mach. Learn. Knowl. Extr. 2024, 6 481

Table 2. Cont.

Reference Input Data Technique Evaluation Notes

LSTM model to handle missing
LSTM networks with a
The study evaluates MAE for values (integration of RNN) was
generalized training rule for
[223] MRI predicting MRI biomarkers and AUC proposed leading to improved
handling missing predictor
for clinical AD diagnosis. prediction of MRI biomarkers and
and target values.
AD diagnosis.
RNN-based model for
Alzheimer’s disease
An RNN-based model tracking AD
progression prediction using
This model was evaluated for its progression in MCI individuals was
cognitive measures and MRI
[224] MRI accuracy in predicting AD progression proposed. Accordingly, cognitive
scans. RNN captures
early in individuals with MCI. data and baseline MRI scans were
temporal patterns for
investigated.
accurate prognostic
predictions.
RNN models (LSTM and GRU)
Accuracy, F-score, sensitivity, and
RNNs with long short-term outperformed conventional
specificity for classification. Low
memory (LSTM) and gated classification methods (SVM, SVR,
[225] MRI, PET RMSE, high correlation coefficient for
recurrent unit (GRU) and ridge regression) for
regression;. Outperforms SVM, SVR,
architectures. multimodal image data towards
and ridge regression models.
classification.
Proposed RNN model predicts AD
Predicting diagnosis, cognition, and
diagnosis, cognition, and ventricular
ventricular volume. Compared to
[226] MRI, PET Minimal RNN model. volume effectively using MRI/PET
baseline algorithms and handling
data as demonstrated in the
missing data.
TADPOLE challenge.
Deep recurrent network for
joint prediction of missing Performance measured using various
Using deep RNN, missing data were
values, phenotypic metrics, comparison with competing
[227] MRI handled to improve AD predictions
measurements, trajectory methods in the literature, exhaustive
in the TADPOLE challenge cohort.
estimation of cognitive analyses, and ablation studies.
scores, and clinical status.
The study employed multi-task
Longitudinal data
learning for tracking Alzheimer’s
(MRI volumetric Multi-task learning
Improvement in mAUC, BCA, and disease and resulted in improved
[228] measurements, framework with adaptive
MAE (ADAS-Cog13 and ventricles). performance towards gmAUC, BCA,
cognitive score, imputation and prediction.
and MAE (ADAS-Cog13
clinical status)
and ventricles).
Proposed framework predicted AD
Neuropsychologica
RNN with LSTM and fully progression using RNN-LSTM and
Rlecurrent
[229] connected neural network Accuracy of 88.24%. fully connected neural networks
measures and
layers. with reported 88.24% accuracy on
MRI biomarkers
the ADNI dataset.
CRNN for brain disease
classification using rs-fMRI CRNN method was evaluated on
(rs-fMRI) data, Using rs-fMRI data and dFC, the
data. Sliding window 174 subjects with 563 rs-fMRI scans for
specifically dFC proposed method automated brain
strategy, convolutional and binary and multicategory classification
[230] networks derived disease classification but it required
LSTM layers for feature tasks. The study demonstrated its
from the generalization on a larger dataset
extraction and temporal effectiveness in accurately classifying
rs-fMRI data. due to limited samples.
dynamics, fully connected brain diseases.
layers for classification.
Proposed GP-ELM-RNN network
GP-ELM-RNN network (a
achieves an accuracy (around
combination of genetic
Accuracy, specificity, and comparison 99.23%) in classifying AD stages
[231] CT programming, extreme
with ELM and RNN models. with CT brain scans, but validation
learning machines, and
is required over a larger dataset
recurrent neural networks).
for generalization.

Recurrent neural networks (RNNs) have emerged as a popular deep learning technique
for analyzing temporal data, making them well-suited for Alzheimer’s disease research.
This discussion section will highlight the various methods that have utilized RNNs in
AD research, provide an overview of their approaches, compare their performance, and
present meaningful insights for further discussion.

5.2.5. Performance Comparison

Comparing the performance of different RNN-based methods in AD research can
be challenging due to variations in datasets, evaluation metrics, and experimental setups.
Mach. Learn. Knowl. Extr. 2024, 6 482

However, several studies have reported high accuracy, sensitivity, and specificity in AD di-
agnosis and prediction tasks using RNNs. For example, LSTM-based models have achieved
accuracies ranging from 80% to over 90% in AD classification. TCNs have demonstrated
competitive performance in predicting cognitive decline, with high AUC scores. Encoder–
decoder architectures with attention mechanisms have shown improvements in disease
progression prediction compared to traditional LSTM models. Hybrid models combining
RNNs with other architectures have reported enhanced performance by leveraging spatial
and temporal information.

5.2.6. Meaningful Insights

RNNs, such as LSTMs, are well-suited for capturing long-term dependencies in se-
quential data. In the context of AD research, this capability allows for identifying subtle
temporal patterns and predicting disease progression. By analyzing longitudinal data,
RNNs can potentially detect early signs of cognitive decline and facilitate early intervention
strategies.
RNNs can also be effectively used for data augmentation in AD research. Synthetic
sequences can be generated using generative models, such as variational autoencoders
(VAEs) or GANs, to increase the diversity and size of the training dataset. This augmented
data can enhance the robustness and generalizability of RNN models, leading to improved
diagnostic accuracy and generalization to unseen data.
In addition, RNNs offer interpretability and explainability in AD research. By ana-
lyzing the temporal patterns learned by the models, researchers can gain insights into the
underlying disease progression mechanisms. This information can aid in understanding
the neurobiological processes associated with AD and provide valuable clues for potential
therapeutic interventions.
Moreover, RNNs can handle multimodal data sources, such as combining brain imag-
ing (e.g., MRI, PET scans) with clinical assessments or genetic information. Integrating
multiple modalities can provide a more comprehensive understanding of AD, capturing
both structural and functional changes in the brain along with clinical markers. RNN-based
models enable the fusion of diverse data sources to improve the accuracy and reliability of
AD diagnosis and prognosis.
RNNs trained on large-scale datasets can learn robust representations that generalize
well to unseen data. Pre-training RNN models on large cohorts or external datasets and
finetuning them on specific AD datasets can facilitate knowledge transfer and enhance the
performance of AD classification and prediction tasks. Transfer learning approaches enable
the utilization of existing knowledge and leverage the expertise gained from related tasks
or domains.
While RNNs have shown promise in AD research, there are still challenges to address.
One major challenge is the limited availability of large-scale, longitudinal AD datasets.
Acquiring and curating diverse datasets with longitudinal follow-up is crucial for training
RNN models effectively. Additionally, incorporating uncertainty estimation and quan-
tifying model confidence in predictions can further enhance the reliability and clinical
applicability of RNN-based methods.
Furthermore, exploring the combination of RNNs with other advanced techniques,
such as attention mechanisms, graph neural networks, or reinforcement learning, holds
promise for improving AD diagnosis, understanding disease progression, and guiding
personalized treatment strategies. Integrating multimodal data sources, such as imaging,
genetics, and omics data, can provide a more comprehensive view of AD pathophysiology.
In conclusion, RNN-based approaches have emerged as powerful tools for AD re-
search, enabling accurate diagnosis, prediction of disease progression, and data augmen-
tation. Various RNN architectures, such as LSTMs, TCNs, and encoder–decoder models,
have been applied to different AD tasks with notable success. These models showcase
the ability to capture long-term temporal dependencies, enhance interpretability, and
integrate multimodal data sources. Nonetheless, further advancements are needed to
Mach. Learn. Knowl. Extr. 2024, 6 483

address challenges related to data availability, uncertainty estimation, and the integration
of cutting-edge techniques. By continuing to explore and refine RNN-based methods, we
can pave the way for improved understanding, early diagnosis, and personalized treatment
of Alzheimer’s disease.

5.3. Generative Modeling for AD Detection

Generative modelling techniques have gained attention in medical imaging for Alzheimer’s
disease (AD) detection. These models are capable of generating new samples that follow the
distribution of the training data, enabling them to capture the underlying patterns and variations
in AD-related imaging data. By leveraging generative models, researchers aim to enhance early
detection, improve classification accuracy, and gain insights into the underlying mechanisms
of AD.
In this section, we present a systematic review focused on the application of generative
modelling techniques for AD detection from 2018 to 2023. Our objective is to provide an
overview of the existing literature and summarize the key findings regarding the use of
generative models in this domain. By examining the different studies’ approaches, input
types/modalities, techniques, evaluation metrics, and additional notes, we aim to highlight
the potential of generative modelling as a valuable tool for AD detection.

5.3.1. GANs for Image Generation

One prominent application of generative modelling in Alzheimer’s disease is the
generation of synthetic brain images for diagnostic and research purposes. GANs have
been used to generate realistic brain images that mimic the characteristics of Alzheimer’s
disease, such as the presence of amyloid beta plaques and neurofibrillary tangles. These
synthetic images can be valuable for augmenting datasets, addressing data scarcity issues,
and improving classification performance.

5.3.2. Conditional GANs for Disease Progression Modeling

Conditional GANs (cGANs) have been employed to model the progression of Alzheimer’s
disease over time. By conditioning the generator on longitudinal data, cGANs can generate
synthetic brain images that capture disease progression stages, ranging from normal to mild
cognitive impairment (MCI) and finally to Alzheimer’s disease. This enables the generation of
realistic images representing the transition from healthy to pathological brain states.

5.3.3. Variational Autoencoders (VAEs) for Feature Extraction

In addition to GANs, variational autoencoders (VAEs) have been utilized to extract
informative features from brain images for Alzheimer’s disease classification. VAEs can
learn a compressed representation of the input images, known as latent space, which
captures relevant features associated with the disease. By sampling from the latent space,
new images can be generated, and the extracted features can be used for classification tasks.

5.3.4. Hybrid Approaches

Some studies have explored hybrid approaches that combine different generative
models to leverage their respective advantages. For example, combining GANs and VAEs
can harness the generative power of GANs while benefiting from the probabilistic nature
and interpretability of VAEs. These hybrid models aim to generate high-quality images
while preserving the meaningful representations learned by VAEs.

5.3.5. Overview of Generative Modeling Studies (GAN) for AD Detection

Table 3 provides a summary of major studies that have utilized generative modelling
techniques for AD detection. It includes relevant information such as the study name, date,
input type/modalities, technique, evaluation metric, and additional notes. This table serves
as a quick reference for researchers and practitioners interested in exploring generative
modelling approaches for AD detection.
Mach. Learn. Knowl. Extr. 2024, 6 484

Table 3. Overview of generative modeling studies for AD detection.

Reference Input Data Technique Evaluation Notes

Two-stage deep learning for
AD diagnosis using MRI and
PET data. Stage 1: Impute The proposed method used
Quality of synthesized PET images was
missing PET data from MRI multi-instance NN for AD diagnosis
assessed using 3D-cGAN. The performance
using 3D-cGAN. Stage 2: for improved AD diagnosis by
[232] MRI, PET of the two-stage deep learning framework
Use a deep multi-instance addressing missing data issues in
was compared to state-of-the-art methods
neural network for AD stage 1, i.e., impute PET data
in AD diagnosis.
diagnosis and MCI from MRI.
conversion prediction with
complete MRI and PET data.
GANs model AD
progressionin MR images.
The GAN-based approach was evaluated
Synthetic images with
through experiments and comparisons GANs were trained on synthetic
varying AD features were
with oserved changes in AD-like features. images to learn AD features and,
generated. Image arithmetic
[233] MRI The modified GAN training was assessed subsequently, a modified GAN was
manipulated AD-like
for encoding and reconstructing real deployed to make predictions on
features in specific brain
images with high atrophy and unusual MR images.
regions. A modified GAN
features.
training handled extreme
AD cases.
Wasserstein GANs to
artificially age individual
The brain ageing model was evaluated on Method utilized Wasserstein GANs
brain images. A novel
healthy and demented subjects, predicting to assess age from brain images and
[234] MRI recursive generator model
conversion from MCI to AD using GAN predict individual brain ageing and
was developed to generate
and pre-trained CNN classifier. MCI to AD conversion.
brain image time series
based on longitudinal data.
Disease progression
prediction framework: Performance measured using SSIM to Future brain MRI generated by
3D mi-GAN generates evaluate the quality of generated MRI GAN and, subsequently,
[235] 3D sMRI future brain MRI images, images and accuracy improvement for classification of AD stage using
3D DenseNet-classifier differentiating between pMCI and sMCI mi-GAN with focal loss
predicts clinical stage using stages. optimization.
focal loss.
GAN to reconstruct missing GAN-based augmentation method
Densely connected model evaluated on
PET images. A densely to address missing PET data
ADNI dataset. Reconstructed images
connected convolutional improved classification model
[236] PET improved classification for
network is then developed performance on imbalanced
class-imbalanced data. Noisy dimensions’
as the classification model datasets, as demonstrated on the
influence was assessed using metrics.
for binary classification. ADNI dataset.
GAN data augmentation for Combination of multi-scale MRI
accurate differential Proposed framework evaluated with features, GAN augmentation, and
[237] sMRI diagnosis between normal 10-fold cross-validation on 1954 images ensemble classifier led to high
ageing, AD, and FTD using achieved 88.28% accuracy. classification accuracy for normal
multi-scale MRI features. ageing, AD, and FTD samples.
Innovative approach: GAN High-performance evaluation:
GAN integration with AD diagnosis
for PET synthesis with AD state-of-the-art results in three- and
for PET image synthesis, leading to
[238] MRI, PET diagnosis integration. four-class AD classification tasks using
state-of-the-art AD classification
Fine-tuned architecture for synthesized PET images. Effective AD
results.
optimized AD classification. diagnosis demonstrated.
GAN called BEGAN for slice
Proposed SVM classifier (based on
selective learning to address
The model was evaluated on the severance GAN features) reported a good
PET imaging environment
[18F] FDG and ADNI datasets using accuracy, performance on the ADNI dataset
[239] differences. The extracted
PET, CT sensitivity, and specificity metrics, and the for AD and NC classification, i.e.,
unbiased features are used
results were statistically compared. less sensitive to acquisition
to train an SVM classifier for
conditions.
AD and NC classification.
Mach. Learn. Knowl. Extr. 2024, 6 485

Table 3. Cont.

Reference Input Data Technique Evaluation Notes

cGAN architecture
synthesizes MRI at various
AD stages using a 2D
generator and 2D/3D
It was evaluated by generating synthetic
discriminators to assess GAN-based image synthesis for AD
3D MR images at different conditions and
[240] MRI image realism. The evolution (at different conditions)
comparing their quality with those
optimization process by evaluation of 2D/3D losses.
generated by 2D or 3D cGANs.
involves both 2D and 3D
GAN losses for evaluating
consecutive 2D images in
3D space.
GAN-based framework enhanced
GAN model generated It was evaluated using SNR, BRISQUE, and
AD classification using synthetic
T1-weighted synthetic 1.5T MRI images NIQE. The classification model’s
[241] 1.5T MRI images, leading to
MRI used by the FCN for AD performance was measured using AUC on
improved performance across
status prediction. various datasets.
multiple datasets.
CNN and GANs for AD
classification using
neuroimaging data.
CNNs’ potential for AD
Three-dimensional CNNs
classification using neuroimaging
handle multimodal AD classification performance was
was tested. Ensemble learning,
PET/MRI data, while GANs evaluated using neuroimaging data and
[242] MRI, PET including PET/MRI and GANs, was
address limited data by metrics like accuracy, sensitivity, specificity,
used to show effectiveness towards
generating synthetic and AUC.
early detection and disease
samples. EL enhances model
understanding.
robustness and classification
performance by combining
multiple models.
GAN-generated brain PET images
GAN-based method generated
Deep GANs used for evaluated using classification model (72%
Alzheimer’s disease images from
[243] PET synthesizing brain PET accuracy) for AlD stages. Quality was
limited data, promising improved
images across AD stages. measured with PSNR (avg. 82, 72, 73) and
diagnosis model accuracy.
SSIM (avg. 25.6, 22.6, 22.8) scores.
The model’s performance was assessed by Proposed method used GAN-based
GAN to harmonize the MRI comparing AD classification accuracy harmonized MR images for
[244] MRI
images. using harmonized MR images and original computing AD classification
non-harmonized datasets. performance w.r.t original dataset.
GAN-based method for AD
GAN-based deep learning GAN-based deep learning methods were
classification outperformed
methods utilized for AD evaluated using accuracy, odds ratios
[245] MRI, PET non-GAN methods, but
classification and compared (ORs), pooled sensitivity, pooled specificity,
improvement is required for
with non-GAN methods. and AUC in a meta-analysis.
differentiating pMCI vs. sMCI.
3D end-to-end generative
BPGAN generated high-quality PET
adversarial network
images from MRI scans, enhancing
T1-weighted (BPGAN) that learns a The performance of BPGAN was evaluated
[246] AD diagnosis accuracy in
sMRI, PET mapping function to using MAE, PSNR, SSIM.
multi-modal medical image
generate PET scans
analysis.
from MRI.
GAN-based approach for
Pairwise feature-based GAN model
AD diagnosis generates PET The effectiveness of the proposed method
for AD diagnosis, using the
features from brain images was evaluated through extensive
[247] MRI, PET attention mechanics model
using attention mechanisms experiments, demonstrating promising
generated PET features from MRI to
for structural information results in the diagnosis of AD.
diagnose AD.
retention.
The proposed approach is
DCGAN-based unsupervised
based on an unsupervised
learning for AD diagnosis using
deep learning model using a The model achieved an AUROC of 0.7951,
sMRI images. Method showed
[248] sMRI deep convolutional precision of 0.8228, recall of 0.7386, and
promising results to discriminate
generative adversarial accuracy of 74.44% for AD diagnosis.
AD and non-AD cases with
network (DCGAN) using
accuracy of 75%.
brain MRIs without labels.
Proposed multimodal
Deep multimodal fusion, including
generative data fusion
Proposed model excelled in AD vs. healthy neuroimaging and genomics data,
framework addresses
[249] fMRI control classification, handling missing handled missing modalities using
missing modalities using
modalities effectively. GAN for improved AD
GANs for accurate
classification.
predictions.
Mach. Learn. Knowl. Extr. 2024, 6 486

Table 3. Cont.

