Received November 27, 2020, accepted December 26, 2020, date of publication December 30, 2020,

date of current version January 20, 2021.

Digital Object Identifier 10.1109/ACCESS.2020.3048172

A Review on Traditional Machine Learning and

Deep Learning Models for WBCs Classification
in Blood Smear Images
SIRAJ KHAN1 , MUHAMMAD SAJJAD 1,2 , TANVEER HUSSAIN 3 , (Student Member, IEEE),
AMIN ULLAH 3 , (Member, IEEE), AND ALI SHARIQ IMRAN 2 , (Member, IEEE)
1 Digital Image Processing Laboratory, Department of Computer Science, Islamia College University Peshawar, Peshawar 25000, Pakistan
2 Department of Computer Science, Norwegian University of Science and Technology (NTNU), Gjøvik, Norway
3 Intelligent Media Laboratory, Department of Software, Sejong University, Seoul 143-747, South Korea
Corresponding authors: Ali Shariq Imran ([email protected]) and Muhammad Sajjad ([email protected])
This work was supported in part by ERCIM ‘Alain Benoussan’ Fellowship Programme, and in part by the ColorLab at the Department of
Computer Science, NTNU, Gjøvik. The work of Muhammad Sajjad was supported under Contract 2019–40.

ABSTRACT In computer vision, traditional machine learning (TML) and deep learning (DL) methods have
significantly contributed to the advancements of medical image analysis (MIA) by enhancing prediction
accuracy, leading to appropriate planning and diagnosis. These methods substantially improved the diagnoses
of automatic brain tumor and leukemia/blood cancer detection and can assist the hematologist and doctors by
providing a second opinion. This review provides an in-depth analysis of available TML and DL techniques
for MIA with a significant focus on leukocytes classification in blood smear images and other medical
imaging domains, i.e., magnetic resonance imaging (MRI), CT images, X-ray, and ultrasounds. The proposed
review’s main impact is to find the most suitable TML and DL techniques in MIA, especially for leukocyte
classification in blood smear images. The advanced DL techniques, particularly the evolving convolutional
neural networks-based models in the MIA domain, are deeply investigated in this review article. The related
literature study reveals that mainstream TML methods are vastly applied to microscopic blood smear images
for white blood cells (WBC) analysis. They provide valuable information to medical specialists and help
diagnose various hematic diseases such as AIDS and blood cancer (Leukaemia). Based on WBC related
literature study and its extensive analysis presented in this study, we derive future research directions for
scientists and practitioners working in the MIA domain.

INDEX TERMS Blood smear images, CNN, deep learning, medical image analysis, traditional machine
learning, WBCs classification.
I. INTRODUCTION hard and soft tissues of different body organs for diag-
Traditional machine learning (TML) and Deep learning (DL) nostic and research purposes [3]. MIA has dense contribu-
techniques are widely used for various applications and are tions for computer vision experts in the investigated topic,
extensively applied in the medical image analysis (MIA) where TML and DL play a significant role in leukocyte
domain [1]. In modern healthcare systems, MIA is an essen- segmentation, cancer detection, classification, medical image
tial attribute, assisting medical experts wisely. MIA plays annotation, and image retrieval in computer-aided diagnosis
a vital role in diagnosing several diseases such as brain (CAD). The CAD and computer aided-detection (CADx)
tumors, lung cancer, anemia, leukemia, and malaria. MIA rely on effective TML and DL schemes because their per-
processes various image modalities such as MRI, CT-Scan, formance directly affects clinical diagnosis and treatment
Ultrasounds, Positron Emission Tomography (PET), Blood process [4], [5]. It further assists the doctors in the diag-
Smear images, and hybrid modalities [2]. In MIA, the image nostic and treatment process, easing their traditional work-
modalities play a vital role in detecting and classifying ing mechanisms. The recent developments in information
technology, such as high-speed computational resources,
The associate editor coordinating the review of this manuscript and hardware design, and storage capabilities significantly
approving it for publication was Wai-keung Fung . impact CAD.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 9, 2021 10657
 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

FIGURE 1. Leukocytes types [6]; (a) Lymphocyte, (b) Monocyte, (c) Neutrophil, (d) Eosinophil and (e) Basophil.

FIGURE 2. General overview of TML and DL models for leukocytes classification in blood smear images.

Formerly, key application areas of CAD system via TML techniques rely on manual analysis of WBCs in blood smear
and DL involve early-stage brain tumor detection in MR images, a time-consuming, challenging task, and prone to
images and leukocytes analysis. It provides valuable infor- errors [6]–[9]. Automatic and CAD systems have a crucial
mation to medical experts, helping them diagnose differ- role in clinical diagnosis and appropriate treatment [10]–[13].
ent hematic problems such as AIDS and blood cancer Therefore, automatic analysis of WBCs in microscopic
(Leukaemia). The main aim of MIA is to assist medical blood smear images is gaining popularity because it can
experts, doctors, hematologists, pathologists, radiologists in decrease the workload on hematologists and provide quick,
the diagnostic and treatment process more effectively and efficient, and accurate results to assist medical experts in the
efficiently. In the medical field, it has been perceived that diagnostic process [14]. There are mainly two ways to achieve
the mainstream human body’s diseases are recognized by automated WBCs classification in blood smear images, i.e.,
analyzing leukocytes/WBCs [9]. The increase or decrease TML and DL techniques, which have great potentials to
of leukocytes/WBCs and their morphological structure, such develop such automatic systems that can make medical hema-
as size, shape, and color variations in blood smear images, tology more efficient [14]–[16]. The General overview of
indicate different human body disorders. TML and DL Models for WBCs classification in blood smear
There are different types of blood cells, such as WBCs images is shown in Fig. 2. Different CAD systems can auto-
(leukocytes), RBCs (erythrocytes), and platelets (thrombo- matically diagnose numerous hematic types, such as AIDS
cytes). Leucocytes are further divided into five subcategories: and blood cancer (Leukemia), by analyzing leucocytes [15].
monocyte, lymphocyte, neutrophil, basophil, and eosinophil, In TML, there are interconnected steps involved, such as
as shown in Fig. 1. Various TML and DL techniques have segmenting ROI and extracting features followed by optimal
emerged in the last two decades to segment and classify classification. A variety of TML techniques are available,
WBCs in microscopic blood smear images. Conventional i.e., manual, semi-automatic, and automatic segmentation

10658 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

techniques to segment ROI from an image [16]. Features both manual and automatic searches to get the most rele-
extraction is another step in the TML approach. However, vant research articles by following the Preferred Reporting
selecting an optimal feature extractor is challenging due Items for Systematic Reviews and Meta-Analyses (PRISMA)
to varying feature dynamics, such as geometric invariance model [20], [21].
and photometric invariance. Nowadays, the vast emergence We performed both manual and automatic searches to
of DL approaches has resulted in high-performance MIA fetch the most relevant content. Our searching strategy begins
models, especially in clinical hematology using blood smear with an automatic search on electronic databases to retrieve
images [18]–[33]. relevant data followed by verification of results by MIA
This research provides a comprehensive survey of the and leucocytes classification experts. In the proposed survey,
available TML and DL techniques and their medical imaging we search for articles from the period of 2014 to 2020. All
applications, mainly targeting WBCs classification in blood the included sources are searched automatically as well as
microscopic images. There have been several surveys on MIA manually using the predefined keywords. Keywords for the
using TML and DL techniques and future trends focusing on search are decided by the authors and other research partici-
MRI, CT, X-rays, but microscopic blood smear is a rarely pants (a group of four research students working in the area of
addressed problem [17], [18]. Therefore, this study intends MIA). These keywords include ‘‘Medical Image Analysis,’’
to fill this gap by analyzing state-of-the-art TML and DL AND ‘‘Leucocytes Classification,’’ OR ‘‘WBC’s Detection in
techniques for MIA, particularly leucocytes classification Blood Smear Images’’. Search keywords are defined based on
methods in blood smear images. The proposed research’s the following steps:
primary focus is to provide a comprehensive review of the (a) The major terms are derived from research questions
use of TML and DL in MIA. (b) Alternating synonyms or spellings are identified for the
In the proposed study, a novel categorization is employed major terms.
to find the most common TML and DL methods that are (c) Keywords are identified from relevant research articles
reviewed in separate groups according to the research focus and books.
and employed technique. This research also helps identify (d) Boolean operator OR is used for alternating spellings
future research directions by following TML and DL tech- or synonyms.
niques to classify leucocytes in microscopic blood smear (e) The major terms are linked by Boolean AND operator
images. The followings are some of the significant contribu- and the search string is formed after the analysis of the
tions of the proposed review study: keywords in order to retrieve the relevant information
• The outlines of this paper investigate different applica-
from the databases.
tions and uses of TML and DL models in MIA.
• This This research study also aims to identify available The above-mentioned keywords and string are checked on
machine learning techniques for leukocyte classification each database and its pattern is modified based on relevant
and analyze the extent of accuracy, applications, and results retrieval. Numerous keywords associated with the
MIA contributions. study’s primary focus is based on the four research ques-
• We address the key challenges and requirements of TML tions (RQ) that are designed keeping in view the Patient,
and DL models, followed by its future directions and Intervention, Comparison, Outcome (PICO) framework [22].
solutions for future research in MIA. RQ -1: What are the different TML and DL techniques for
The remaining paper is structured as follows; Section II leukocyte classification in blood smear images?
describes the review methodology and papers scrutiniza- RQ -2: What are the different applications of TML and
tion process in detail. Section III gives a brief introduction DL techniques in medical analysis, especially leukocytes
about MIA. In Section IV, we present the detailed summary classification?
and applications of the artificial neural network and leu- RQ -3: How are TML and DL techniques used in MIA, par-
cocytes classification in microscopic blood smear images. ticularly for leukocyte classification in blood smear images?
In Section V, the current challenges and requirements are RQ -4: What type of machine learning is practical and
discussed. Future directions of the proposed review study are efficient for analyzing leukocytes in blood smear images?
described in section VI. In the last section VII, we discuss
about recent advancements in DL models, followed by con- A. RETRIEVED PAPERS SCRUTINIZATION CRITERION
clusions of the proposed review work. The initially retrieved papers are subject to inclu-
sion/exclusion criteria by following the PRISMA guidelines.
II. REVIEW METHODOLOGY Table 1 represents the inclusion and exclusion criteria to
This section provides a detailed discussion about digital filter out irrelevant articles. The selection of research articles
libraries used for conducting a formal research process in is completed in three steps. Firstly, duplicate papers are
the proposed review study. A planned searching procedure removed. Secondly, the paper title, abstract, and keywords are
is required to find the available literature that fulfills the investigated for relevancy, and finally, the remaining research
searching criteria, to utilize the available digital resources papers are included after a thorough investigation. The pro-
purposefully [19]. In the proposed study, we incorporated cess of exclusion and inclusion criteria is applied to eliminate

VOLUME 9, 2021 10659

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 1. Criteria for research papers inclusion and exclusion in the proposed study.

FIGURE 3. A comprehensive illustration for article selection process.

FIGURE 4. Number of research publications in the field of medical

imaging in the past two decades.
conflict analysis and biasedness. A total of 1436 research
papers are collected to review the literature based on the
research focus during the article selection process. In the
initial selection process, manual filtering is incorporated, and
the papers are filtered using the relevant title, and 1106 papers
are obtained. These 1106 articles are then filtered by observ-
ing the abstract and conclusion, finally leaving 922 papers.
These papers are filtered by methodology and results in the
next step, and 725 articles are obtained. Then the articles are
filtered after reading the full contents, leaving 216 articles.
We checked the remaining articles’ quality by evaluating the
methodology, full-proof results, journal’s impact factor, and
citations. After checking all these parameters, 80 papers are
picked for the proposed study. After the completion of the
paper’s scrutinization process (paper inclusion and exclu-
sion), the quality assessment is performed. Each research FIGURE 5. The exponential growth in the research of TML and DL for
leucocytes classification in blood smear images.
article is assessed against the scrutinization criteria. All
research articles are reviewed, and the quality of the papers
with respect to each research question is assessed. Each of the X-ray, MRI, microscopic blood smear images, PET, and
selected articles is read and analyzed manually by the authors. ultrasound [17], [23], [24]. These imaging technologies play
The publication channels used for the article searching and an essential role in MIA; doctors and medical experts can
the stepwise selection process are presented in Fig. 3. automatically detect and diagnose different chronic diseases
by analyzing these images. They can also visualize different
III. MIA body organs for research [37]. The number of research papers
The process that can provide visual information of the explored in this field is shown in Fig. 4. The last two decades
human body to assist the radiologists and doctors in an have witnessed extensive medical imaging usage in CAD,
efficient diagnostic and treatment is called medical imag- for instance, in applications such as for leucocytes segmenta-
ing [23]. There are many image modalities upon which tion and classification, tumor segmentation and classification,
the doctors and medical experts rely for diagnosing dis- detection and classification of breast cancer, image-guided
eases and prescribing treatment. These modalities include CT, therapy, and medical image annotation [25]–[28]. It has,