Reference Input Data Technique Evaluation Notes

HSIA-GAN uses
hypergraph structural
information aggregation, HSIA-GAN model integrated
HSIA-GAN model evaluated in three AD
capturing low-order multi-level information(structural)
neuroimaging classification tasks for
[250] fMRI, SNP relations with vertex and for AD analysis, improving disease
accurate sample classification and feature
edge graphs, and extracting classification with informative
extraction.
structural information using features.
generator and discriminator
components.
The proposed approach uses
Proposed work enhanced AD
an adversarial The proposed approach improves
classifier using adversarial
counterfactual augmentation Alzheimer’s disease classification and
[251] MRI counterfactual augmentation to
scheme to address classifier addresses spurious correlations and
mitigate spurious correlations and
weaknesses by leveraging catastrophic forgetting.
forgetting.
the generative model.
Model: Trained on ADNI NCs, validated GANCMLAE model combined
GANCMLAE, combining on Xuanwu cohort. Evaluation: SSIM, GAN and autoencoder for accurate
GANs and multiple loss PSNR, MSE for image reconstruction; MCI atrophy pattern depiction in AD
[252] sMRI
autoencoder to depict subtype atrophy pattern identification; and MCI, outperforming the t-test
individual atrophy patterns. AUC-ROC for AD and MCI vs. NC model with promising precision in
classification. AD and MCI.
The system uses a
GAN-based DCNN for AD, A system for multi-type dementia
PD, and FTD diagnosis, The model achieved an accuracy of 97.7%, classification using FDG-PET brain
[253] FDG-PET
addressing distribution with sensitivity and specificity both at 0.97. scans with an accuracy of (97.7%) to
issues and handling feature identify AD, FTD, and PD.
learning and classification.

Generative modelling, particularly through approaches like generative adversarial

networks (GANs), has emerged as a promising technique in the field of Alzheimer’s disease
research. This discussion section will provide an overview of the various methods used in
generative modelling for Alzheimer’s disease, compare their strengths and limitations, and
highlight meaningful insights for further exploration and discussion.

5.3.6. Comparative Analysis

When comparing the different generative modelling methods in Alzheimer’s disease
research, several factors should be considered:
• Image Quality: The primary goal of generative modelling is to generate high-quality
brain images that closely resemble real data. GANs have demonstrated remarkable
success in producing visually realistic images, while VAEs tend to produce slightly
blurred images due to the nature of their probabilistic decoding process.
• Feature Extraction: While GANs excel in image generation, VAEs are more suitable
for feature extraction and latent space representation. VAEs can capture meaningful
features that reflect disease progression and provide interpretability, making them
valuable for understanding the underlying mechanisms of Alzheimer’s disease.
• Data Scarcity: Alzheimer’s disease datasets are often limited in size, posing challenges
for training deep learning models. Generative modelling techniques, especially GANs,
can help address data scarcity by generating synthetic samples that augment the
training data and improve model generalization.
• Interpretability: VAEs offer an advantage in terms of interpretability because they
learn a structured latent space that captures meaningful variations in the data. This
can aid in understanding disease patterns and identifying potential biomarkers.

5.3.7. Meaningful Insights

Generative modelling in Alzheimer’s disease research holds great promise for ad-
vancing diagnosis, disease progression modelling, and understanding the underlying
mechanisms of the disease. By generating realistic brain images and capturing disease-
Mach. Learn. Knowl. Extr. 2024, 6 487

related features, these techniques can complement traditional diagnostic methods and
provide new avenues for personalized treatment and intervention strategies.
One meaningful insight from the application of generative modelling is the potential to
address data scarcity issues. Alzheimer’s disease datasets are often limited in size and subject to
variability in imaging protocols and data acquisition. By using generative models like GANs
and VAEs, researchers can generate synthetic data that closely resemble real brain images. This
augmentation of the dataset not only increases the sample size but also captures a wider range
of disease characteristics and progression patterns. Consequently, it enhances the robustness
and generalizability of machine learning models trained on these augmented datasets.
Moreover, generative modelling techniques provide a unique opportunity to simulate
disease progression and explore hypothetical scenarios. By conditioning the generative mod-
els on various disease stages, researchers can generate synthetic brain images that represent
different pathological states, from early stages of mild cognitive impairment to advanced
Alzheimer’s disease. This capability allows for the investigation of disease progression dynam-
ics, identification of critical biomarkers, and evaluation of potential intervention strategies.
Furthermore, the combination of generative models with other deep learning tech-
niques, such as convolutional neural networks (CNNs) or recurrent neural networks
(RNNs), can further enhance the performance of Alzheimer’s disease classification and
prediction tasks. These hybrid models can leverage the strengths of different architectures
and generate more accurate and interpretable results. For example, combining GANs for
image generation with CNNs for feature extraction and classification can lead to improved
diagnostic accuracy and a better understanding of the underlying disease mechanisms.
However, despite the promising results and potential benefits, there are several chal-
lenges and considerations that need to be addressed in future research. Firstly, the inter-
pretability of generative models remains a topic of investigation. While GANs and VAEs
can generate realistic images or extract informative features, understanding the specific
disease-related factors they capture is still an ongoing challenge. Developing methods to
interpret and validate the generated features or images can further enhance their clinical
relevance and utility.
Secondly, the generalizability of the generated synthetic data and models across
different populations, imaging modalities, and data acquisition protocols needs to be
carefully evaluated. It is crucial to ensure that the generated samples accurately represent
the true population distribution and do not introduce biases or artifacts that may limit their
applicability in real-world scenarios.
Lastly, the ethical implications of using generative models in Alzheimer’s disease
research should be considered. The generation of synthetic brain images raises concerns
about privacy, informed consent, and the potential impact on patients’ emotional well-being.
Guidelines and protocols should be established to address these ethical considerations and
ensure the responsible and ethical use of generative modelling techniques.
In conclusion, generative modelling techniques, such as GANs and VAEs, offer promis-
ing avenues for advancing Alzheimer’s disease research. The ability to generate realistic
brain images, model disease progression, and extract meaningful features provides valu-
able insights for diagnosis, prognosis, and treatment planning. By addressing data scarcity,
enhancing interpretability, and combining with other deep learning approaches, generative
modelling can contribute to more accurate and personalized approaches in Alzheimer’s
disease management. However, further research is needed to overcome challenges related
to interpretability, generalizability, and ethical considerations to fully realize the potential
of generative modelling in Alzheimer’s disease research and clinical practice.

6. Trending Technologies in AD Studies

In recent years, there has been a surge of interest in applying deep learning techniques
to Alzheimer’s disease (AD) detection and diagnosis. While convolutional neural networks
(CNNs), recurrent neural networks (RNNs), and generative models have received signifi-
cant attention in the field, there are several other emerging technologies that hold promise
Mach. Learn. Knowl. Extr. 2024, 6 488

for advancing AD research. In this section, we explore some of these trending technologies
and their potential applications in AD studies.

6.1. Graph Convolutional Networks (GCNs)

Table 4 offers an overview of studies that have utilized GCNs for AD detection. Graph
convolutional networks (GCNs) have gained attention for their ability to effectively analyze
graph-structured data, making them particularly suitable for modelling brain connectivity
networks in AD. GCNs can provide insights into the underlying structural and functional
changes associated with AD by capturing relationships between brain regions. Recent
studies have shown promising results in using GCNs to classify AD patients from healthy
controls based on brain network connectivity data.

Table 4. Overview of GCN studies for AD detection.

Reference Input Data Technique Evaluation Notes

Method utilized GCNs to merge
GCNs combined with
Framework evaluated on ABIDE and ADNI imaging and non-imaging data,
imaging and non-imaging
[254] MRI datasets. Assessed for disease prediction leading to improved classification
information integration in a
accuracy. accuracy, reaching 70.4% for ABIDE
sparse graph representation.
and 80.0% for ADNI.
The method was evaluated using VBM-MRI,
sMRI, Interpretable graph Proposed multi-modality
FDG-PET, and AV45-PET modalities.
FDG-PET, convolutional network imaging-based GCN for AD
[255] Assessed on clinical score prediction, disease
and (GCN) framework extended classification for effective ROI
status identification, and biomarker
AV45-PET with GradCAM technique. quantification on the ADNI dataset.
identification for AD and MCI.
Two-phase framework that Proposed two-phase framework for
iteratively assigns weights AD diagnosis, leading to reduced
Compared to the state-of-the-art in
[256] MRI to samples and features to biasing and improved
classification and interpretability.
address training set bias and interpretability for binary
improve interpretability. classification on ADNI dataset.
FSNet is a dual interpretable
FSNet overcomes GCN limitations
graph convolutional The FSNet model demonstrates superior
with the dual interpretable
network for enhancing classification performance and
[257] MRI framework, outperforming
model performance and interpretability compared to recent
state-of-the-art methods in ADNI
interpretability in medical state-of-the-art methods.
dataset classification.
diagnosis.
GCNs coupled with Integration of feature learning and
The performance of the proposed method
interpretable feature dynamic graph learning into GCN
[258] MRI was evaluated based on its diagnosis
learning and dynamic for robust and personalized disease
accuracy for early AD detection.
graph learning. diagnosis with improved accuracy.

6.2. Attention Mechanisms

Table 5 offers an overview of studies that have utilized Attention mechanisms for
AD detection. Attention mechanisms have emerged as a powerful tool in deep learning,
allowing for models to focus on relevant features or regions of interest. In the context of AD
studies, attention mechanisms can aid in identifying critical brain regions or biomarkers
that contribute significantly to disease progression. By selectively attending to informative
regions, attention-based models can improve the interpretability of predictions and enhance
our understanding of AD pathology.

Table 5. Overview of attention mechanism studies for AD detection.

Reference Input Data Technique Evaluation Notes

Proposed method evaluated on ADNI MRI DL method using connected CNNs
Densely connected CNN with data of AD vs. healthy, MCI converter vs. for AD detection resulted in
[154] MRI
an attention mechanism. healthy, and MCI converter vs. improved AD classification rate
non-converter using accuracy. MCI predictions.
Combining image filtering,
Evaluation of classification performance Study explored image filtering and
pyramid squeeze attention
using accuracy, considering image filtering PSA impact on AD classification,
[259] MRI (PSA) mechanism, FCN, and
approaches and attention mechanisms’ with a reported accuracy of 98.85%
MLP for improved image
impact on AD diagnosis. for classification.
analysis.
Mach. Learn. Knowl. Extr. 2024, 6 489

6.3. Transfer Learning

Table 6 offers an overview of studies that have utilized Transfer learning for AD
detection. Transfer learning, a technique that leverages knowledge learned from one
task to improve performance on another related task, has shown promise in AD research.
Pretrained models on large-scale datasets, such as ImageNet, can be fine-tuned using
AD-specific data to extract discriminative features. Transfer learning enables the utilization
of knowledge from diverse domains and can enhance the generalization ability of AD
detection models, especially when data scarcity is challenging.

Table 6. Overview of transfer learning studies for AD detection.

Reference Input Data Technique Evaluation Notes

The evaluation metric used was accuracy,
Automated system for precise AD
Transfer learning on a measuring the system’s performance in
detection using multi-class
[208] MRI multiclass classification classifying MRI images into different
approach achieved 91.70% accuracy
model using deep learning. Alzheimer’s disease stages: MD, MOD,
to predict the stage of disease.
ND, and VMD.
The evaluation is based on the accuracy of
The method explored the sequential
the system for AD detection using MRI
Transfer learning for 2D relationship of transfer learning and
[260] MRI scans. The performance is compared
CNNs followed by RNN. RNN for Alzheimer’s detection
between using a 2D CNN alone and using
improvements.
a combination of a 2D CNN and an RNN.
VGG as the pre-trained
The proposed model was evaluated on AD
model for transfer learning The study proposed transfer
vs. NC, AD vs. MCI, and MCI vs. NC
on MRI images. Fine-tuning learning followed by intelligent
[261] MRI classification tasks. It outperformed
with layer-wise tuning tuning for improved AD
state-of-the-art methods in terms of
improves efficiency with classification over small datasets.
accuracy and other performance metrics.
smaller datasets.
The proposed system
The system’s performance was evaluated
employs transfer learning Transfer learning validated for AD
using various metrics, including overall
with AlexNet for image detection on brain MRI with 92.85%
[262] MRI accuracy for binary (AD vs. non-AD) and
classification, tested on both accuracy on
multiclass (four dementia stages)
segmented and segmented/unsegmented imagery.
classification.
unsegmented images.
Deep learning models with
transfer learning are used, Transfer learning improved AD
Models were evaluated using accuracy,
including 3D CNNs and detection accuracy up to 98.20% and
sensitivity, specificity, precision, and
[263] MRI pretrained network-based prognostic prediction accuracy up to
F1-score to assess AD classification and
architectures, to extract 87.78%; however; the dataset used
disease progression prediction.
high-level features from was limited.
neuroimaging data.

6.4. Autoencoders
Table 7 offers an overview of studies that have utilized Autoencoders for AD detection.
Autoencoders are unsupervised learning models that learn to encode and decode data,
often used for dimensionality reduction or data reconstruction. In AD studies, autoencoders
have been employed for anomaly detection by reconstructing normal brain patterns and
identifying deviations indicative of AD pathology. By capturing the underlying structure
of AD-related changes, autoencoders can contribute to early detection and monitoring of
disease progression.

Table 7. Overview of autoencoder studies for AD detection.

Reference Input Data Technique Evaluation Notes

Proposed deep convoltional
Deep convolutional autoencoders The study analyzed extracted
autoencoders extracted AD-related
are used for exploratory data features using regression,
imaging features, with strong
analysis of Alzheimer’s disease. classification, and correlation
correlations (>0.6) to clinical data,
[264] MRI They extract abstract features from techniques. It evaluated their
achieving 80% diagnosis accuracy,
MRI images, representing the data relationship with clinical variables
and showcasing deep learning’s
distribution in low-dimensional and measured AD diagnosis
potential in understanding AD’s
manifolds. accuracy.
clinical features.
Mach. Learn. Knowl. Extr. 2024, 6 490

Table 7. Cont.

Reference Input Data Technique Evaluation Notes

Study used supervised switching
autoencoders (SSAs) for AD Supervised switching autoencoders
The proposed model supervised
classification. Models trained on (SSAs) accuracy was assessed for
switching autoencoders (SSAs)
2D slice patches, exploring healthy vs. AD-demented subjects,
[265] sMRI classified AD using one MRI slice by
sizes/parameters. Patch-level comparing identified regions with
combining patch representations
classification identifies disease prior studies and medical
and achieved high accuracy.
regions based on knowledge.
accuracy densities.
The ANNs were evaluated based on
Two ANNs, with stacked hidden
accuracy for classifying ADD and ANNs stacked hidden layers
layers for input recreation, classify
control participants using rsEEG, effectively to distinguish AD from
ADD using LORETA source
[266] rsEEG, sMRI sMRI, and combined features. healthy controls, i.e., a combination
estimates and sMRI variables. The
Specialized ANNs for ADD and of rsEEG and sMRI features yields
task involves discriminating
controls were also assessed with the improved accuracy.
between AD and healthy controls.
same features.
The method involves two steps: Evaluation metrics included
(1) Extracting image features using accuracy, sensitivity, and specificity. Two-step approach resulted in high
[267] 3D MRI a pre-trained autoencoder Accuracy rates were 95% for accuracy for Alzheimer’s disease
ensemble, and (2) Diagnosing AD/NC, 90% for AD/MCI, and diagnosis using 3D images.
Alzheimer’s disease with a CNN. 92.5% for MCI/NC classification.

7. Highlights
Recent advancements in Alzheimer’s disease (AD) research have elucidated a di-
verse spectrum of disease subtypes, revealing at least five distinct variants, each char-
acterized by unique anatomical pathologies divergent from traditional markers such as
Thal or Braak staging [268]. Through meticulous neuropathological and neuroimaging
analyses, researchers have consistently identified three primary subtypes: typical AD,
limbic-predominant AD, and hippocampal-sparing AD, with the emergence of a fourth
subtype, minimal atrophy AD [269]. Additionally, a subgroup devoid of discernible atro-
phy has been delineated as a distinct AD subtype. These subtypes have been discerned
through intricate patterns of brain atrophy and neuropathological characteristics, exhibiting
heterogeneous clinical and cognitive features, with certain variants demonstrating slower
disease progression compared to the prototypical AD presentation [270]. Understanding
the intricacies of these subtypes is paramount for elucidating the heterogeneity of AD, with
implications for enhancing discrimination, accurate diagnosis, and targeted therapeutic
interventions [271]. Moreover, it is posited that an individual’s positionings along the
typicality and severity spectra are shaped by a complex interplay of protective factors, risk
factors, and diverse brain pathologies, giving rise to the delineation of four unique AD
subtypes: typical AD, limbic-predominant AD, hippocampal-sparing AD, and minimum
atrophy AD [272].
Alzheimer’s disease (AD) is widely recognized for its inherent heterogeneity, both in
terms of disease manifestation and demographic factors. Importantly, it is rare to encounter
pure cases of AD, as individuals often present with a complex interplay of multiple diseases.
This aspect is crucial when assessing disease progression, developing new analyses, or
classifying deep learning methods. Recent research has introduced the concept of at least
five distinct AD subtypes, each characterized by unique anatomical pathologies beyond
traditional markers like Thal or Braak staging [268]. While this classification enhances our
understanding of AD diversity, it also poses challenges in diagnosis and necessitates a
nuanced approach to disease characterization.
To expand the heterogeneous nature of AD, there is a need to emphasize implications
for both clinical practice and research endeavors. Clinicians and researchers should know
that AD cases often manifest as a composite of different subtypes, which makes it challeng-
ing to identify pure cases [269]. A thorough approach is necessary since the occurrence
of multiple subtypes hampers diagnostic attempts significantly. Furthermore, various
variables, including different brain disorders and protective and risk factors, influence how
diseases proceed in different people [270].
Mach. Learn. Knowl. Extr. 2024, 6 491

Nevertheless, the diagnosis of pure AD cases presents a formidable challenge due

to the disease’s inherent heterogeneity, compounded by mixed pathologies, coexisting
conditions, and clinical syndromes [273]. The coexistence of multiple subtypes compli-
cates diagnostic efforts, making identifying pure AD cases arduous. It is imperative to
acknowledge that these subtypes are not mutually exclusive, as AD cases often mani-
fest as a composite of different subtypes [274]. The presence of varied subtypes further
complicates the diagnostic process, posing challenges in accurately delineating pure AD
cases. Alzheimer’s disease presents a notable diversity among individuals, evident in
both its clinical expression and underlying pathological mechanisms. Although memory
impairment is a primary symptom, the impact on other cognitive functions, such as ex-
ecutive function, language, and visuospatial skills, varies among patients. The pace of
disease progression differs significantly from person to person. Various factors contribute
to this diversity, including age at onset, genetic predisposition, presence of other health
conditions, environmental influences, and individual variances in brain resilience and
compensatory mechanisms [275]. Moreover, AD frequently co-occurs with other neurode-
generative conditions, with stroke and vascular dementia emerging as the most prevalent
comorbidities [276]. The diagnostic landscape is further convoluted by significant clinical
variability in the age of onset and in neurological and cognitive characteristics. AD patients
may present with concomitant illnesses such as TDP-43 proteinopathy, Lewy body disease,
and cerebrovascular disease, further confounding diagnosis [277]. Notably, the diagnosis
of AD necessitates the presence of intracellular neurofibrillary tangles and extracellular
amyloid plaques, adding layers of complexity to the diagnostic process [278].
In conclusion, the multifaceted nature of AD subtypes underscores the need for a
nuanced and comprehensive approach to disease characterization and diagnosis. By recog-
nizing the diverse manifestations of AD and its associated subtypes, clinicians can tailor
interventions to individual patients, optimizing clinical outcomes and enhancing patient
care. Moreover, continued research into AD subtypes holds promise for the development
of targeted therapeutic strategies that address specific pathological mechanisms underlying
different AD variants, ultimately improving outcomes for individuals affected by this
devastating disease.

8. Challenges
Deep learning architectures, such as recurrent neural networks (RNNs), convolutional
neural networks (CNNs), and generative modelling, have emerged as powerful tools in
Alzheimer’s disease (AD) research. These architectures have shown great potential in
analyzing various types of data, including imaging, genetic, and clinical data, to advance
our understanding of the disease. However, despite their successes, they also face a number
of challenges that need to be addressed in order to maximize their impact and applicability
in AD research.
One of the challenges faced by RNNs is the limited availability of longitudinal datasets.
RNNs excel at modelling temporal dependencies and capturing sequential patterns, making
them well-suited for analyzing disease progression over time. However, acquiring large-
scale longitudinal datasets with diverse AD populations is crucial to training robust RNN
models. Additionally, the heterogeneity of AD data poses a challenge for RNNs. AD is a
complex and multifaceted disease, and there is significant variability in data acquisition
protocols and demographic factors across different studies. This heterogeneity requires
researchers to develop more sophisticated modelling techniques to effectively capture and
generalize the patterns in AD data.
Interpretability and explainability are also important challenges for RNNs in AD
research. RNNs are often regarded as black-box models, making interpreting and explain-
ing their predictions difficult. To address this, researchers need to explore methods for
extracting meaningful features, visualizing temporal patterns, and providing explanations
for RNN-based predictions. This will help gain insights into the underlying neurobiological
processes and enhance the clinical utility of RNN models.
Mach. Learn. Knowl. Extr. 2024, 6 492

CNNs, on the other hand, have demonstrated remarkable performance in analyzing

medical imaging data, including brain MRI and PET scans. However, they face their own
set of challenges in AD research. One such challenge is the need for large and diverse
datasets to train CNN models effectively. AD data are often limited in size and can exhibit
class imbalances, requiring careful data augmentation strategies and techniques to address
these issues. Furthermore, CNNs struggle with generalizing across different imaging
modalities and acquisition protocols. AD studies often involve multi-site collaborations
and variations in imaging protocols that can introduce unwanted variability. Developing
robust techniques to handle these challenges and ensure model generalization is a key area
of research.
Generative modelling approaches, such as generative adversarial networks (GANs),
offer exciting possibilities for data augmentation, image synthesis, and generating realistic
brain images. However, there are challenges that need to be addressed in this domain
as well. Training GANs for AD research requires access to large and diverse datasets,
which can be difficult to obtain due to privacy concerns and data availability. Additionally,
ensuring that the generated images are biologically plausible and representative of the un-
derlying AD pathology is a critical challenge. Striking a balance between data augmentation
and maintaining the integrity of AD-specific features is a topic of ongoing research.
While deep learning architectures have shown promise in AD research, they also
face challenges that need to be overcome to fully harness their potential. Addressing
the limitations and developing innovative solutions in data availability, heterogeneity,
interpretability, generalization, and biological plausibility will contribute to advancing AD
research and ultimately improve our understanding and management of the disease.

9. Future Perspectives and Recommendations

Deep learning architectures, including recurrent neural networks (RNNs), convolu-
tional neural networks (CNNs), and generative modelling, have shown great promise in
the field of Alzheimer’s disease (AD) research. These advanced techniques have provided
valuable insights and improved our understanding of the disease. As we look to the
future, several perspectives and recommendations can guide further advancements in deep
learning for AD research.
One important perspective is the integration of multiple modalities. Deep learning
models should continue to explore the combination of various data sources, such as neu-
roimaging, genetics, and clinical data. By leveraging the complementary information from
these modalities, we can enhance the accuracy of AD diagnosis, prognosis, and treatment
response prediction. Integrating multimodal data can provide a more comprehensive view
of the disease and enable the development of personalized treatment strategies.
Another key perspective is the analysis of longitudinal data. AD is a progressive
disease that unfolds over time, and capturing the dynamic changes is crucial for under-
standing its trajectory. Deep learning architectures can be further developed to effectively
model and analyze longitudinal data, enabling researchers to track disease progression and
identify early biomarkers of AD. Longitudinal analysis can provide valuable insights into
disease mechanisms and aid in developing targeted interventions.
Furthermore, it is important to address the challenges associated with limited data
availability in AD research. Deep learning techniques often require large amounts of
labelled data for optimal performance. However, AD datasets are typically limited due to
the difficulty and cost of data collection. To overcome this challenge, researchers can explore
transfer learning techniques, where pre-trained models on related tasks or datasets are fine-
tuned for AD analysis. Additionally, data augmentation strategies can artificially increase
the available data’s size and diversity, enabling more robust and generalizable models.
In terms of model interpretability, future research should focus on developing tech-
niques to enhance the transparency and explainability of deep learning models in AD
diagnosis and prediction. Interpretability in medical applications is crucial to gain the trust
and acceptance of clinicians and ensure the ethical use of AI technologies. Efforts should be
Mach. Learn. Knowl. Extr. 2024, 6 493

made to incorporate interpretable components, such as attention mechanisms or saliency

maps, into deep learning architectures for AD analysis.
Establishing standardized benchmarks and evaluation protocols for AD-related deep
learning tasks is recommended to promote collaboration and accelerate progress in the
field. This would allow for fair comparisons between different models and facilitate
the reproducibility of research findings. Furthermore, the sharing of well-curated and
annotated datasets can help overcome the limitations of data scarcity and encourage the
development of novel algorithms and methodologies.
Deep learning architectures hold great potential for advancing our understanding of
AD and improving diagnosis, prognosis, and treatment. By integrating multiple modalities,
analyzing longitudinal data, addressing data limitations, enhancing interpretability, and
fostering collaboration, we can pave the way for more accurate, efficient, and interpretable
deep learning models in AD research. These efforts have the potential to transform clinical
practice and contribute to the development of personalized and targeted interventions for
individuals at risk or affected by AD.

10. Conclusions
In conclusion, this systematic literature review has provided valuable insights into
the current state of Alzheimer’s disease (AD) detection using deep learning approaches.
This review highlights the potential of deep models, particularly in neuroimaging, for
accurate AD detection and emphasizes the importance of highly discriminative feature
representations.
The analysis of various biomarkers, features, and pre-processing techniques for neu-
roimaging data from single-modality and multi-modality studies has demonstrated the
versatility of deep learning models in capturing the complex patterns associated with AD.
Specifically, deep learning architectures such as convolutional neural networks (CNNs),
recurrent neural networks (RNNs), and generative models have been examined for their
performance in AD detection.
Despite the promising results, this review also identifies several challenges that need
to be addressed. The limited availability of datasets and the need for robust training
procedures pose significant hurdles in achieving optimal performance with deep learning
models. These challenges highlight the importance of developing benchmark platforms and
standardized evaluation protocols to facilitate comparative analysis and foster collaboration
in the field.
Looking ahead, future research directions should focus on overcoming the limitations
identified in this review. The development of highly discriminative feature representations
that can effectively differentiate AD from similar brain patterns is crucial. Additionally,
advancements in model architectures and training methodologies are necessary to enhance
the performance and generalizability of deep learning models for AD detection.
The findings of this review underscore the potential of deep learning in improving the
diagnostic accuracy of AD. However, it is essential to recognize that deep learning is not
a standalone solution, and it should be integrated with other clinical data and diagnostic
tools to achieve comprehensive and accurate AD detection.
In summary, deep learning holds significant promise for advancing AD detection.
However, further advancements in models and methodologies are necessary to overcome
the challenges associated with limited datasets and training procedures. By addressing
these challenges and promoting collaboration and standardization, deep learning can
contribute to the development of practical diagnostic methods for AD, leading to earlier
detection and intervention for improved patient outcomes.

Author Contributions: M.G.A.: conceptualization, methodology, software, visualization, validation,

data curation, and writing—original draft preparation; S.L. and K.S.: methodology, supervision,
project administration, and writing—review and editing. All authors have read and agreed to the
published version of the manuscript.
Mach. Learn. Knowl. Extr. 2024, 6 494

Funding: This study received no external funding.

Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Vaupel, J.W. Biodemography of human ageing. Nature 2010, 464, 536–542. [CrossRef]
2. Samir, K.C.; Lutz, W. The human core of the shared socioeconomic pathways: Population scenarios by age, sex and level of
education for all countries to 2100. Glob. Environ. Chang. 2017, 42, 181–192. [CrossRef]
3. Godyń, J.; Jończyk, J.; Panek, D.; Malawska, B. Therapeutic strategies for Alzheimer’s disease in clinical trials. Pharmacol. Rep.
2016, 68, 127–138. [CrossRef]
4. Blaikie, L.; Kay, G.; Lin, P.K.T. Current and emerging therapeutic targets of Alzheimer’s disease for the design of multi-target
directed ligands. MedChemComm 2019, 10, 2052–2072. [CrossRef]
5. Gutierrez, B.A.; Limon, A. Synaptic disruption by soluble oligomers in patients with Alzheimer’s and Parkinson’s disease.
Biomedicines 2022, 10, 1743. [CrossRef]
6. Zvěřová, M. Clinical aspects of Alzheimer’s disease. Clin. Biochem. 2019, 72, 3–6. [CrossRef]
7. Bronzuoli, M.R.; Iacomino, A.; Steardo, L.; Scuderi, C. Targeting neuroinflammation in Alzheimer’s disease. J. Inflamm. Res. 2016,
9, 199–208. [CrossRef] [PubMed]
8. Aljunid, S.M.; Maimaiti, N.; Ahmed, Z.; Nur, A.M.; Nor, N.M.; Ismail, N.; Haron, S.A.; Shafie, A.A.; Salleh, M.; Yusuf, S.; et al.
Development of clinical pathway for mild cognitive impairment and dementia to quantify cost of age-related cognitive disorders
in Malaysia. Malays. J. Public Health Med. 2014, 14, 88–96.
9. Hu, K.; Wang, Y.; Chen, K.; Hou, L.; Zhang, X. Multi-scale features extraction from baseline structure MRI for MCI patient
classification and AD early diagnosis. Neurocomputing 2016, 175, 132–145. [CrossRef]
10. Eid, A.; Mhatre, I.; Richardson, J.R. Gene-environment interactions in Alzheimer’s disease: A potential path to precision medicine.
Pharmacol. Ther. 2019, 199, 173–187. [CrossRef]
11. McKhann, G.; Drachman, D.; Folstein, M.; Katzman, R.; Price, D.; Stadlan, E.M. Clinical diagnosis of Alzheimer’s disease:
Report of the NINCDS—ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on
Alzheimer’s Disease. Neurology 1984, 34, 939. [CrossRef] [PubMed]
12. Dubois, B.; Feldman, H.H.; Jacova, C.; DeKosky, S.T.; Barberger-Gateau, P.; Cummings, J.L.; Delacourte, A.; Galasko, D.; Gauthier,
S.; Jicha, G.A.; et al. Research criteria for the diagnosis of Alzheimer’s disease: Revising the NINCDS–ADRDA criteria. Lancet
Neurol. 2007, 6, 734–746. [CrossRef] [PubMed]
13. Jack, C.R., Jr.; Bennett, D.A.; Blennow, K.; Carrillo, M.C.; Dunn, B.; Haeberlein, S.B.; Holtzman, D.M.; Jagust, W.; Jessen, F.;
Karlawish, J.; et al. NIA-AA research framework: Toward a biological definition of Alzheimer’s disease. Alzheimer’s Dement. 2018,
14, 535–562. [CrossRef]
14. Arevalo-Rodriguez, I.; Smailagic, N.; I Figuls, M.R.; Ciapponi, A.; Sanchez-Perez, E.; Giannakou, A.; Pedraza, O.L.; Cosp, X.B.;
Cullum, S. Mini-Mental State Examination (MMSE) for the detection of Alzheimer’s disease and other dementias in people with
mild cognitive impairment (MCI). Cochrane Database Syst. Rev. 2015, 5, 3. [CrossRef]
15. Wei, Y.-C.; Chen, C.-K.; Lin, C.; Chen, P.-Y.; Hsu, P.-C.; Lin, C.-P.; Shyu, Y.-C.; Huang, W.-Y. Normative Data of Mini-Mental State
Examination, Montreal Cognitive Assessment, and Alzheimer’s Disease Assessment Scale-Cognitive Subscale of Community-
Dwelling Older Adults in Taiwan. Dement. Geriatr. Cogn. Disord. 2022, 51, 365–376. [CrossRef]
16. Qiao, H.; Chen, L.; Zhu, F. Ranking convolutional neural network for Alzheimer’s disease mini-mental state examination
prediction at multiple time-points. Comput. Methods Programs Biomed. 2022, 213, 106503. [CrossRef]
17. Morris, J.C. Clinical dementia rating: A reliable and valid diagnostic and staging measure for dementia of the Alzheimer type.
Int. Psychogeriatr. 1997, 9, 173–176. [CrossRef] [PubMed]
18. Cedarbaum, J.M.; Jaros, M.; Hernandez, C.; Coley, N.; Andrieu, S.; Grundman, M.; Vellas, B.; Alzheimer’s Disease Neuroimaging
Initiative. Rationale for use of the Clinical Dementia Rating Sum of Boxes as a primary outcome measure for Alzheimer’s disease
clinical trials. Alzheimer’s Dement. 2013, 9, S45–S55. [CrossRef]
19. Williams, M.M.; Storandt, M.; Roe, C.M.; Morris, J.C. Progression of Alzheimer’s disease as measured by Clinical Dementia
Rating Sum of Boxes scores. Alzheimer’s Dement. 2013, 9, S39–S44. [CrossRef]
20. Audronyte, E.; Pakulaite-Kazliene, G.; Sutnikiene, V.; Kaubrys, G. Properties of odor identification testing in screening for
early-stage Alzheimer’s disease. Sci. Rep. 2023, 13, 6075. [CrossRef]
21. Rodriguez-Santiago, M.A.; Sepulveda, V.; Valentin, E.M.; Arnold, S.E.; Jiménez-Velázquez, I.Z.; Wojna, V. Diagnosing Alzheimer
Disease: Which Dementia Screening Tool to Use in Elderly Puerto Ricans with Mild Cognitive Impairment and Early Alzheimer
Disease? Alzheimer’s Dement. 2022, 18, e062560. [CrossRef]
22. Tsoy, E.; VandeVrede, L.; Rojas, J.C.; Possin, K.L. Cognitive assessment in diverse populations: Implications for Alzheimer’s
disease clinical trials. Alzheimer’s Dement. 2022, 18, e064114. [CrossRef]
23. Celik, S.; Onur, O.; Yener, G.; Kessler, J.; Özbek, Y.; Meyer, P.; Frölich, L.; Teichmann, B. Cross-cultural comparison of MMSE and
RUDAS in German and Turkish patients with Alzheimer’s disease. Neuropsychology 2022, 36, 195–205. [CrossRef] [PubMed]
Mach. Learn. Knowl. Extr. 2024, 6 495

24. Chen, X.; Zhang, W.; Lin, Z.; Zheng, C.; Chen, S.; Zhou, H.; Liu, Z. Preliminary evidence for developing safe and efficient fecal
microbiota transplantation as potential treatment for aged related cognitive impairments. Front. Cell. Infect. Microbiol. 2023,
13, 211. [CrossRef] [PubMed]
25. Adair, T.; Temple, J.; Anstey, K.J.; Lopez, A.D. Is the rise in reported dementia mortality real? Analysis of multiple-cause-of-death
data for Australia and the United States. Am. J. Epidemiol. 2022, 191, 1270–1279. [CrossRef] [PubMed]
26. Patel, D.; Montayre, J.; Karamacoska, D.; Siette, J. Progressing dementia risk reduction initiatives for culturally and linguistically
diverse older adults in Australia. Australas. J. Ageing 2022, 41, 579–584. [CrossRef]
27. Brennan, F.; Chapman, M.; Gardiner, M.D.; Narasimhan, M.; Cohen, J. Our dementia challenge: Arise palliative care. Intern. Med.
J. 2023, 53, 186–193. [CrossRef] [PubMed]
28. Ebrahimighahnavieh, A.; Luo, S.; Chiong, R. Deep learning to detect Alzheimer’s disease from neuroimaging: A systematic
literature review. Comput. Methods Programs Biomed. 2020, 187, 105242. [CrossRef]
29. Fathi, S.; Ahmadi, M.; Dehnad, A. Early diagnosis of Alzheimer’s disease based on deep learning: A systematic review. Comput.
Biol. Med. 2022, 146, 105634. [CrossRef] [PubMed]
30. Binaco, R.; Calzaretto, N.; Epifano, J.; McGuire, S.; Umer, M.; Emrani, S.; Wasserman, V.; Libon, D.J.; Polikar, R. Machine learning
analysis of digital clock drawing test performance for differential classification of mild cognitive impairment subtypes versus
Alzheimer’s disease. J. Int. Neuropsychol. Soc. 2020, 26, 690–700. [CrossRef]
31. Karikari, T.K.; Ashton, N.J.; Brinkmalm, G.; Brum, W.S.; Benedet, A.L.; Montoliu-Gaya, L.; Lantero-Rodriguez, J.; Pascoal, T.A.;
Suárez-Calvet, M.; Rosa-Neto, P.; et al. Blood phospho-tau in Alzheimer disease: Analysis, interpretation, and clinical utility. Nat.
Rev. Neurol. 2022, 18, 400–418. [CrossRef]
32. Wang, M.; Song, W.M.; Ming, C.; Wang, Q.; Zhou, X.; Xu, P.; Krek, A.; Yoon, Y.; Ho, L.; Orr, M.E.; et al. Guidelines for
bioinformatics of single-cell sequencing data analysis in Alzheimer’s disease: Review, recommendation, implementation and
application. Mol. Neurodegener. 2022, 17, 1–52. [CrossRef]
33. Zhang, Y.; Wu, K.-M.; Yang, L.; Dong, Q.; Yu, J.-T. Tauopathies: New perspectives and challenges. Mol. Neurodegener. 2022, 17, 28.
[CrossRef]
34. Klyucherev, T.O.; Olszewski, P.; Shalimova, A.A.; Chubarev, V.N.; Tarasov, V.V.; Attwood, M.M.; Syvänen, S.; Schiöth, H.B.
Advances in the development of new biomarkers for Alzheimer’s disease. Transl. Neurodegener. 2022, 11, 1–24. [CrossRef]
[PubMed]
35. Zhao, K.; Duka, B.; Xie, H.; Oathes, D.J.; Calhoun, V.; Zhang, Y. A dynamic graph convolutional neural network framework
reveals new insights into connectome dysfunctions in ADHD. NeuroImage 2022, 246, 118774. [CrossRef]
36. Jiang, Y.; Zhou, X.; Ip, F.C.; Chan, P.; Chen, Y.; Lai, N.C.; Cheung, K.; Lo, R.M.; Tong, E.P.; Wong, B.W.; et al. Large-scale plasma
proteomic profiling identifies a high-performance biomarker panel for Alzheimer’s disease screening and staging. Alzheimer’s
Dement. 2022, 18, 88–102. [CrossRef] [PubMed]
37. Gómez-Isla, T.; Frosch, M.P. Lesions without symptoms: Understanding resilience to Alzheimer disease neuropathological
changes. Nat. Rev. Neurol. 2022, 18, 323–332. [CrossRef]
38. Liu, C.; Xiang, X.; Han, S.; Lim, H.Y.; Li, L.; Zhang, X.; Ma, Z.; Yang, L.; Guo, S.; Soo, R.; et al. Blood-based liquid biopsy: Insights
into early detection and clinical management of lung cancer. Cancer Lett. 2022, 524, 91–102. [CrossRef]
39. Hernandez, M.; Ramon-Julvez, U.; Ferraz, F.; with the ADNI Consortium. Explainable AI toward understanding the performance
of the top three TADPOLE Challenge methods in the forecast of Alzheimer’s disease diagnosis. PLoS ONE 2022, 17, e0264695.
[CrossRef] [PubMed]
40. Lydon, E.A.; Nguyen, L.T.; Shende, S.A.; Chiang, H.-S.; Spence, J.S.; Mudar, R.A. EEG theta and alpha oscillations in early versus
late mild cognitive impairment during a semantic Go/NoGo task. Behav. Brain Res. 2022, 416, 113539. [CrossRef]
41. Yadav, S.; Zhou Shu, K.; Zachary, Z.; Yueyang, G.; Lana, X.; ADNI Consortium. Integrated Metabolomics and Transcriptomics
Analysis Identifies Molecular Subtypes within the Early and Late Mild Cognitive Impairment Stages of Alzheimer’s Disease.
medRxiv 2023. [CrossRef]
42. Jeyavathana, R.B.; Balasubramanian, R.; Pandian, A.A. A survey: Analysis on pre-processing and segmentation techniques for
medical images. Int. J. Res. Sci. Innov. (IJRSI) 2016, 3, 113–120.
43. James, A.P.; Dasarathy, B.V. Medical image fusion: A survey of the state of the art. Inf. Fusion 2014, 19, 4–19. [CrossRef]
44. Balagurunathan, Y.; Mitchell, R.; El Naqa, I. Requirements and reliability of AI in the medical context. Phys. Medica 2021, 83,
72–78. [CrossRef] [PubMed]
45. Nithya, V.P.; Mohanasundaram, N. An Extensive Survey of Various Deep Learning Approaches for Predicting Alzheimer’s
Disease. Ann. Rom. Soc. Cell Biol. 2021, 25, 848–860.
46. Pan, Y.; Liu, M.; Lian, C.; Xia, Y.; Shen, D. Spatially-constrained fisher representation for brain disease identification with
incomplete multi-modal neuroimages. IEEE Trans. Med. Imaging 2020, 39, 2965–2975. [CrossRef]
47. Liu, M.; Zhang, J.; Adeli, E.; Shen, D. Landmark-based deep multi-instance learning for brain disease diagnosis. Med. Image Anal.
2017, 43, 157–168. [CrossRef]
48. Ebrahimi, A.; Luo, S.; Alzheimer’s Disease Neuroimaging Initiative. Convolutional neural networks for Alzheimer’s disease
detection on MRI images. J. Med. Imaging 2021, 8, 024503. [CrossRef]
49. Lee, J.-G.; Jun, S.; Cho, Y.-W.; Lee, H.; Kim, G.B.; Seo, J.B.; Kim, N. Deep learning in medical imaging: General overview. Korean J.
Radiol. 2017, 18, 570–584. [CrossRef]
Mach. Learn. Knowl. Extr. 2024, 6 496

50. Mirzaei, G.; Adeli, H. Machine learning techniques for diagnosis of Alzheimer disease, mild cognitive disorder, and other types
of dementia. Biomed. Signal Process. Control 2020, 72, 103293. [CrossRef]
51. Feng, J.; Zhang, S.-W.; Chen, L.; Zuo, C. Detection of Alzheimer’s disease using features of brain region-of-interest-based
individual network constructed with the sMRI image. Comput. Med. Imaging Graph. 2022, 98, 102057. [CrossRef] [PubMed]
52. El-Sappagh, S.; Ali, F.; Abuhmed, T.; Singh, J.; Alonso, J.M. Automatic detection of Alzheimer’s disease progression: An efficient
information fusion approach with heterogeneous ensemble classifiers. Neurocomputing 2022, 512, 203–224. [CrossRef]
53. Dubois, B.; Picard, G.; Sarazin, M. Early detection of Alzheimer’s disease: New diagnostic criteria. Dialog. Clin. Neurosci. 2022, 10,
35. [CrossRef] [PubMed]
54. Hamdi, M.; Bourouis, S.; Rastislav, K.; Mohmed, F. Evaluation of Neuro Images for the Diagnosis of Alzheimer’s Disease Using
Deep Learning Neural Network. Front. Public Health 2022, 10, 834032. [CrossRef]
55. Kim, J.S.; Han, J.W.; Bin Bae, J.; Moon, D.G.; Shin, J.; Kong, J.E.; Lee, H.; Yang, H.W.; Lim, E.; Sunwoo, L.; et al. Deep learning-based
diagnosis of Alzheimer’s disease using brain magnetic resonance images: An empirical study. Sci. Rep. 2022, 12, 18007. [CrossRef]
56. Qiu, S.; Miller, M.I.; Joshi, P.S.; Lee, J.C.; Xue, C.; Ni, Y.; Wang, Y.; De Anda-Duran, I.; Hwang, P.H.; Cramer, J.A.; et al. Multimodal
deep learning for Alzheimer’s disease dementia assessment. Nat. Commun. 2022, 13, 3404. [CrossRef] [PubMed]
57. Goel, T.; Sharma, R.; Tanveer, M.; Suganthan, P.N.; Maji, K.; Pilli, R. Multimodal Neuroimaging based Alzheimer’s Disease
Diagnosis using Evolutionary RVFL Classifier. IEEE J. Biomed. Health Inform. 2023, 1–9. [CrossRef]
58. Tu, Y.; Lin, S.; Qiao, J.; Zhuang, Y.; Zhang, P. Alzheimer’s disease diagnosis via multimodal feature fusion. Comput. Biol. Med.
2022, 148, 105901. [CrossRef]
59. Jia, H.; Lao, H. Deep learning and multimodal feature fusion for the aided diagnosis of Alzheimer’s disease. Neural Comput. Appl.
2022, 34, 19585–19598. [CrossRef]
60. Kong, Z.; Zhang, M.; Zhu, W.; Yi, Y.; Wang, T.; Zhang, B. Multi-modal data Alzheimer’s disease detection based on 3D convolution.
Biomed. Signal Process. Control 2022, 75, 103565. [CrossRef]
61. AlSaeed, D.; Omar, S.F. Brain MRI analysis for Alzheimer’s disease diagnosis using CNN-based feature extraction and machine
learning. Sensors 2022, 22, 2911. [CrossRef] [PubMed]
62. Rossini, P.M.; Miraglia, F.; Vecchio, F. Early dementia diagnosis, MCI-to-dementia risk prediction, and the role of machine learning
methods for feature extraction from integrated biomarkers, in particular for EEG signal analysis. Alzheimer’s Dement. 2022, 18,
2699–2706. [CrossRef]
63. Taghavirashidizadeh, A.; Sharifi, F.; Vahabi, S.A.; Hejazi, A.; SaghabTorbati, M.; Mohammed, A.S. WTD-PSD: Presentation of
novel feature extraction method based on discrete wavelet transformation and time-dependent power spectrum descriptors for
diagnosis of Alzheimer’s disease. Comput. Intell. Neurosci. 2022, 2022, 9554768. [CrossRef] [PubMed]
64. Mahendran, N.; PM, D.R.V. A deep learning framework with an embedded-based feature selection approach for the early
detection of the Alzheimer’s disease. Comput. Biol. Med. 2022, 141, 105056. [CrossRef] [PubMed]
65. Shankar, V.G.; Sisodia, D.S.; Chandrakar, P. A novel discriminant feature selection–based mutual information extraction from MR
brain images for Alzheimer’s stages detection and prediction. Int. J. Imaging Syst. Technol. 2022, 32, 1172–1191. [CrossRef]
66. Zhou, K.; Liu, Z.; He, W.; Cai, J.; Hu, L. Application of 3D Whole-Brain Texture Analysis and the Feature Selection Method Based
on within-Class Scatter in the Classification and Diagnosis of Alzheimer’s Disease. Ther. Innov. Regul. Sci. 2022, 56, 561–571.
[CrossRef]
67. Amrutesh, A.; CG, G.B.; Amruthamsh, A.; KP, A.R.; Gowrishankar, S. Alzheimer’s Disease Prediction using Machine Learning
and Transfer Learning Models. In Proceedings of the 2022 6th International Conference on Computation System and Information
Technology for Sustainable Solutions (CSITSS), Bangalore, India, 21–23 December 2022; pp. 1–6.
68. Weerasinghe, M.S.U. Pre-Detection of Dementia Using Machine Learning Mechanism. Master’s Thesis, University of Colombo,
Colombo, Sri Lanka, 2022.
69. Valliani, A.A.-A.; Ranti, D.; Oermann, E.K. Deep learning and neurology: A systematic review. Neurol. Ther. 2019, 8, 351–365.
[CrossRef]
70. Pang, G.; Shen, C.; Cao, L.; Van Den Hengel, A. Deep learning for anomaly detection: A review. ACM Comput. Surv. (CSUR) 2021,
54, 1–38. [CrossRef]
71. Jo, T.; Nho, K.; Saykin, A.J. Deep learning in Alzheimer’s disease: Diagnostic classification and prognostic prediction using
neuroimaging data. Front. Aging Neurosci. 2019, 11, 220. [CrossRef]
72. Wang, N.; Wang, Y.; Er, M.J. Review on deep learning techniques for marine object recognition: Architectures and algorithms.
Control Eng. Pract. 2020, 118, 104458. [CrossRef]
73. Francies, M.L.; Ata, M.M.; Mohamed, M.A. A robust multiclass 3D object recognition based on modern YOLO deep learning
algorithms. Concurr. Comput. Pract. Exp. 2022, 34, e6517. [CrossRef]
74. Balamurugan, D.; Aravinth, S.S.; Reddy, P.C.S.; Rupani, A.; Manikandan, A. Multiview objects recognition using deep learning-
based wrap-CNN with voting scheme. Neural Process. Lett. 2022, 54, 1495–1521. [CrossRef]
75. Ghasemi, Y.; Jeong, H.; Choi, S.H.; Park, K.-B.; Lee, J.Y. Deep learning-based object detection in augmented reality: A systematic
review. Comput. Ind. 2022, 139, 103661. [CrossRef]
76. Zaidi, S.S.A.; Ansari, M.S.; Aslam, A.; Kanwal, N.; Asghar, M.; Lee, B. A survey of modern deep learning based object detection
models. Digit. Signal Process. 2022, 126, 103514. [CrossRef]
Mach. Learn. Knowl. Extr. 2024, 6 497