10660 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

FIGURE 6. General architecture of TML based leucocytes segmentation and classification.

therefore, became an integral part of today’s modern health- algorithm [31], which are the most crucial steps towards
care systems [29]. efficiency and accuracy of the adopted technique. The general
overview of TML is shown in Fig. 6.
Many conventional supervised learning methods have been
A. TML AND DL FOR LEUCOCYTES CLASSIFICATION IN used to classify leucocytes in microscopic blood smear
BLOOD SMEAR IMAGES images, such as Support Vector Machine (SVM) [32]–[34],
The literature includes a sufficient number of recently pub- Naive Bayes (NB) [35]–[37], K-Nearest Neighbor (KNN)
lished review articles on TML and DL techniques used in [38]–[40], and Artificial Neural Network (ANN) [41]–[43].
MIA. The most recent and relevant research works about Some popular WBCs nuclei detection techniques are identi-
TML and DL methods in medical imaging, particularly for fied and reviewed, which are presented in Table 2.
the classification of leucocytes in blood smear images [30],
are discussed in the subsequent sections. In the proposed
B. LEUKOCYTES CLASSIFICATION USING SVM
study, the most relevant and recent studies are searched out
There are numerous supervised learning techniques available
using keywords ‘‘leucocytes detection’’ or ‘‘leucocytes clas-
to deal with leucocyte classification, such as SVM, ANN,
sification’’ by filtering the recent papers. During searching,
Naïve Bayesian, and Decision Trees. Hegde et al. [70] pro-
we found that there is an exponential research growth of using
posed a novel technique in which the authors first segmented
TML and DL methods for leukocytes analysis in blood smear
the WBCs and then employed SVM to classify WBC cells
images. Fig. 5. represents the overall research results of DL
into a normal or leukemic cell. Zhao et al. [69] proposed
and TML techniques for MIA and its exponential growth in
a novel technique to segment and classify Leukocytes in
the last two decades.
blood smear images. Color co-relation and morphological
TML approaches involve interconnected steps, i.e., image
based segmentation are applied, followed by texture features
pre-processing, segmentation, feature extraction, feature
extraction and classification using SVM to classify WBCs
selection, and classification. The pre-processing step includes
into its five subclasses [90]. Table 3 elaborates on the key
image enhancement such as contrast adjustment, noise
contributions and applications of SVM for leucocytes classi-
removal, and image sharpening. All these steps are applied
fication in blood smear images.
to the input image before image segmentation [41]. There are
numerous pre-processing techniques such as median filter,
low pass filter, high pass filter, and Gabor filter. These are C. ENSEMBLES, HYBRIDS, BAYESIAN, K-NN AND
used normally for image contrast adjustment, image sharpen- DECISION TREES FOR LEUKOCYTES CLASSIFICATION
ing, and noise removal before image segmentation. TML has In addition to ANNs and SVMs, which have significant
been addressed by several researchers for leucocytes detec- contributions to MIA, hybrids, Bayesian, Ensembles, K-
tion and classification. However, accurate nuclei detection, NN, and Tree models have also been applied to solve the
separation of borders to recover overlapped cells, segment- problems in different sub-domains of medical imaging such
ing ROI, robust features extraction, and best features selec- as brain tumor detection, lung cancer detection, leukocytes
tion have become challenging and time-consuming using classification, etc. Abdulkadir et al. [71] proposed a hybrid
these approaches [31]–[33]. In this approach, after segment- approach for WBC classification in blood smear images.
ing ROI, the next step is feature extraction. In traditional Sajjad et al. [15] proposed a smartphone-based quality health-
supervised learning techniques, the classification depends on care system for smart cities, in which an ensemble multi-class
choosing robust features descriptor and best features selection SVM is used to classify WBCs in blood smear images.

VOLUME 9, 2021 10661

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 2. List of various research studies for WBCS nuclei detection in blood smear images.

10662 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 2. (Continued.) List of various research studies for WBCS nuclei detection in blood smear images.

TABLE 3. Notable contributions and application of SVM for leucocytes classification.

Tantikitti et al. [72] proposed a computer-aided diagnos- the SVM classifier. A transfer learning is also utilized for
ing system to diagnose dengue fever disease. A multi-level further classification of leukocytes, as comprehensively given
threshold technique is used to segment leukocytes in blood in Table 4.
smear images. This research has two decision tree models
for classification. The first model was used to classify the IV. ANN FOR LEUCOCYTES CLASSIFICATION
type of white blood cells that are lymphocytes or Phago- ANN is a supervised learning technique inspired by the
cytes. The second model is used to classify the dengue biological nervous system of the human brain. It involves
virus infection as positive or negative. In [73], a novel tech- input, output, and hidden layers that are linked together via
nique is proposed in which WBCs nucleus and cytoplasm weighted connections. The performance of any ANN tech-
are segmented using simple thresholding. After segmenta- nique depends on these weights, which are numerical values.
tion, some morphological operations are performed using The output layer generates results given the inputs based
ellipse curve fitting, followed by feature extraction. For fea- on weights, error function, and neurons in the hidden layer.
ture selection, the sequential forward selection technique is Several research studies have applied ANN in the context of
incorporated, and finally, a naïve Bayes classifier is used to MIA due to its enormous applications, including leucocytes
classify WBCs. Vogado et al. [74] used a hybrid approach classification, brain tumor classification, breast cancer detec-
for the classification and segmentation of leukocytes. In their tion, and lung cancer detection. Some notable contributions
proposed technique, CNN features are used as input to train and applications are summarized in Table 5.

VOLUME 9, 2021 10663

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 4. Notable contributions and applications of bayesian, ensembles, hybrids, K-NN and trees models for leucocytes classification in blood smear
images.

A. LEUKOCYTES CLASSIFICATION BASED

ON DEEP LEARNING
DL allows us to define a system in which the feature extrac-
tion is not designed by human engineers but learned from
data using a general-purpose learning procedure [79]. In the
field of MIA, deep learning achieved satisfactory perfor-
mance and relatively easy to build an end-to-end network
using CNN [80]. TML models are trained on manually
extracted features, or they learn features via other simple FIGURE 7. The general overview of deep learning.
machine learning techniques to perform different classifi-
cation tasks. Therefore, DL techniques have attracted the data. These methods are particularly suitable to those areas
researcher’s attention and motivated them to explore DL’s where human-like intelligence is required to analyze large
benefits for WBCs classification. Currently, DL has become amounts of data. Additionally, good knowledge is needed to
a powerful research tool in artificial intelligence, speech extract rich features from a massive raw data volume [88].
analysis [81], natural language processing (NLP) [82], and However, this task is challenging and time-consuming
medical imaging [83]. DL’s use is also becoming an essen- when a vast collection of data is to be handled efficiently.
tial aspect as a pattern recognition tool in the field of DL provides end-to-end learning and eliminates all extra
MIA [84]–[86]. According to a recent review on DL based overheads of selecting feature descriptors and feature selec-
MIA [87], DL algorithms and particularly convolutional net- tion, as shown in Fig.7. DL methods’ significant advantage
works, have become a choice for many for analyzing medical is learning and automatically extracting semantically rich

10664 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 5. Representative key contributions and applications of ANN for WBCS classification.

features from the raw data [82]. This is the main difference early detection of WBCs in blood smear images. In [14],
between TML and DL models. DL’s unmatched benefits have the authors proposed a Regional-Based CNN using transfer
attracted a large research community and industries to use learning approaches to classify WBCs in peripheral blood
DL-based approaches for MIA. smear images. The overview of some recent articles using DL
DL models can be classified into different categories for leukocyte classification is shown in Table 6.
such as convolutional neural networks [95], deep belief
networks [96], Long short-term memory networks [97],
Recurrent Neural Networks (RNN) [98], and deep auto- V. CURRENT CHALLENGES AND REQUIREMENTS
encoders [99]. Convolutional neural networks (CNNs) is In this extensive literature review, we found the major
widely used in medical imaging [17]. research challenges and requirements, several key features,
their applications, and advantages of TML and DL techniques
for MIA, particularly for WBC’s classification in blood smear
B. LEUKOCYTES CLASSIFICATION USING images. In the last few years, there are certain standard and
CONVOLUTIONAL NEURAL NETWORKS (CNN) powerful TML and DL models developed for MIA, such as
CNN consists of multiple convolutional, pooling, and fully brain tumor localization and classification from MRI, leuko-
interconnected layers with activation functions. It is trained cytes detection and classification in blood smear images,
using gradient descent and backpropagation as any standard and lung cancer detection in CT images [124]. Still, there
ANNs (see Fig. 8) [100]. Typical CNNs generally have a exist some significant challenges that the research community
successive convolutional and pooling layer followed by a either has to accept or try to overcome. These challenges
fully connected layer. A Softmax function is used at the include the unavailability of publicly available large and good
output nodes to classify WBC’s into its five respective cat- quality datasets, dedicated medical experts, and lightweight
egories, i.e., monocyte, lymphocyte, neutrophil, basophil, TML and DL techniques. Some of the challenges are related
and eosinophil. Banik al. [101] proposed a novel CNN to the mathematical and theoretical underpinnings of many
model for WBCs classification by fusing the features of DML techniques [123], [124]. To overcome these challenges,
first and last convolutional layers using the BCCD database. unsupervised or semi-supervised systems are required [83].
Choi et al. [102] proposed a CNN model with eight layers The proficiency of semi-supervised and unsupervised meth-
for WBCs classification. Karthikeyan et al. [103] presented ods in MIA will be compromised to avoid these issues.
an LSM-TIDC method to classify WBCs in blood smear It is also challenging to move from supervised learning to
images. Firstly, images are pre-processed, then texture and unsupervised learning approaches without affecting the sys-
geometrical features are extracted using a multi-directional tem’s accuracy and efficiency. MIA applications and systems
model. Finally, the extracted features are fed as a fea- employing TML and DL methods are still far from perfect,
ture vector to deep convolutional networks for efficient and leaving significant space for improvements.

VOLUME 9, 2021 10665

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

FIGURE 8. Building blocks of general CNN architecture for leukocytes classification. A slight modification of a figure in [17].

A. UNAVAILABILITY OF PUBLICALLY AVAILABLE DATASETS solutions that minimize the training error. More surprisingly,
The major problem in the field of medical image analysis is the networks learned this way exhibit good generalization
the unavailability of publicly available datasets. To address abilities [128], even when the number of parameters is sig-
this issue, the researchers need to encourage health orga- nificantly larger than the amount of training data [129].
nizations to make their medical data available; it can be During model training, only minimizing the training error
interesting if quality data is publicly available for researchers. is not enough. Picking the wrong global minima can also
Moreover, initiatives that encourage open data from different lead to bad generalization behavior for the predictor. In
health institutions worldwide are encouraged; some operation such situations, generalization behavior depends implicitly
are also necessary (e.g., data from hospitals and conditional on the algorithm used to minimize the training error. Different
access to datasets). In all these cases, incentive mechanisms algorithmic choices for optimization, such as the initializa-
can be related to financial return, entertainment, or services tion, update rules, learning rate, and stopping condition, will
to these institutions while providing quality data. The topic lead to different global minima with different generalization
becomes more interesting for research when the data is avail- abilities.
able in massive amounts, just like other fields (e.g., video
summarization [125], IoT [126], energy management [127], C. TRUST-WORTHY METHODS TO BE FUNCTIONAL IN
and so on.). It is vital to collect extensive and quality datasets REAL-WORLD ENVIROMENTS
with ground-truth labels for specific MIA applications. More- The existing TML and DL techniques are not good enough to
over, such datasets can be used for benchmarking as well as be trusted without medical expertise to function in real world
hosting different competitions. health diagnosis systems [130]. There must be an expert as
well as technical skills to train a learning model for MIA and
leucocytes classification. We need to explore such precise and
B. TRAINED PREDICTOR GENERALIZATION ABILITIES
trustworthy methods which do not need health experts and are
The key issue with MIA and leucocytes detection and clas-
implementable in real-world health applications.
sification is to train a predictor. An ideal learning technique
with a better balance of generalization ability and a computa-
tionally efficient heuristic model is required to overcome this VI. FUTURE RESEARCH DIRECTIONS
problem. A learning paradigm that uses true or random labels Considering the major challenges encountered by the MIA
and provides effective tools to deal with available datasets community outlined in section V, extensive work is demanded
and efficient training algorithms are needed to train a model from the biomedical industry and research community to
with remarkable generalization abilities. Learning with deep contribute to MAI and especially leukocytes analysis in blood
neural networks has enjoyed huge empirical success in recent smear images.
years across a wide variety of tasks in the field of MIA,
i.e., brain tumor detection, lung cancer, breast cancer detec- A. DATA AUGMENTATION TECHNIQUES TO FILL THE
tion, and leucocytes classification. Despite being a complex, DATASETS DEFICIENCY
non-convex optimization problem, simple methods such as In this study, we have focused on the most frequently men-
stochastic gradient descent (SGD) can recover reasonable tioned problem of unavailability of datasets in the field