77. Kang, J.; Tariq, S.; Oh, H.; Woo, S.S. A survey of deep learning-based object detection methods and datasets for overhead imagery.
IEEE Access 2022, 10, 20118–20134. [CrossRef]
78. Chu, P.; Wang, J.; You, Q.; Ling, H.; Liu, Z. Transmot: Spatial-temporal graph transformer for multiple object tracking. In
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA, 3–7 January
2023; pp. 4870–4880.
79. Tang, X.; Zhang, Z.; Qin, Y. On-road object detection and tracking based on radar and vision fusion: A review. IEEE Intell. Transp.
Syst. Mag. 2022, 14, 103–128. [CrossRef]
80. Vaquero, L.; Brea, V.M.; Mucientes, M. Tracking more than 100 arbitrary objects at 25 FPS through deep learning. Pattern Recognit.
2022, 121, 108205. [CrossRef]
81. Krentzel, D.; Shorte, S.L.; Zimmer, C. Deep learning in image-based phenotypic drug discovery. Trends Cell Biol. 2023, 33, 538–554.
[CrossRef]
82. Zhou, S.; Canchila, C.; Song, W. Deep learning-based crack segmentation for civil infrastructure: Data types, architectures, and
benchmarked performance. Autom. Constr. 2023, 146, 104678. [CrossRef]
83. He, S.; Bao, R.; Li, J.; Grant, P.E.; Ou, Y. Accuracy of segment-anything model (sam) in medical image segmentation tasks. arXiv
2023, arXiv:2304.09324.
84. Hourri, S.; Kharroubi, J. A deep learning approach for speaker recognition. Int. J. Speech Technol. 2020, 23, 123–131. [CrossRef]
85. Hourri, S.; Nikolov, N.S.; Kharroubi, J. A deep learning approach to integrate convolutional neural networks in speaker
recognition. Int. J. Speech Technol. 2020, 23, 615–623. [CrossRef]
86. Waldemar, G.; Dubois, B.; Emre, M.; Georges, J.; McKeith, I.G.; Rossor, M.; Scheltens, P.; Tariska, P.; Winblad, B. Recommendations
for the diagnosis and management of Alzheimer’s disease and other disorders associated with dementia: EFNS guideline. Eur. J.
Neurol. 2007, 14, e1–e26. [CrossRef]
87. Shenton, M.E.; Hamoda, H.M.; Schneiderman, J.S.; Bouix, S.; Pasternak, O.; Rathi, Y.; Vu, M.-A.; Purohit, M.P.; Helmer, K.; Koerte,
I.; et al. A review of magnetic resonance imaging and diffusion tensor imaging findings in mild traumatic brain injury. Brain
Imaging Behav. 2012, 6, 137–192. [CrossRef]
88. Matthews, P.M.; Peter, J. Functional magnetic resonance imaging. J. Neurol. Neurosurg. Psychiatry 2004, 75, 6–12.
89. Klunk, W.E.; Mathis, C.A.; Price, J.C.; DeKosky, S.T.; Lopresti, B.J.; Tsopelas, N.D.; Judith, A.S.; Robert, D.N. Amyloid imaging
with PET in Alzheimer’s disease, mild cognitive impairment, and clinically unimpaired subjects. In PET in the Evaluation of
Alzheimer’s Disease and Related Disorders; Springer: New York, NY, USA, 2009; pp. 119–147.
90. Basser, P.J. Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. NMR Biomed.
1995, 8, 333–344. [CrossRef] [PubMed]
91. Braak, H.; Alafuzoff, I.; Arzberger, T.; Kretzschmar, H.; Del Tredici, K. Staging of Alzheimer disease-associated neurofibrillary
pathology using paraffin sections and immunocytochemistry. Acta Neuropathol. 2006, 112, 389–404. [CrossRef] [PubMed]
92. Mueller, S.G.; Weiner, M.W.; Thal, L.J.; Petersen, R.C.; Jack, C.; Jagust, W.; Trojanowski, J.Q.; Toga, A.W.; Beckett, L. The
Alzheimer’s disease neuroimaging initiative. Neuroimaging Clin. 2005, 15, 869–877. [CrossRef] [PubMed]
93. Jagust, W. Imaging the evolution and pathophysiology of Alzheimer disease. Nat. Rev. Neurosci. 2018, 19, 687–700. [CrossRef]
[PubMed]
94. Weiner, M.W.; Veitch, D.P.; Aisen, P.S.; Beckett, L.A.; Cairns, N.J.; Green, R.C.; Harvey, D.; Jack, C.R.; Jagust, W.; Liu, E.; et al.
The Alzheimer’s Disease Neuroimaging Initiative: A review of papers published since its inception. Alzheimer’s Dement. 2013, 9,
e111–e194. [CrossRef]
95. Klein, A.; Andersson, J.; Ardekani, B.A.; Ashburner, J.; Avants, B.; Chiang, M.-C.; Christensen, G.E.; Collins, D.L.; Gee, J.; Hellier,
P.; et al. Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration. NeuroImage 2009, 46,
786–802. [CrossRef]
96. Ashburner, J.; Friston, K.J. Voxel-based morphometry—The methods. Neuroimage 2000, 11, 805–821. [CrossRef]
97. Choy, G.; Khalilzadeh, O.; Michalski, M.; Synho, D.; Samir, A.E.; Pianykh, O.S.; Geis, J.R.; Pandharipande, P.V.; Brink, J.A.; Dreyer,
K.J. Current applications and future impact of machine learning in radiology. Radiology 2018, 288, 318–328. [CrossRef] [PubMed]
98. Smith, S.M.; Jenkinson, M.; Woolrich, M.W.; Beckmann, C.F.; Behrens, T.E.; Johansen-Berg, H.; Bannister, P.R.; De Luca, M.;
Drobnjak, I.; Flitney, D.E.; et al. Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage
2004, 23, S208–S219. [CrossRef]
99. Jenkinson, M.; Bannister, P.; Brady, M.; Smith, S. Improved optimization for the robust and accurate linear registration and motion
correction of brain images. Neuroimage 2002, 17, 825–841. [CrossRef]
100. Hashem IA, T.; Yaqoob, I.; Anuar, N.B.; Mokhtar, S.; Gani, A.; Khan, S.U. The rise of “big data” on cloud computing: Review and
open research issues. Inf. Syst. 2015, 47, 98–115. [CrossRef]
101. Gorgolewski, K.J.; Auer, T.; Calhoun, V.D.; Craddock, R.C.; Das, S.; Duff, E.P.; Flandin, G.; Ghosh, S.S.; Glatard, T.; Halchenko,
Y.O.; et al. The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments. Sci.
Data 2016, 3, 160044. [CrossRef] [PubMed]
102. Kandel, S.; Heer, J.; Plaisant, C.; Kennedy, J.; van Ham, F.; Riche, N.H.; Weaver, C.; Lee, B.; Brodbeck, D.; Buono, P. Research
directions in data wrangling: Visualizations and transformations for usable and credible data. Inf. Vis. 2011, 10, 271–288.
[CrossRef]
103. LeCun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature 2015, 521, 436–444. [CrossRef] [PubMed]
Mach. Learn. Knowl. Extr. 2024, 6 498

104. Madrid, L.; Labrador, S.C.; González-Pérez, A.; Sáez, M.E.; Alzheimer’s Disease Neuroimaging Initiative (ADNI and others).
Integrated Genomic, Transcriptomic and Proteomic Analysis for Identifying Markers of Alzheimer’s Disease. Diagnostics 2021, 11,
2303. [CrossRef]
105. Konur, O.; Kingma, D.; Ba, J. Adam: A method for stochastic optimization. In Proceedings of the International Conference on
Learning Representations, ICLR 2015, San Diego, CA, USA, 7–9 May 2015; pp. 1–15. [CrossRef]
106. Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A simple way to prevent neural networks
from overfitting. J. Mach. Learn. Res. 2014, 15, 1929–1958.
107. Jiang, X.; Chang, L.; Zhang, Y.-D. Classification of Alzheimer’s disease via eight-layer convolutional neural network with batch
normalization and dropout techniques. J. Med. Imaging Health Inform. 2020, 10, 1040–1048. [CrossRef]
108. Pan, S.J.; Yang, Q. A survey on transfer learning. IEEE Trans. Knowl. Data Eng. 2010, 22, 1345–1359. [CrossRef]
109. Powers, D.M.W. Evaluation: From precision, recall and F-measure to ROC, informedness, markedness and correlation. arXiv
2020, arXiv:2010.16061.
110. Fawcett, T. An introduction to ROC analysis. Pattern Recogn. Lett. 2006, 27, 861–874. [CrossRef]
111. Kohavi, R. A study of cross-validation and bootstrap for accuracy estimation and model selection. Int. Jt. Conf. Artif. Intell. 1995,
14, 1137–1145.
112. Welton, T.; Kent, D.A.; Auer, D.P.; Dineen, R.A. Reproducibility of graph-theoretic brain network metrics: A systematic review.
Brain Connect. 2015, 5, 193–202. [CrossRef]
113. Moher, D.; Liberati, A.; Tetzlaff, J.; Altman, D.G.; PRISMA Group. Preferred reporting items for systematic reviews and
meta-analyses: The PRISMA statement. Ann. Intern. Med. 2009, 151, 264–269. [CrossRef] [PubMed]
114. De Leon, M.J.; DeSanti, S.; Zinkowski, R.; Mehta, P.D.; Pratico, D.; Segal, S.; Clark, C.; Kerkman, D.; DeBernardis, J.; Li, J.; et al.
MRI and CSF studies in the early diagnosis of Alzheimer’s disease. J. Intern. Med. 2004, 256, 205–223. [CrossRef]
115. Mendelson, A.F. Validating Supervised Learning Approaches to the Prediction of Disease Status in Neuroimaging. Ph.D. Thesis,
UCL (University College London), London, UK, 2017.
116. Yao, H.; Liu, Y.; Zhou, B.; Zhang, Z.; An, N.; Wang, P.; Wang, L.; Zhang, X.; Jiang, T. Decreased functional connectivity of the
amygdala in Alzheimer’s disease revealed by resting-state fMRI. Eur. J. Radiol. 2013, 82, 1531–1538. [CrossRef]
117. Miltiadous, A.; Tzimourta, K.D.; Giannakeas, N.; Tsipouras, M.G.; Afrantou, T.; Ioannidis, P.; Tzallas, A.T. Alzheimer’s disease
and frontotemporal dementia: A robust classification method of eeg signals and a comparison of validation methods. Diagnostics
2021, 11, 1437. [CrossRef]
118. Arena, F. Brain Network Analysis and Deep Learning Models for Studying Neurological Disorders Based on EEG Signal
Processing. Ph.D. Thesis, University of Messina, Messina, Italy, 2019.
119. Ge, Q.; Lin, Z.-C.; Gao, Y.-X.; Zhang, J.-X. A Robust Discriminant framework based on functional biomarkers of EEG and its
potential for diagnosis of Alzheimer’s disease. Healthcare 2020, 8, 476. [CrossRef]
120. Barzegaran, E.; van Damme, B.; Meuli, R.; Knyazeva, M.G. Perception-related EEG is more sensitive to Alzheimer’s disease
effects than resting EEG. Neurobiol. Aging 2016, 43, 129–139. [CrossRef]
121. Lu, L.; Sedor, J.R.; Gulani, V.; Schelling, J.R.; O’brien, A.; Flask, C.A.; Dell, K.M. Use of diffusion tensor MRI to Identify early
changes in diabetic nephropathy. Am. J. Nephrol. 2011, 34, 476–482. [CrossRef]
122. Guskiewicz, K.M.; Marshall, S.W.; Bailes, J.; McCrea, M.; Cantu, R.C.; Randolph, C.; Jordan, B.D. Association between recurrent
concussion and late-life cognitive impairment in retired professional football players. Neurosurgery 2005, 57, 719–726. [CrossRef]
[PubMed]
123. Ebrahimi, A.; Luo, S.; Chiong, R. Introducing transfer learning to 3D ResNet-18 for Alzheimer’s disease detection on MRI images.
In Proceedings of the 2020 35th International Conference on Image and Vision Computing New Zealand (IVCNZ), Wellington,
New Zealand, 25–27 November 2020; pp. 1–6.
124. Liu, Y.; Dwivedi, G.; Boussaid, F.; Sanfilippo, F.; Yamada, M.; Bennamoun, M. Inflating 2D Convolution Weights for Efficient
Generation of 3D Medical Images. Comput. Methods Programs Biomed. 2023, 240, 107685. [CrossRef] [PubMed]
125. Jang, J.; Hwang, D. M3T: Three-dimensional Medical image classifier using Multi-plane and Multi-slice Transformer. In
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, LA, USA, 18–24 June 2022;
pp. 20686–20697.
126. Poldrack, R.A. Region of interest analysis for fMRI. Soc. Cogn. Affect. Neurosci. 2007, 2, 67–70. [CrossRef] [PubMed]
127. Liu, M.; Zhang, J.; Nie, D.; Yap, P.-T.; Shen, D. Anatomical landmark based deep feature representation for MR images in brain
disease diagnosis. IEEE J. Biomed. Health Inform. 2018, 22, 1476–1485. [CrossRef] [PubMed]
128. Acharya, U.R.; Fernandes, S.L.; WeiKoh, J.E.; Ciaccio, E.J.; Fabell, M.K.M.; Tanik, U.J.; Rajinikanth, V.; Yeong, C.H. Automated
detection of Alzheimer’s disease using brain MRI images—A study with various feature extraction techniques. J. Med. Syst. 2019,
43, 302. [CrossRef]
129. Wang, X.; Liang, X.; Jiang, Z.; Nguchu, B.A.; Zhou, Y.; Wang, Y.; Wang, H.; Li, Y.; Zhu, Y.; Wu, F.; et al. Decoding and mapping
task states of the human brain via deep learning. Hum. Brain Mapp. 2020, 41, 1505–1519. [CrossRef]
130. Dua, S.; Srinivasan, P. A non-voxel based feature extraction to detect cognitive states in fMRI. In Proceedings of the 2008 30th Annual
International Conference of the IEEE Engineering in Medicine and Biology Society, Vancouver, BC, Canada, 20–25 August 2008;
pp. 4431–4434.
Mach. Learn. Knowl. Extr. 2024, 6 499