10666 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 6. Some notable key contribution of different deep learning models for leukocytes classification.

VOLUME 9, 2021 10667

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TABLE 6. (Continued.) Some notable key contribution of different deep learning models for leukocytes classification.

of MIA and leucocytes classification. An extensive data D. END-TO-END LEUCOCYTES DETECTION AND
augmentation technique and transfer leaning models are CLASSIFICATION MODELS
recommended to improve MIA and WBC’s detection Traditional learning techniques can be replaced by a deep
classification in blood smear images. There are several neural network (DNN) based models. With the recent
data augmentation techniques used to extend the existing advancement of CNNs [132], end-to-end models are also
data, i.e., classical image transformations like rotating, crop- gaining in popularity due to simplified model-building pro-
ping, zooming, Gaussian blur, sharpening, edge detection, cesses and the ability to classify leucocytes into its five cate-
histogram-based methods, and finishing at Style Transfer and gories. These models are based on data-driven learning meth-
Generative Adversarial Networks. ods and competition with complicated MIA models based
on DNN. Different end-to-end architectures for leucocyte
B. MEDICAL EXPERTISE AND TECHNICAL detection and classification in blood smear images, such as
SKILL ARE REQUIRED attention-based methods [133], [134] and CNN based model
In the future, computer-aided MIA-based diagnostic appli- are also prominent.
cations can benefit from the recent advances in TML and
DL models. These models are already available on multi-
E. UNIVERSAL EVALUATION FOR TML AND DL IN MIA
ple open-source platforms such as Tensorflow, Caffe, and
In MIA, the research community mainly relies on subjec-
Keras [131]. However, selecting and training an appropri-
tive evaluation techniques. However, this task is challenging,
ate machine learning model for a specific MAI problem is
time-consuming, and can be prone to errors. Thus, further
challenging due to limited medical expertise and clinical
research is required to explore universal evaluation tech-
knowledge.
niques that can automatically measure the performance of
C. RESOURCE CONSCIOUS DL MODELS FOR LEUKOCYTES TML and DL models for MIA from different perspectives.
CLASSIFICATION
In recent developments, DL, i.e., GAN’s (Generative Adver- VII. DISCUSSION AND CONCLUSION
sarial Networks), R-CNN, Fast R-CNN, faster R-CNN, and This study provided a comprehensive review of TML and DL
deep fusion of TML and DL techniques models have achieved techniques used for leukocyte classification in blood smear
higher performance in brain tumor detection, leukocytes clas- mages. We reviewed different TML and DL approaches to
sification, breast cancer detection, and other MIA tasks. How- classify WBCs in blood smear images. The data are collected
ever, their primary concerns are high computational cost and from primary studies published during 2014 to 2020. The
high memory requirements. So, computationally efficient and current study’s literature identifies 80 primary studies (arti-
energy-friendly TML and DL models need to be explored cles published in journals, books, conferences, and online
for leukocytes analysis in blood smear images. Furthermore, materials) defining TML and DL techniques for leucocytes
such light weighted models can be easily implemented over classification in blood smear images and its applications in
resource-constrained devices. medical diagnosis. While reviewing the articles, we found

10668 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

that both TML and DL approaches have performed equally [6] M. Sajjad, S. Khan, M. Shoaib, H. Ali, Z. Jan, K. Muhammad, and
well with overall contributions in MIA. This study is focused I. Mehmood, ‘‘Computer aided system for leukocytes classification and
segmentation in blood smear images,’’ in Proc. Int. Conf. Frontiers Inf.
on identifying different applications of TML and DL in MIA Technol. (FIT), Dec. 2016, pp. 99–104.
and leucocytes classification in blood smear images. The [7] N. S. M. Raja, S. L. Fernandes, N. Dey, S. C. Satapathy, and
objective of this study is to gain insight into complex details V. Rajinikanth, ‘‘Contrast enhanced medical MRI evaluation using tsallis
entropy and region growing segmentation,’’ J. Ambient Intell. Humanized
of TML and DL by accumulating and analyzing the knowl- Comput., vol. 2018, pp. 1–12, May 2018.
edge provided in the literature in order to facilitate further [8] Y. Duan, J. Wang, M. Hu, M. Zhou, Q. Li, L. Sun, S. Qiu, and
research in the field of MIA. This study shows that much Y. Wang, ‘‘Leukocyte classification based on spatial and spectral features
of microscopic hyperspectral images,’’ Opt. Laser Technol., vol. 112,
work is still needed to investigate the use of TML and DL pp. 530–538, Apr. 2019.
techniques for useful MIA and leucocytes classification in [9] H. Cao, H. Liu, and E. Song, ‘‘A novel algorithm for segmentation of
blood smear images. This study also aimed at identifying leukocytes in peripheral blood,’’ Biomed. Signal Process. Control, vol. 45,
pp. 10–21, Aug. 2018.
applications of advanced DL models other than leucocyte
[10] E. Abdulhay, M. A. Mohammed, D. A. Ibrahim, N. Arunkumar, and
classification. However, it is found that almost all other med- V. Venkatraman, ‘‘Computer aided solution for automatic segmenting
ical diagnosis applications are either directly or indirectly and measurements of blood leucocytes using static microscope images,’’
related to TML and DL. The accumulation of all this infor- J. Med. Syst., vol. 42, no. 4, p. 58, Apr. 2018.
[11] L. Wang, S. Yang, S. Yang, C. Zhao, G. Tian, Y. Gao, Y. Chen, and
mation in this study will benefit the research community by Y. Lu, ‘‘Automatic thyroid nodule recognition and diagnosis in ultrasound
identifying where they need to start in further research on imaging with the YOLOv2 neural network,’’ World J. Surg. Oncol.,
TML and DL models for MIA. vol. 17, no. 1, pp. 1–9, Dec. 2019.
[12] R. B. Hegde, K. Prasad, H. Hebbar, and B. M. K. Singh, ‘‘Comparison
In future these techniques will have tremendous con- of traditional image processing and deep learning approaches for classifi-
tributions in the development of medical imaging, natural cation of white blood cells in peripheral blood smear images,’’ Biocyber.
language processing and speech analysis. Beside WBCs, Biomed. Eng., vol. 39, no. 2, pp. 382–392, Apr. 2019.
[13] M. Sajjad, S. Khan, K. Muhammad, W. Wu, A. Ullah, and S. W.
TML and DL techniques are also used for the detection Baik, ‘‘Multi-grade brain tumor classification using deep CNN with
and classification of different MIA domains i.e., MRI, CT, extensive data augmentation,’’ J. Comput. Sci., vol. 30, pp. 174–182,
X-ray, Ultrasound images analysis. In the current study, Jan. 2019.
[14] H. Kutlu, E. Avci, and F. Özyurt, ‘‘White blood cells detection and
we reviewed different TML and DL techniques such as SVM
classification based on regional convolutional neural networks,’’ Med.
ANNs, Ensembles, Bayesians, neuro-fuzzy, hybrids, DL and Hypotheses, vol. 135, Feb. 2020, Art. no. 109472.
CNNs which are used to analyzed blood smear image [15], [15] M. Sajjad, S. Khan, Z. Jan, K. Muhammad, H. Moon, J. T. Kwak,
[72]–[78]. In MIA, blood smear images are the emerging S. Rho, S. W. Baik, and I. Mehmood, ‘‘Leukocytes classification and
segmentation in microscopic blood smear: A resource-aware healthcare
domain that achieved great attention by the research com- service in smart cities,’’ IEEE Access, vol. 5, pp. 3475–3489, 2017.
munity since last three decades. Standard contributions and [16] F. Xing and L. Yang, ‘‘Robust nucleus/cell detection and segmentation
applications of TML and DL in MIA are presented in this in digital pathology and microscopy images: A comprehensive review,’’
IEEE Rev. Biomed. Eng., vol. 9, pp. 234–263, Jan. 2016.
study. Furthermore, we also identified the current challenges, [17] A. S. Lundervold and A. Lundervold, ‘‘An overview of deep learning in
future directions and solutions for the advancements of TML medical imaging focusing on MRI,’’ Zeitschrift für Med. Phys., vol. 29,
and DL models in the field of MIA and particularly for no. 2, pp. 102–127, May 2019.
[18] T. Zhou, S. Ruan, and S. Canu, ‘‘A review: Deep learning for medi-
WBCs classification in blood smear images. In future, we aim cal image segmentation using multi-modality fusion,’’ Array, vols. 3–4,
to extend our survey by considering various MIA domains Sep. 2019, Art. no. 100004.
such as MRI, CT, Ultrasound, X-ray images by utilizing the [19] T. Hussain, K. Muhammad, W. Ding, J. Lloret, and S. W. Baik, ‘‘A com-
potentials of TML and DL techniques. prehensive survey of multi-view video summarization,’’ Pattern Recog-
nit., vol. 109, Jan. 2020, Art. no. 107567.
[20] I. Balki, A. Amirabadi, J. Levman, A. L. Martel, Z. Emersic, B. Meden,
REFERENCES A. Garcia-Pedrero, S. C. Ramirez, D. Kong, A. R. Moody, and
P. N. Tyrrell, ‘‘Sample-size determination methodologies for machine
[1] P. Pandey, S. Pallavi, and S. C. Pandey, ‘‘Pragmatic medical image learning in medical imaging research: A systematic review,’’ Can. Assoc.
analysis and deep learning: An emerging trend,’’ in Advancement Radiol. J., vol. 70, no. 4, pp. 344–353, Nov. 2019.
of Machine Intelligence in Interactive Medical Image Analysis.
[21] P. S. Fleming, D. Koletsi, and N. Pandis, ‘‘Blinded by PRISMA: Are sys-
Singapore: Springer, Jan. 2020, pp. 1–18. [Online]. Available:
tematic reviewers focusing on PRISMA and ignoring other guidelines?’’
http://www.springer.com/series/16171, doi: 10.1007/978-981-15-
PLoS ONE, vol. 9, no. 5, May 2014, Art. no. e96407.
1100-4_1.
[22] A. Littlewood and D. Kloukos, ‘‘Searching the literature for studies
[2] S. Asgari Taghanaki, K. Abhishek, J. P. Cohen, J. Cohen-Adad, and for a systematic review. Part 1: Identifying search concepts in a ques-
G. Hamarneh, ‘‘Deep semantic segmentation of natural and medical tion,’’ Amer. J. Orthodontics Dentofacial Orthopedics, vol. 155, no. 2,
images: A review,’’ Artif. Intell. Rev., vol. 4, pp. 1–42, Jun. 2020. pp. 299–301, Feb. 2019.
[3] M. Rahman, S. Antani, and G. Thoma, ‘‘A learning-based similar- [23] S. M. Anwar, M. Majid, A. Qayyum, M. Awais, M. Alnowami,
ity fusion and filtering approach for biomedical image retrieval using and M. K. Khan, ‘‘Medical image analysis using convolutional neu-
SVM classification and relevance feedback,’’ IEEE Trans. Inf. Technol. ral networks: A review,’’ J. Med. Syst., vol. 42, no. 11, p. 226,
Biomed., vol. 15, no. 4, pp. 640–646, Jul. 2011. Nov. 2018.
[4] H. P. Chan, L. M. Hadjiiski, and R. K. Samala, ‘‘Computer-aided diagno- [24] P. Lakhani, D. L. Gray, C. R. Pett, P. Nagy, and G. Shih, ‘‘Hello world deep
sis in the era of deep learning,’’ Med. Phys., vol. 47, no. 5, pp. e218–e227, learning in medical imaging,’’ J. Digit. Imag., vol. 31, no. 3, pp. 283–289,
2020. Jun. 2018.
[5] M. R. Mohebian, H. R. Marateb, M. Mansourian, M. A. Mañanas, and [25] A. Saez, J. Sanchez-Monedero, P. A. Gutierrez, and C. Hervas-Martinez,
F. Mokarian, ‘‘A hybrid computer-aided-diagnosis system for prediction ‘‘Machine learning methods for binary and multiclass classification
of breast cancer recurrence (HPBCR) using optimized ensemble learn- of melanoma thickness from dermoscopic images,’’ IEEE Trans. Med.
ing,’’ Comput. Struct. Biotechnol. J., vol. 15, pp. 75–85, Jan. 2017. Imag., vol. 35, no. 4, pp. 1036–1045, Apr. 2016.