131. Shresta, S.; Senanayake, S.A.; Triloka, J. Advanced Cascaded Anisotropic Convolutional Neural Network Architecture Based
Optimized Feature Selection Brain Tumour Segmentation and Classification. In Proceedings of the 2020 5th International
Conference on Innovative Technologies in Intelligent Systems and Industrial Applications (CITISIA, Sydney, Australia,
25–27 November 2020; pp. 1–11.
132. Mahmood, R.; Ghimire, B. Automatic detection and classification of Alzheimer’s Disease from MRI scans using principal
component analysis and artificial neural networks. In Proceedings of the 2013 20th International Conference on Systems, Signals
and Image Processing (IWSSIP), Bucharest, Romania, 7–9 July 2013; pp. 133–137.
133. López, M.; Ramírez, J.; Górriz, J.M.; Álvarez, I.; Salas-Gonzalez, D.; Segovia, F.; Chaves, R.; Padilla, P.; Gómez-Río, M. Principal
component analysis-based techniques and supervised classification schemes for the early detection of Alzheimer’s disease.
Neurocomputing 2011, 74, 1260–1271. [CrossRef]
134. Rathore, S.; Habes, M.; Iftikhar, M.A.; Shacklett, A.; Davatzikos, C. A review on neuroimaging-based classification studies
and associated feature extraction methods for Alzheimer’s disease and its prodromal stages. NeuroImage 2017, 155, 530–548.
[CrossRef] [PubMed]
135. Cui, Y.; Liu, B.; Luo, S.; Zhen, X.; Fan, M.; Liu, T.; Zhu, W.; Park, M.; Jiang, T.; Jin, J.S.; et al. Identification of conversion from mild
cognitive impairment to Alzheimer’s disease using multivariate predictors. PLoS ONE 2011, 6, e21896. [CrossRef] [PubMed]
136. Jones, S.E.; Mahmoud, S.Y.; Phillips, M.D. A practical clinical method to quantify language lateralization in fMRI using whole-
brain analysis. NeuroImage 2011, 54, 2937–2949. [CrossRef] [PubMed]
137. Jack, C.R., Jr.; Bernstein, M.A.; Fox, N.C.; Thompson, P.; Alexander, G.; Harvey, D.; Borowski, B.; Whitwell, P.J.L.; Ward, J.; Dale,
A.M.; et al. The Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods. J. Magn. Reson. Imaging Off. J. Int. Soc. Magn.
Reson. Med. 2008, 27, 685–691. [CrossRef] [PubMed]
138. Marcus, D.S.; Wang, T.H.; Parker, J.; Csernansky, J.G.; Morris, J.C.; Buckner, R.L. Open Access Series of Imaging Studies (OASIS):
Cross-sectional MRI data in young, middle aged, nondemented, and demented older adults. J. Cogn. Neurosci. 2007, 19, 1498–1507.
[CrossRef] [PubMed]
139. Malone, I.B.; Cash, D.M.; Ridgway, G.R.; MacManus, D.G.; Ourselin, S.; Fox, N.C.; Schott, J.M. MIRIAD–Public release of a
multiple time point Alzheimer’s MR imaging dataset. NeuroImage 2013, 70, 33–36. [CrossRef]
140. Cardoso, B.R.; Hare, D.J.; Bush, A.I.; Li, Q.-X.; Fowler, C.J.; Masters, C.L.; Martins, R.N.; Ganio, K.; Lothian, A.; Mukherjee, S.; et al.
Selenium levels in serum, red blood cells, and cerebrospinal fluid of Alzheimer’s disease patients: A report from the Australian
Imaging, Biomarker & Lifestyle Flagship Study of Ageing (AIBL). J. Alzheimer’s Dis. 2017, 57, 183–193.
141. Gupta, Y.; Lee, K.H.; Choi, K.Y.; Lee, J.J.; Kim, B.C.; Kwon, G.R.; National Research Center for Dementia; Alzheimer’s Disease
Neuroimaging Initiative. Early diagnosis of Alzheimer’s disease using combined features from voxel-based morphometry and
cortical, subcortical, and hippocampus regions of MRI T1 brain images. PLoS ONE 2019, 14, e0222446. [CrossRef]
142. Katako, A.; Shelton, P.; Goertzen, A.L.; Levin, D.; Bybel, B.; Aljuaid, M.; Yoon, H.J.; Kang, D.Y.; Kim, S.M.; Lee, C.S.; et al.
Machine learning identified an Alzheimer’s disease-related FDG-PET pattern which is also expressed in Lewy body dementia
and Parkinson’s disease dementia. Sci. Rep. 2018, 8, 13236. [CrossRef]
143. Freitas, S.; Simões, M.R.; Alves, L.; Santana, I. Montreal cognitive assessment: Validation study for mild cognitive impairment
and Alzheimer disease. Alzheimer Dis. Assoc. Disord. 2013, 27, 37–43. [CrossRef]
144. Desikan, R.S.; Cabral, H.J.; Settecase, F.; Hess, C.P.; Dillon, W.P.; Glastonbury, C.M.; Weiner, M.W.; Schmansky, N.J.; Salat, D.H.;
Fischl, B. Automated MRI measures predict progression to Alzheimer’s disease. Neurobiol. Aging 2010, 31, 1364–1374. [CrossRef]
[PubMed]
145. Sharma, R.; Anand, H.; Badr, Y.; Qiu, R.G. Time-to-event prediction using survival analysis methods for Alzheimer’s disease
progression. Alzheimer’s Dement. Transl. Res. Clin. Interv. 2021, 7, e12229. [CrossRef] [PubMed]
146. Khojaste-Sarakhsi, M.; Haghighi, S.S.; Ghomi, S.F.; Marchiori, E. Deep learning for Alzheimer’s disease diagnosis: A survey. Artif.
Intell. Med. 2022, 130, 102332. [CrossRef] [PubMed]
147. Huang, H.-C.; Jiang, Z.-F. Accumulated amyloid-β peptide and hyperphosphorylated tau protein: Relationship and links in
Alzheimer’s disease. J. Alzheimer’s Dis. 2009, 16, 15–27. [CrossRef]
148. Johnson, K.A.; Fox, N.C.; Sperling, R.A.; Klunk, W.E. Brain imaging in Alzheimer disease. Cold Spring Harb. Perspect. Med. 2012, 2,
a006213. [CrossRef] [PubMed]
149. Thambisetty, M.; An, Y.; Kinsey, A.; Koka, D.; Saleem, M.; Güntert, A.; Kraut, M.; Ferrucci, L.; Davatzikos, C.; Lovestone, S.; et al.
Plasma clusterin concentration is associated with longitudinal brain atrophy in mild cognitive impairment. Neuroimage 2012, 59,
212–217. [CrossRef]
150. Storelli, L.; Azzimonti, M.; Gueye, M.; Vizzino, C.M.; Preziosa, P.M.; Tedeschi, G.; De Stefano, N.M.; Pantano, P.M.; Filippi, M.;
Rocca, M.A. A deep learning approach to predicting disease progression in multiple sclerosis using magnetic resonance imaging.
Investig. Radiol. 2022, 57, 423–432. [CrossRef]
151. Kumar, P.; Nagar, P.; Arora, C.; Gupta, A. U-segnet: Fully convolutional neural network based automated brain tissue seg-
mentation tool. In Proceedings of the 2018 25th IEEE International Conference on Image Processing (ICIP), Athens, Greece,
7–10 October 2018; pp. 3503–3507. [CrossRef]
152. Narayana, P.A.; Coronado, I.; Sujit, S.J.; Sun, X.; Wolinsky, J.S.; Gabr, R.E. Are multi-contrast magnetic resonance images necessary
for segmenting multiple sclerosis brains? A large cohort study based on deep learning. Magn. Reson. Imaging 2020, 65, 8–14.
[CrossRef]
Mach. Learn. Knowl. Extr. 2024, 6 500

153. Narayana, P.A.; Coronado, I.; Sujit, S.J.; Wolinsky, J.S.; Lublin, F.D.; Gabr, R.E. Deep-learning-based neural tissue segmentation of
MRI in multiple sclerosis: Effect of training set size. J. Magn. Reson. Imaging 2020, 51, 1487–1496. [CrossRef]
154. Zhang, J.; Zheng, B.; Gao, A.; Feng, X.; Liang, D.; Long, X. A 3D densely connected convolution neural network with connection-
wise attention mechanism for Alzheimer’s disease classification. Magn. Reson. Imaging 2021, 78, 119–126. [CrossRef]
155. Ben Ahmed, O.; Mizotin, M.; Benois-Pineau, J.; Allard, M.; Catheline, G.; Ben Amar, C.; Alzheimer’s Disease Neuroimaging
Initiative. Alzheimer’s disease diagnosis on structural MR images using circular harmonic functions descriptors on hippocampus
and posterior cingulate cortex. Comput. Med. Imaging Graph. 2015, 44, 13–25. [CrossRef]
156. Chen, L.; Qiao, H.; Zhu, F. Alzheimer’s Disease Diagnosis With Brain Structural MRI Using Multiview-Slice Attention and 3D
Convolution Neural Network. Front. Aging Neurosci. 2022, 14, 871706. [CrossRef]
157. Li, Z.; Liu, F.; Yang, W.; Peng, S.; Zhou, J. A survey of convolutional neural networks: Analysis, applications, and prospects. IEEE
Trans. Neural Netw. Learn. Syst. 2021, 33, 6999–7019. [CrossRef]
158. Bin Bae, J.; Lee, S.; Jung, W.; Park, S.; Kim, W.; Oh, H.; Han, J.W.; Kim, G.E.; Kim, J.S.; Kim, J.H.; et al. Identification of Alzheimer’s
disease using a convolutional neural network model based on T1-weighted magnetic resonance imaging. Sci. Rep. 2020, 10, 22252.
[CrossRef]
159. Aderghal, K.; Benois-Pineau, J.; Afdel, K. Classification of sMRI for Alzheimer’s disease diagnosis with CNN: Single Siamese
networks with 2D+? Approach and fusion on ADNI. In Proceedings of the 2017 ACM on International Conference on Multimedia
Retrieval, Bucharest, Romania, 6–9 June 2017; pp. 494–498.
160. Khvostikov, A.; Aderghal, K.; Krylov, A.; Catheline, G.; Benois-Pineau, J. 3D inception-based CNN with sMRI and MD-DTI data
fusion for Alzheimer’s disease diagnostics. arXiv 2018, arXiv:1809.03972.
161. Khagi, B.; Lee, C.G.; Kwon, G.-R. Alzheimer’s disease classification from brain MRI based on transfer learning from CNN.
In Proceedings of the 2018 11th Biomedical Engineering International Conference (BMEiCON), Chiang Mai, Thailand,
21–24 November 2018; pp. 1–4.
162. Sahumbaiev, I.; Popov, A.; Ramirez, J.; Gorriz, J.M.; Ortiz, A. 3D-CNN HadNet classification of MRI for Alzheimer’s Disease
diagnosis. In Proceedings of the 2018 IEEE Nuclear Science Symposium and Medical Imaging Conference Proceedings (NSS/MIC),
Sydney, NSW, Australia, 10–17 November 2018; pp. 1–4.
163. Roy, S.S.; Sikaria, R.; Susan, A. A deep learning based CNN approach on MRI for Alzheimer’s disease detection. Intell. Decis.
Technol. 2019, 13, 495–505. [CrossRef]
164. Kruthika, K.R.; Maheshappa, H.D.; Alzheimer’s Disease Neuroimaging Initiative. CBIR system using Capsule Networks and 3D
CNN for Alzheimer’s disease diagnosis. Inform. Med. Unlocked 2019, 14, 59–68. [CrossRef]
165. Huang, Y.; Xu, J.; Zhou, Y.; Tong, T.; Zhuang, X.; Alzheimer’s Disease Neuroimaging Initiative (ADNI). Diagnosis of Alzheimer’s
disease via multi-modality 3D convolutional neural network. Front. Neurosci. 2019, 13, 509. [CrossRef] [PubMed]
166. Jain, R.; Jain, N.; Aggarwal, A.; Hemanth, D.J. Convolutional neural network based Alzheimer’s disease classification from
magnetic resonance brain images. Cogn. Syst. Res. 2019, 57, 147–159. [CrossRef]
167. Folego, G.; Weiler, M.; Casseb, R.F.; Pires, R.; Rocha, A. Alzheimer’s disease detection through whole-brain 3D-CNN MRI. Front.
Bioeng. Biotechnol. 2020, 8, 534592. [CrossRef] [PubMed]
168. Dua, M.; Makhija, D.; Manasa, P.Y.; Mishra, P. A CNN–RNN–LSTM based amalgamation for Alzheimer’s disease detection. J.
Med. Biol. Eng. 2020, 40, 688–706. [CrossRef]
169. Hussain, E.; Hasan, M.; Hassan, S.Z.; Azmi, T.H.; Rahman, A.; Parvez, M.Z. Deep learning based binary classification for
Alzheimer’s disease detection using brain mri images. In Proceedings of the 2020 15th IEEE Conference on Industrial Electronics
and Applications (ICIEA), Kristiansand, Norway, 9–13 November 2020; pp. 1115–1120.
170. Xia, Z.; Yue, G.; Xu, Y.; Feng, C.; Yang, M.; Wang, T.; Lei, B. A novel end-to-end hybrid network for Alzheimer’s disease detection
using 3D CNN and 3D CLSTM. In Proceedings of the 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI),
Iowa City, IA, USA, 3–7 April 2020; pp. 1–4.
171. Wen, J.; Thibeau-Sutre, E.; Diaz-Melo, M.; Samper-González, J.; Routier, A.; Bottani, S.; Dormont, D.; Durrleman, S.; Burgos, N.;
Colliot, O. Convolutional neural networks for classification of Alzheimer’s disease: Overview and reproducible evaluation. Med.
Image Anal. 2020, 63, 101694. [CrossRef]
172. Liu, S.; Yadav, C.; Fernandez-Granda, C.; Razavian, N. On the design of convolutional neural networks for automatic detection of
Alzheimer’s disease. In Machine Learning for Health Workshop; PMLR: Vancouver, BC, Canada, 2020; pp. 184–201.
173. Liu, M.; Li, F.; Yan, H.; Wang, K.; Ma, Y.; Alzheimer’s Disease Neuroimaging Initiative; Shen, L.; Xu, M. A multi-model deep
convolutional neural network for automatic hippocampus segmentation and classification in Alzheimer’s disease. NeuroImage
2020, 208, 116459. [CrossRef]
174. Pan, D.; Zeng, A.; Jia, L.; Huang, Y.; Frizzell, T.; Song, X. Early detection of Alzheimer’s Disease using magnetic resonance
imaging: A novel approach combining convolutional neural networks and ensemble learning. Front. Neurosci. 2020, 14, 259.
[CrossRef]
175. Ramana, T.V. Alzheimer disease detection and classification on magnetic resonance imaging (MRI) brain images using improved
expectation maximization (IEM) and convolutional neural network (CNN). Turk. J. Comput. Math. Educ. (TURCOMAT) 2021, 12,
5998–6006.
Mach. Learn. Knowl. Extr. 2024, 6 501