VOLUME 9, 2021 10669

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

[26] A. Rehman, S. Naz, M. I. Razzak, F. Akram, and M. Imran, [45] H. Danyali, M. S. Helfroush, and Z. Moshavash, ‘‘Robust leukocyte
‘‘A deep learning-based framework for automatic brain tumors classifi- segmentation in blood microscopic images based on intuitionistic fuzzy
cation using transfer learning,’’ Circuits, Syst., Signal Process., vol. 39, divergence,’’ in Proc. 22nd Iranian Conf. Biomed. Eng. (ICBME),
no. 2, pp. 757–775, Dec. 2013. Nov. 2015, pp. 275–280.
[27] H. Irshad, A. Veillard, L. Roux, and D. Racoceanu, ‘‘Methods for nuclei [46] S. Manik, L. M. Saini, and N. Vadera, ‘‘Counting and classification of
detection, segmentation, and classification in digital histopathology: white blood cell using artificial neural network (ANN),’’ in Proc. IEEE
A review—Current status and future potential,’’ IEEE Rev. Biomed. Eng., 1st Int. Conf. Power Electron., Intell. Control Energy Syst. (ICPEICES),
vol. 7, pp. 97–114, 2014. Jul. 2016, pp. 1–5.
[28] A. Brunetti, L. Carnimeo, G. F. Trotta, and V. Bevilacqua, ‘‘Computer- [47] Y. Li, R. Zhu, L. Mi, Y. Cao, and D. Yao, ‘‘Segmentation of white
assisted frameworks for classification of liver, breast and blood neoplasias blood cell from acute lymphoblastic leukemia images using dual-
via neural networks: A survey based on medical images,’’ Neurocomput- threshold method,’’ Comput. Math. Methods Med., vol. 2016, May 2016,
ing, vol. 335, pp. 274–298, Mar. 2019. Art. no. 9514707.
[29] T. Hussain, K. Muhammad, S. Khan, A. Ullah, M. Y. Lee, and S. W. Baik, [48] P. Ghosh, D. Bhattacharjee, and M. Nasipuri, ‘‘Blood smear analyzer
‘‘Intelligent baby behavior monitoring using embedded vision in IoT for for white blood cell counting: A hybrid microscopic image analyzing
smart healthcare centers,’’ J. Artif. Intell. Syst., vol. 1, no. 15, p. 2019, technique,’’ Appl. Soft Comput., vol. 46, pp. 629–638, Sep. 2016.
2019.
[49] Q. Wang, L. Chang, M. Zhou, Q. Li, H. Liu, and F. Guo, ‘‘A spectral
[30] M. I. Razzak and S. Naz, ‘‘Microscopic blood smear segmentation and and morphologic method for white blood cell classification,’’ Opt. Laser
classification using deep contour aware CNN and extreme machine learn- Technol., vol. 84, pp. 144–148, Oct. 2016.
ing,’’ in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. Workshops
[50] S. Ravikumar, ‘‘Image segmentation and classification of white blood
(CVPRW), Jul. 2017, pp. 49–55.
cells with the extreme learning machine and the fast relevance vec-
[31] S. Kar, K. Das Sharma, and M. Maitra, ‘‘Optimised feature selection
tor machine,’’ Artif. Cells, Nanomedicine, Biotechnol., vol. 44, no. 3,
for identification of carcinogenic leukocytes using weighted aggregation
pp. 985–989, 2016.
based transposition PSO,’’ IETE J. Res., vol. 4, pp. 1–14, Nov. 2019.
[32] J. Li, L. Lu, Y.-H. Zhang, Y. Xu, M. Liu, K. Feng, L. Chen, X. Kong, [51] O. Sarrafzadeh and A. M. Dehnavi, ‘‘Nucleus and cytoplasm segmenta-
T. Huang, and Y.-D. Cai, ‘‘Identification of leukemia stem cell expression tion in microscopic images using K-means clustering and region grow-
signatures through Monte Carlo feature selection strategy and support ing,’’ Adv. Biomed. Res., vol. 4, Dec. 2015.
vector machine,’’ Cancer Gene Therapy, vol. 27, nos. 1–2, pp. 56–69, [52] K. Yadanar Win, S. Choomchuay, and K. Hamamoto, ‘‘K mean clustering
Feb. 2020. based automated segmentation of overlapping cell nuclei in pleural effu-
[33] S. Kumar, A. L. Fred, H. A. Kumar, P. S. Varghese, and S. A. Jacob, sion cytology images,’’ in Proc. Int. Conf. Adv. Technol. Commun. (ATC),
‘‘Segmentation of anomalies in abdomen CT images by convolution Oct. 2017, pp. 265–269.
neural network and classification by fuzzy support vector machine,’’ in [53] A. Vard, N. Ghane, A. Talebi, and P. Nematollahy, ‘‘Segmentation of
Hybrid Machine Intelligence for Medical Image Analysis (Studies in white blood cells from microscopic images using a novel combination of
Computational Intelligence), vol. 841. Singapore: Springer, Aug. 2019, K-Means clustering and modified watershed algorithm,’’ J. Med. Signals
pp. 157–196. [Online]. Available: http://www.springer.com/series/7092, Sensors, vol. 7, no. 2, p. 92, 2017.
doi: 10.1007/978-981-13-8930-6. [54] A. S. Negm, O. A. Hassan, and A. H. Kandil, ‘‘A decision support system