176. Miah, A.S.M.; Rashid, M.M.; Rahman, M.R.; Hossain, M.T.; Sujon, M.S.; Nawal, N.; Hasan, M.; Shin, J. Alzheimer’s disease
detection using CNN based on effective dimensionality reduction approach. In Intelligent Computing and Optimization: Proceedings
of the 3rd International Conference on Intelligent Computing and Optimization 2020 (ICO 2020); Springer: Berlin/Heidelberg, Germany,
2021; pp. 801–811.
177. AbdulAzeem, Y.; Bahgat, W.M.; Badawy, M. A CNN based framework for classification of Alzheimer’s disease. Neural Comput.
Appl. 2021, 33, 10415–10428. [CrossRef]
178. Al-Khuzaie, F.E.K.; Bayat, O.; Duru, A.D. Diagnosis of Alzheimer disease using 2D MRI slices by convolutional neural network.
Appl. Bionics Biomech. 2021, 2021, 6690539. [CrossRef]
179. Janghel, R.; Rathore, Y. Deep convolution neural network based system for early diagnosis of Alzheimer’s disease. IRBM 2021, 42,
258–267. [CrossRef]
180. Kang, W.; Lin, L.; Zhang, B.; Shen, X.; Wu, S. Multi-model and multi-slice ensemble learning architecture based on 2D convolu-
tional neural networks for Alzheimer’s disease diagnosis. Comput. Biol. Med. 2021, 136, 104678. [CrossRef]
181. Basher, A.; Kim, B.C.; Lee, K.H.; Jung, H.Y. Volumetric feature-based Alzheimer’s disease diagnosis from sMRI data using a
convolutional neural network and a deep neural network. IEEE Access 2021, 9, 29870–29882. [CrossRef]
182. Singh, A.; Kharkar, N.; Priyanka, P.; Parvartikar, S. Alzheimer’s disease detection using deep learning-CNN. In Ambient
Communications and Computer Systems: Proceedings of RACCCS 2021; Springer: Berlin/Heidelberg, Germany, 2022; pp. 529–537.
183. Davuluri, R.; Rengaswamy, R. Improved Classification Model using CNN for Detection of Alzheimer’s Disease. J. Comput. Sci.
2022, 18, 415–425. [CrossRef]
184. Angkoso, C.V.; Tjahyaningtijas, H.P.A.; Purnama, I.; Purnomo, M.H. Multiplane Convolutional Neural Network (Mp-CNN) for
Alzheimer’s Disease Classification. Int. J. Intell. Eng. Syst. 2022, 15, 329–340.
185. Techa, C.; Ridouani, M.; Hassouni, L.; Anoun, H. Alzheimer’s disease multi-class classification model based on CNN and StackNet
using brain MRI data. In International Conference on Advanced Intelligent Systems and Informatics; Springer: Berlin/Heidelberg,
Germany, 2022; pp. 248–259.
186. Divya, R.; Kumari, R.S.S.; Alzheimer’s Disease Neuroimaging Initiative. Detection of Alzheimer’s disease from temporal lobe
grey matter slices using 3D CNN. Imaging Sci. J. 2022, 70, 578–587. [CrossRef]
187. Dar, G.M.U.D.; Bhagat, A.; Ansarullah, S.I.; Ben Othman, M.T.; Hamid, Y.; Alkahtani, H.K.; Ullah, I.; Hamam, H. A novel
framework for classification of different Alzheimer’s disease stages using CNN model. Electronics 2023, 12, 469. [CrossRef]
188. Lanjewar, M.G.; Parab, J.S.; Shaikh, A.Y. Development of framework by combining CNN with KNN to detect Alzheimer’s disease
using MRI images. Multimed. Tools Appl. 2023, 82, 12699–12717. [CrossRef]
189. Khalid, A.; Senan, E.M.; Al-Wagih, K.; Al-Azzam, M.M.A.; Alkhraisha, Z.M. Automatic Analysis of MRI Images for Early
Prediction of Alzheimer’s Disease Stages Based on Hybrid Features of CNN and Handcrafted Features. Diagnostics 2023, 13, 1654.
[CrossRef] [PubMed]
190. Ravikumar, R.; Sasipriyaa, N.; Thilagaraj, T.; Raj, R.H.; Abishek, A.; Kannan, G.G. Design and Implementation of Alzheimer’s
Disease Detection using cGAN and CNN. In Proceedings of the 2023 International Conference on Computer Communication and
Informatics (ICCCI), Coimbatore, India, 23–25 January 2023; pp. 1–7.
191. Shastri, S.S.; Bhadrashetty, A.; Kulkarni, S. Detection and Classification of Alzheimer’s Disease by Employing CNN. Int. J. Intell.
Syst. Appl. 2023, 15, 14–22. [CrossRef]
192. Venkatasubramanian, S.; Dwivedi, J.N.; Raja, S.; Rajeswari, N.; Logeshwaran, J.; Kumar, A.P. Prediction of Alzheimer’s Disease
Using DHO-Based Pretrained CNN Model. Math. Probl. Eng. 2023, 2023, 1110500. [CrossRef]
193. Ilesanmi, A.E.; Ilesanmi, T.; Idowu, O.P.; Torigian, D.A.; Udupa, J.K. Organ segmentation from computed tomography images
using the 3D convolutional neural network: A systematic review. Int. J. Multimed. Inf. Retr. 2022, 11, 315–331. [CrossRef]
194. Irshad, S.; Gomes, D.P.S.; Kim, S.T. Improved abdominal multi-organ segmentation via 3d boundary-constrained deep neural
networks. IEEE Access 2023, 11, 35097–35110. [CrossRef]
195. Rickmann, A.-M.; Senapati, J.; Kovalenko, O.; Peters, A.; Bamberg, F.; Wachinger, C. AbdomenNet: Deep neural network for
abdominal organ segmentation in epidemiologic imaging studies. BMC Med. Imaging 2022, 22, 1–11. [CrossRef] [PubMed]
196. Hassan, S.M.; Maji, A.K. Plant disease identification using a novel convolutional neural network. IEEE Access 2022, 10, 5390–5401.
[CrossRef]
197. Ashwinkumar, S.; Rajagopal, S.; Manimaran, V.; Jegajothi, B. Automated plant leaf disease detection and classification using
optimal MobileNet based convolutional neural networks. Mater. Today Proc. 2022, 51, 480–487. [CrossRef]
198. Jiang, J.; Liu, H.; Zhao, C.; He, C.; Ma, J.; Cheng, T.; Zhu, Y.; Cao, W.; Yao, X. Evaluation of Diverse Convolutional Neural
Networks and Training Strategies for Wheat Leaf Disease Identification with Field-Acquired Photographs. Remote Sens. 2022,
14, 3446. [CrossRef]
199. Nirthika, R.; Manivannan, S.; Ramanan, A.; Wang, R. Pooling in convolutional neural networks for medical image analysis: A
survey and an empirical study. Neural Comput. Appl. 2022, 34, 5321–5347. [CrossRef]
200. Mattia, G.M.; Sarton, B.; Villain, E.; Vinour, H.; Ferre, F.; Buffieres, W.; Le Lann, M.-V.; Franceries, X.; Peran, P.; Silva, S.
Multimodal MRI-Based Whole-Brain Assessment in Patients In Anoxoischemic Coma by Using 3D Convolutional Neural
Networks. Neurocritical Care 2022, 37 (Suppl. S2), 303–312. [CrossRef]
201. Warren, S.L.; Moustafa, A.A. Functional magnetic resonance imaging, deep learning, and Alzheimer’s disease: A systematic
review. J. Neuroimaging 2023, 33, 5–18. [CrossRef]
Mach. Learn. Knowl. Extr. 2024, 6 502

202. Dan, T.; Huang, Z.; Cai, H.; Laurienti, P.J.; Wu, G. Learning brain dynamics of evolving manifold functional MRI data using
geometric-attention neural network. IEEE Trans. Med. Imaging 2022, 41, 2752–2763. [CrossRef]
203. Guo, X.; Zhou, B.; Pigg, D.; Spottiswoode, B.; Casey, M.E.; Liu, C.; Dvornek, N.C. Unsupervised inter-frame motion correction for
whole-body dynamic PET using convolutional long short-term memory in a convolutional neural network. Med. Image Anal.
2022, 80, 102524. [CrossRef] [PubMed]
204. Bin Tufail, A.; Anwar, N.; Ben Othman, M.T.; Ullah, I.; Khan, R.A.; Ma, Y.-K.; Adhikari, D.; Rehman, A.U.; Shafiq, M.; Hamam, H.
Early-stage Alzheimer’s disease categorization using PET neuroimaging modality and convolutional neural networks in the 2d
and 3d domains. Sensors 2022, 22, 4609. [CrossRef]
205. Estudillo-Romero, A.; Haegelen, C.; Jannin, P.; Baxter, J.S.H. Voxel-based diktiometry: Combining convolutional neural networks
with voxel-based analysis and its application in diffusion tensor imaging for Parkinson’s disease. Hum. Brain Mapp. 2022, 43,
4835–4851. [CrossRef]
206. Liu, S.; Liu, Y.; Xu, X.; Chen, R.; Liang, D.; Jin, Q.; Liu, H.; Chen, G.; Zhu, Y. Accelerated cardiac diffusion tensor imaging using
deep neural network. Phys. Med. Biol. 2022, 68, 025008. [CrossRef]
207. Park, S.; Yu, J.; Woo, H.-H.; Park, C.G. A novel network architecture combining central-peripheral deviation with image-based
convolutional neural networks for diffusion tensor imaging studies. J. Appl. Stat. 2022, 50, 3294–3311. [CrossRef]
208. Ghazal, T.M.; Abbas, S.; Munir, S.; Ahmad, M.; Issa, G.F.; Zahra, S.B.; Khan, M.A.; Hasan, M.K. Alzheimer disease detection
empowered with transfer learning. Comput. Mater. Contin. 2022, 70, 5005–5019. [CrossRef]
209. Liu, S.; Masurkar, A.V.; Rusinek, H.; Chen, J.; Zhang, B.; Zhu, W.; Fernandez-Granda, C.; Razavian, N. Generalizable deep
learning model for early Alzheimer’s disease detection from structural MRIs. Sci. Rep. 2022, 12, 17106. [CrossRef]
210. Loddo, A.; Buttau, S.; Di Ruberto, C. Deep learning based pipelines for Alzheimer’s disease diagnosis: A comparative study and
a novel deep-ensemble method. Comput. Biol. Med. 2021, 141, 105032. [CrossRef]
211. Cui, R.; Liu, M.; Li, G. Longitudinal analysis for Alzheimer’s disease diagnosis using RNN. In Proceedings of the 2018 IEEE 15th
International Symposium on Biomedical Imaging (ISBI 2018), Washington, DC, USA, 4–7 April 2018; 2018; pp. 1398–1401.
212. Feng, C.; Elazab, A.; Yang, P.; Wang, T.; Lei, B.; Xiao, X. 3D convolutional neural network and stacked bidirectional recurrent
neural network for Alzheimer’s disease diagnosis. In PRedictive Intelligence in MEdicine: First International Workshop, PRIME 2018,
Held in Conjunction with MICCAI 2018, Granada, Spain, September 16, 2018, Proceedings 1; Springer: Berlin/Heidelberg, Germany,
2018; pp. 138–146.
213. Nguyen, M.; Sun, N.; Alexander, D.C.; Feng, J.; Yeo, B.T.T. Modeling Alzheimer’s disease progression using deep recurrent neural
networks. In Proceedings of the 2018 International Workshop on Pattern Recognition in Neuroimaging (PRNI), Singapore, 12–14
June 2018; pp. 1–4.
214. Wang, T.; Qiu, R.G.; Yu, M. Predictive modeling of the progression of Alzheimer’s disease with recurrent neural networks. Sci.
Rep. 2018, 8, 9161. [CrossRef]
215. Liu, M.; Cheng, D.; Yan, W.; Alzheimer’s Disease Neuroimaging Initiative. Classification of Alzheimer’s disease by combination
of convolutional and recurrent neural networks using FDG-PET images. Front. Aging Neurosci. 2018, 12, 35. [CrossRef]
216. Cui, R.; Liu, M.; Alzheimer’s Disease Neuroimaging Initiative. RNN-based longitudinal analysis for diagnosis of Alzheimer’s
disease. Comput. Med. Imaging Graph. 2019, 73, 1–10. [CrossRef]
217. Li, F.; Liu, M.; The Alzheimer’s Disease Neuroimaging Initiative. A hybrid convolutional and recurrent neural network for
hippocampus analysis in Alzheimer’s disease. J. Neurosci. Methods 2019, 323, 108–118. [CrossRef]
218. Velazquez, M.; Anantharaman, R.; Velazquez, S.; Lee, Y. RNN-based Alzheimer’s disease prediction from prodromal stage using
diffusion tensor imaging. In Proceedings of the 2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM),
San Diego, CA, USA, 18–21 November 2019; pp. 1665–1672.
219. Jabason, E.; Ahmad, M.O.; Swamy, M. Hybrid Feature Fusion Using RNN and Pre-trained CNN for Classification of Alzheimer’s
Disease (Poster). In Proceedings of the 2019 22th International Conference on Information Fusion (FUSION), Ottawa, ON, Canada,
2–5 July 2019.
220. Hong, X.; Lin, R.; Yang, C.; Zeng, N.; Cai, C.; Gou, J.; Yang, J. Predicting Alzheimer’s disease using LSTM. IEEE Access 2019, 7,
80893–80901. [CrossRef]
221. Feng, C.; Elazab, A.; Yang, P.; Wang, T.; Zhou, F.; Hu, H.; Xiao, X.; Lei, B. Deep learning framework for Alzheimer’s disease
diagnosis via 3D-CNN and FSBi-LSTM. IEEE Access 2019, 7, 63605–63618. [CrossRef]
222. Pan, Q.; Wang, S.; Zhang, J. Prediction of Alzheimer’s disease based on bidirectional LSTM. J. Phys. Conf. Ser. 2019, 1187, 052030.
[CrossRef]
223. Ghazi, M.M.; Nielsen, M.; Pai, A.; Cardoso, M.J.; Modat, M.; Ourselin, S.; Sørensen, L. Training recurrent neural networks robust
to incomplete data: Application to Alzheimer’s disease progression modeling. Med. Image Anal. 2019, 53, 39–46. [CrossRef]
224. Li, H.; Fan, Y. Early Prediction Of Alzheimer’s disease dementia based on baseline hippocampal mri and 1-year follow-up
cognitive measures using deep recurrent neural networks. In Proceedings of the 2019 IEEE 16th International Symposium on
Biomedical Imaging (ISBI), Venice, Italy, 8–11 April 2019; pp. 368–371.
225. Tabarestani, S.; Aghili, M.; Shojaie, M.; Freytes, C.; Cabrerizo, M.; Barreto, A.; Rishe, N.; Curiel, R.E.; Loewenstein, D.; Duara, R.;
et al. Longitudinal prediction modeling of Alzheimer disease using recurrent neural networks. In Proceedings of the 2019 IEEE
EMBS International Conference on Biomedical & Health Informatics (BHI), Chicago, IL, USA, 19–22 May 2019; pp. 1–4.
Mach. Learn. Knowl. Extr. 2024, 6 503

226. Nguyen, M.; He, T.; An, L.; Alexander, D.C.; Feng, J.; Yeo, B.T. Predicting Alzheimer’s disease progression using deep recurrent
neural networks. NeuroImage 2020, 222, 117203. [CrossRef]
227. Jung, W.; Jun, E.; Suk, H.-I.; Alzheimer’s Disease Neuroimaging Initiative. Deep recurrent model for individualized prediction of
Alzheimer’s disease progression. NeuroImage 2021, 237, 118143. [CrossRef]
228. Liang, W.; Zhang, K.; Cao, P.; Liu, X.; Yang, J.; Zaiane, O. Rethinking modeling Alzheimer’s disease progression from a multi-task
learning perspective with deep recurrent neural network. Comput. Biol. Med. 2021, 138, 104935. [CrossRef]
229. Aqeel, A.; Hassan, A.; Khan, M.A.; Rehman, S.; Tariq, U.; Kadry, S.; Majumdar, A.; Thinnukool, O. A long short-term memory
biomarker-based prediction framework for Alzheimer’s disease. Sensors 2022, 22, 1475. [CrossRef]
230. Lin, K.; Jie, B.; Dong, P.; Ding, X.; Bian, W.; Liu, M. Convolutional recurrent neural network for dynamic functional MRI analysis
and brain disease identification. Front. Neurosci. 2022, 16, 933660. [CrossRef]
231. Algani, Y.M.A.; Vidhya, S.; Ghai, B.; Acharjee, P.B.; Kathiravan, M.N.; Dwivedi, V.K. Innovative Method for Alzheimer Disease
Prediction using GP-ELM-RNN. In Proceedings of the 2023 2nd International Conference on Applied Artificial Intelligence and
Computing (ICAAIC), Salem, India, 4–6 May 2023; pp. 723–728.
232. Pan, Y.; Liu, M.; Lian, C.; Zhou, T.; Xia, Y.; Shen, D. Synthesizing missing PET from MRI with cycle-consistent generative
adversarial networks for Alzheimer’s disease diagnosis. In Medical Image Computing and Computer Assisted Intervention–MICCAI
2018: 21st International Conference, Granada, Spain, September 16–20, 2018, Proceedings, Part III 11; Springer: Berlin/Heidelberg,
Germany, 2018; pp. 455–463.
233. Bowles, C.; Gunn, R.; Hammers, A.; Rueckert, D. Modelling the progression of Alzheimer’s disease in MRI using generative
adversarial networks. In Medical Imaging 2018: Image Processing; SPIE: Houston, TX, USA, 2018; Volume 10574, pp. 397–407.
234. Wegmayr, V.; Hörold, M.; Buhmann, J.M. Generative aging of brain MR-images and prediction of Alzheimer progression. In
Pattern Recognition: 41st DAGM German Conference, DAGM GCPR 2019, Dortmund, Germany, September 10–13, 2019, Proceedings 41;
Springer: Berlin/Heidelberg, Germany, 2019; pp. 247–260.
235. Zhao, Y.; Ma, B.; Jiang, P.; Zeng, D.; Wang, X.; Li, S. Prediction of Alzheimer’s disease progression with multi-information
generative adversarial network. IEEE J. Biomed. Health Inform. 2020, 25, 711–719. [CrossRef] [PubMed]
236. Hu, S.; Yu, W.; Chen, Z.; Wang, S. Medical image reconstruction using generative adversarial network for Alzheimer disease
assessment with class-imbalance problem. In Proceedings of the 2020 IEEE 6th International Conference on Computer and
Communications (ICCC), Chengdu, China, 11–14 December 2020; pp. 1323–1327.
237. Ma, D.; Lu, D.; Popuri, K.; Wang, L.; Beg, M.F.; Alzheimer’s Disease Neuroimaging Initiative. Differential diagnosis of
frontotemporal dementia, Alzheimer’s disease, and normal aging using a multi-scale multi-type feature generative adversarial
deep neural network on structural magnetic resonance images. Front. Neurosci. 2020, 14, 853. [CrossRef] [PubMed]
238. Shin, H.C.; Ihsani, A.; Xu, Z.; Mandava, S.; Sreenivas, S.T.; Forster, C.; Cha, J.; Alzheimer’s Disease Neuroimaging Initiative.
GANDALF: Generative adversarial networks with discriminator-adaptive loss fine-tuning for Alzheimer’s disease diagnosis
from MRI. In Medical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru,
October 4–8, 2020, Proceedings, Part II 23; Springer: Berlin/Heidelberg, Germany, 2020; pp. 688–697.
239. Kim, H.W.; Lee, H.E.; Lee, S.; Oh, K.T.; Yun, M.; Yoo, S.K. Slice-selective learning for Alzheimer’s disease classification using a
generative adversarial network: A feasibility study of external validation. Eur. J. Nucl. Med. 2020, 47, 2197–2206. [CrossRef]
240. Jung, E.; Luna, M.; Park, S.H. Conditional generative adversarial network for predicting 3d medical images affected by
Alzheimer’s diseases. In Predictive Intelligence in Medicine: Third International Workshop, PRIME 2020, Held in Conjunction
with MICCAI 2020, Lima, Peru, October 8, 2020, Proceedings 3; Springer: Berlin/Heidelberg, Germany, 2020; pp. 79–90.
241. Zhou, X.; Qiu, S.; Joshi, P.S.; Xue, C.; Killiany, R.J.; Mian, A.Z.; Chin, S.P.; Au, R.; Kolachalama, V.B. Enhancing magnetic resonance
imaging-driven Alzheimer’s disease classification performance using generative adversarial learning. Alzheimer’s Res. Ther. 2021,
13, 497. [CrossRef]
242. Logan, R.; Williams, B.G.; da Silva, M.F.; Indani, A.; Schcolnicov, N.; Ganguly, A.; Miller, S.J. Deep convolutional neural networks
with ensemble learning and generative adversarial networks for Alzheimer’s disease image data classification. Front. Aging
Neurosci. 2021, 13, 720226. [CrossRef]
243. Sajjad, M.; Ramzan, F.; Khan, M.U.G.; Rehman, A.; Kolivand, M.; Fati, S.M.; Bahaj, S.A. Deep convolutional generative adversarial
network for Alzheimer’s disease classification using positron emission tomography (PET) and synthetic data augmentation.
Microsc. Res. Tech. 2021, 84, 3023–3034. [CrossRef]
244. Sinha, S.; Thomopoulos, S.I.; Lam, P.; Muir, A.; Thompson, P.M. Alzheimer’s disease classification accuracy is improved by MRI
harmonization based on attention-guided generative adversarial networks. In Proceedings of the 17th International Symposium
on Medical Information Processing and Analysis, Campinas, Brazil, 17–19 November 2021; SPIE: Bellingham, WA, USA, 2021;
pp. 180–189.
245. Qu, C.; Zou, Y.; Ma, Y.; Chen, Q.; Luo, J.; Fan, H.; Jia, Z.; Gong, Q.; Chen, T. Diagnostic performance of generative adversarial
network-based deep learning methods for Alzheimer’s disease: A systematic review and meta-analysis. Front. Aging Neurosci.
2022, 14, 841696. [CrossRef]
246. Zhang, J.; He, X.; Qing, L.; Gao, F.; Wang, B. BPGAN: Brain PET synthesis from MRI using generative adversarial network for
multi-modal Alzheimer’s disease diagnosis. Comput. Methods Programs Biomed. 2022, 217, 106676. [CrossRef] [PubMed]
247. Ye, H.; Zhu, Q.; Yao, Y.; Jin, Y.; Zhang, D. Pairwise feature-based generative adversarial network for incomplete multi-modal
Alzheimer’s disease diagnosis. Vis. Comput. 2022, 39, 2235–2244. [CrossRef]
Mach. Learn. Knowl. Extr. 2024, 6 504

248. Cabreza, J.N.; Solano, G.A.; Ojeda, S.A.; Munar, V. Anomaly detection for Alzheimer’s disease in brain MRIS via unsupervised
generative adversarial learning. In Proceedings of the 2022 International Conference on Artificial Intelligence in Information and
Communication (ICAIIC), Jeju Island, Republic of Korea, 21–24 February 2022; pp. 1–5.
249. Dolci, G.; Rahaman, A.; Chen, J.; Duan, K.; Fu, Z.; Abrol, A.; Menegaz, G.; Calhoun, V.D. A deep generative multimodal imaging
genomics framework for Alzheimer’s disease prediction. In Proceedings of the 2022 IEEE 22nd International Conference on
Bioinformatics and Bioengineering (BIBE), Taichung, Taiwan, 7–9 November 2022; pp. 41–44.
250. Bi, X.-A.; Wang, Y.; Luo, S.; Chen, K.; Xing, Z.; Xu, L. Hypergraph Structural Information Aggregation Generative Adversarial
Networks for Diagnosis and Pathogenetic Factors Identification of Alzheimer’s Disease With Imaging Genetic Data. IEEE Trans.
Neural Netw. Learn. Syst. 2022, 1–15. [CrossRef]
251. Xia, T.; Sanchez, P.; Qin, C.; Tsaftaris, S.A. Adversarial counterfactual augmentation: Application in Alzheimer’s disease
classification. Front. Radiol. 2022, 2, 1039160. [CrossRef]
252. Shi, R.; Sheng, C.; Jin, S.; Zhang, Q.; Zhang, S.; Zhang, L.; Ding, C.; Wang, L.; Wang, L.; Han, Y.; et al. Generative adversarial
network constrained multiple loss autoencoder: A deep learning-based individual atrophy detection for Alzheimer’s disease and
mild cognitive impairment. Hum. Brain Mapp. 2022, 44, 1129–1146. [CrossRef]
253. Noella, R.S.N.; Priyadarshini, J. Diagnosis of Alzheimer’s, Parkinson’s disease and frontotemporal dementia using a generative
adversarial deep convolutional neural network. Neural Comput. Appl. 2023, 35, 2845–2854. [CrossRef]
254. Parisot, S.; Ktena, S.I.; Ferrante, E.; Lee, M.; Guerrero, R.; Glocker, B.; Rueckert, D. Disease prediction using graph convolutional
networks: Application to autism spectrum disorder and Alzheimer’s disease. Med. Image Anal. 2018, 48, 117–130. [CrossRef]
255. Zhou, H.; He, L.; Zhang, Y.; Shen, L.; Chen, B. Interpretable graph convolutional network of multi-modality brain imaging for
Alzheimer’s disease diagnosis. In Proceedings of the 2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI),
Kolkata, India, 28–31 March 2022; pp. 1–5.
256. Zeng, L.; Li, H.; Xiao, T.; Shen, F.; Zhong, Z. Graph convolutional network with sample and feature weights for Alzheimer’s
disease diagnosis. Inf. Process. Manag. 2022, 59, 102952. [CrossRef]
257. Li, H.; Shi, X.; Zhu, X.; Wang, S.; Zhang, Z. FSNet: Dual Interpretable Graph Convolutional Network for Alzheimer’s Disease
Analysis. IEEE Trans. Emerg. Top. Comput. Intell. 2022, 7, 15–25. [CrossRef]
258. Zhu, Y.; Ma, J.; Yuan, C.; Zhu, X. Interpretable learning based dynamic graph convolutional networks for Alzheimer’s disease
analysis. Inf. Fusion 2022, 77, 53–61. [CrossRef]
259. Yan, B.; Li, Y.; Li, L.; Yang, X.; Li, T.-Q.; Yang, G.; Jiang, M. Quantifying the impact of Pyramid Squeeze Attention mechanism and
filtering approaches on Alzheimer’s disease classification. Comput. Biol. Med. 2022, 148, 105944. [CrossRef]
260. Ebrahimi-Ghahnavieh, A.; Luo, S.; Chiong, R. Transfer learning for Alzheimer’s disease detection on MRI images. In Proceedings
of the 2019 IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT), Bali,
Indonesia, 1–3 July 2019; pp. 133–138.
261. Khan, N.M.; Abraham, N.; Hon, M. Transfer learning with intelligent training data selection for prediction of Alzheimer’s disease.
IEEE Access 2019, 7, 72726–72735. [CrossRef]
262. Maqsood, M.; Nazir, F.; Khan, U.; Aadil, F.; Jamal, H.; Mehmood, I.; Song, O.-Y. Transfer learning assisted classification and
detection of Alzheimer’s disease stages using 3D MRI scans. Sensors 2019, 19, 2645. [CrossRef]
263. Agarwal, D.; Marques, G.; de la Torre-Díez, I.; Martin, M.A.F.; Zapirain, B.G.; Rodríguez, F.M. Transfer learning for Alzheimer’s
disease through neuroimaging biomarkers: A systematic review. Sensors 2021, 21, 7259. [CrossRef]
264. Martinez-Murcia, F.J.; Ortiz, A.; Gorriz, J.-M.; Ramirez, J.; Castillo-Barnes, D. Studying the manifold structure of Alzheimer’s
disease: A deep learning approach using convolutional autoencoders. IEEE J. Biomed. Health Inform. 2019, 24, 17–26. [CrossRef]
265. Mendoza-Léon, R.; Puentes, J.; Uriza, L.F.; Hoyos, M.H. Single-slice Alzheimer’s disease classification and disease regional
analysis with Supervised Switching Autoencoders. Comput. Biol. Med. 2020, 116, 103527. [CrossRef]
266. Ferri, R.; Babiloni, C.; Karami, V.; Triggiani, A.I.; Carducci, F.; Noce, G.; Lizio, R.; Pascarelli, M.T.; Soricelli, A.; Amenta, F.; et al.
Stacked autoencoders as new models for an accurate Alzheimer’s disease classification support using resting-state EEG and MRI
measurements. Clin. Neurophysiol. 2021, 132, 232–245. [CrossRef]
267. Hedayati, R.; Khedmati, M.; Taghipour-Gorjikolaie, M. Deep feature extraction method based on ensemble of convolutional auto
encoders: Application to Alzheimer’s disease diagnosis. Biomed. Signal Process. Control 2021, 66, 102397. [CrossRef]
268. Jellinger, K.A. Neuropathology of the Alzheimer’s continuum: An update. Free Neuropathol. 2020, 1, 32. [CrossRef]
269. Devi, G. A how-to guide for a precision medicine approach to the diagnosis and treatment of Alzheimer’s disease. Front. Aging
Neurosci. 2023, 15. [CrossRef]
270. Fiford, C.M. Disentangling the Relationship between White Matter Disease, Vascular Risk, Alzheimer’s Disease Pathology and
Brain Atrophy: Timing of Events and Location of Changes. Ph.D. Dissertation, UCL (University College London), London,
UK, 2019.
271. Nbo, S.N. Unsupervised Discovery of Mental Disorder Factors Using MRI. Ph.D. Dissertation, National University of Singapore,
Singapore, 2020.
272. Veitch, D.P.; Weiner, M.W.; Miller, M.; Aisen, P.S.; Ashford, M.A.; Beckett, L.A.; Green, R.C.; Harvey, D.; Jack, C.R., Jr.; Jagust, W.;
et al. The Alzheimer’s Disease Neuroimaging Initiative in the era of Alzheimer’s disease treatment: A review of ADNI studies
from 2021 to 2022. Alzheimer’s Dement. 2024, 20, 652–694. [CrossRef] [PubMed]
Mach. Learn. Knowl. Extr. 2024, 6 505

273. Xiao, J.; Li, J.; Wang, J.; Zhang, X.; Wang, C.; Peng, G.; Hu, H.; Liu, H.; Liu, J.; Shen, L.; et al. 2023 China Alzheimer’s disease:
Facts and figures. Hum. Brain 2023, 2. [CrossRef]
274. Tönges, L.; Buhmann, C.; Klebe, S.; Klucken, J.; Kwon, E.H.; Müller, T.; Pedrosa, D.J.; Schröter, N.; Riederer, P.; Lingor, P.
Blood-based biomarker in Parkinson’s disease: Potential for future applications in clinical research and practice. J. Neural Transm.
2022, 129, 1201–1217. [CrossRef] [PubMed]
275. Blasiak, J.; Sobczuk, P.; Pawlowska, E.; Kaarniranta, K. Interplay between aging and other factors of the pathogenesis of age-related
macular degeneration. Ageing Res. Rev. 2022, 81, 101735. [CrossRef] [PubMed]
276. Paolini Paoletti, F.; Simoni, S.; Parnetti, L.; Gaetani, L. The contribution of small vessel disease to neurodegeneration: Focus on
Alzheimer’s disease, Parkinson’s disease and multiple sclerosis. Int. J. Mol. Sci. 2021, 22, 4958. [CrossRef] [PubMed]
277. Erkkinen, M.G.; Kim, M.O.; Geschwind, M.D. Clinical neurology and epidemiology of the major neurodegenerative diseases.
Cold Spring Harb. Perspect. Biol. 2018, 10, a033118. [CrossRef] [PubMed]
278. van Oostveen, W.M.; de Lange, E.C. Imaging techniques in Alzheimer’s disease: A review of applications in early diagnosis and
longitudinal monitoring. Int. J. Mol. Sci. 2021, 22, 2110. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.