[34] Y. Wang, F. Shi, L. Cao, N. Dey, Q. Wu, A. S. Ashour, R. S. Sherratt, for acute leukaemia classification based on digital microscopic images,’’
V. Rajinikanth, and L. Wu, ‘‘Morphological segmentation analysis and Alexandria Eng. J., vol. 57, no. 4, pp. 2319–2332, Dec. 2018.
texture-based support vector machines classification on mice liver fibro- [55] V. P. Ananthi and P. Balasubramaniam, ‘‘A new thresholding technique
sis microscopic images,’’ Current Bioinf., vol. 14, no. 4, pp. 282–294, based on fuzzy set as an application to leukocyte nucleus segmentation,’’
Apr. 2019. Comput. Methods Programs Biomed., vol. 134, pp. 165–177, Oct. 2016.
[35] A. H. Lubis, P. Sihombing, and E. B. Nababan, ‘‘Analysis of accuracy [56] M. A. Khosrosereshki and M. B. Menhaj, ‘‘A fuzzy based classifier
improvement in K-nearest neighbor using principal component analysis for diagnosis of acute lymphoblastic leukemia using blood smear image
(PCA),’’ J. Phys., Conf. Ser., vol. 1566, Jun. 2020, Art. no. 012062. processing,’’ in Proc. 5th Iranian Joint Congr. Fuzzy Intell. Syst. (CFIS),
[36] A. Gautam, P. Singh, B. Raman, and H. Bhadauria, ‘‘Automatic clas- Mar. 2017, pp. 13–18.
sification of leukocytes using morphological features and Naïve bayes [57] K. AL-Dulaimi, I. Tomeo-Reyes, J. Banks, and V. Chandran, ‘‘Evaluation
classifier,’’ in Proc. IEEE Region 10 Conf. (TENCON), Nov. 2016, and benchmarking of level set-based three forces via geometric active
pp. 1023–1027. contours for segmentation of white blood cell nuclei shape,’’ Comput.
[37] H. Gao, X. Zeng, and C. Yao, ‘‘Application of improved distributed naive Biol. Med., vol. 2012, pp. 1–12, Jan. 2020, Art. no. 923946.
Bayesian algorithms in text classification,’’ J. Supercomput., vol. 75, [58] A. Bouchet, S. Montes, V. Ballarin, and I. Díaz, ‘‘Intuitionistic fuzzy
no. 9, pp. 5831–5847, Sep. 2019. set and fuzzy mathematical morphology applied to color leukocytes
[38] E. Purwanti and E. Calista, ‘‘Detection of acute lymphocyte leukemia segmentation,’’ Signal, Image Video Process., vol. 14, no. 3, pp. 557–564,
using k-nearest neighbor algorithm based on shape and histogram fea- Apr. 2019.
tures,’’ J. Phys., Conf. Ser., vol. 853, May 2017, Art. no. 012011.
[59] N. I. Che Marzuki, N. H. Mahmood, and M. A. Abdul Razak, ‘‘Segmenta-
[39] M. B. Rajesh and S. Sathiamoorthy, ‘‘Classification of leukemia image
tion of white blood cell nucleus using active contour,’’ J. Teknol., vol. 74,
using genetic based K-nearest neighbor (G-KNN),’’ Asian J. Comput. Sci.
no. 6, pp. 115–118, May 2015.
Technol., vol. 7, no. 2, pp. 113–117, 2018.
[40] E. S. Wiharto, S. Palgunadi, Y. R. Putra, and E. Suryani, ‘‘Cells iden- [60] K. K. Jha and H. S. Dutta, ‘‘Mutual information based hybrid model and
tification of acute myeloid leukemia AML m0 and AML m1 using K- deep learning for acute lymphocytic leukemia detection in single cell
nearest neighbour based on morphological images,’’ in Proc. Int. Conf. blood smear images,’’ Comput. Methods Programs Biomed., vol. 179,
Data Softw. Eng. (ICoDSE), Nov. 2017, pp. 1–6. Oct. 2019, Art. no. 104987.
[41] S. N. M. Safuan, M. R. Md Tomari, and W. N. Wan Zakaria, ‘‘White blood [61] Z. Liu, J. Liu, X. Xiao, H. Yuan, X. Li, J. Chang, and C. Zheng, ‘‘Segmen-
cell (WBC) counting analysis in blood smear images using various color tation of white blood cells through nucleus mark watershed operations
segmentation methods,’’ Measurement, vol. 116, pp. 543–555, Feb. 2018. and mean shift clustering,’’ Sensors, vol. 15, no. 9, pp. 22561–22586,
[42] S. Mishra, B. Majhi, and P. K. Sa, ‘‘Texture feature based classification Sep. 2015.
on microscopic blood smear for acute lymphoblastic leukemia detection,’’ [62] L. Xie, J. Qi, L. Pan, and S. Wali, ‘‘Integrating deep convolutional
Biomed. Signal Process. Control, vol. 47, pp. 303–311, Jan. 2019. neural networks with marker-controlled watershed for overlapping nuclei
[43] M. A. Mohammed, M. K. Abd Ghani, N. Arunkumar, R. I. Hamed, segmentation in histopathology images,’’ Neurocomputing, vol. 376,
S. A. Mostafa, M. K. Abdullah, and M. A. Burhanuddin, ‘‘Decision pp. 166–179, Feb. 2020.
support system for nasopharyngeal carcinoma discrimination from endo- [63] S. Dai, K. Lu, J. Dong, Y. Zhang, and Y. Chen, ‘‘A novel approach of lung
scopic images using artificial neural network,’’ J. Supercomput., vol. 76, segmentation on chest CT images using graph cuts,’’ Neurocomputing,
no. 2, pp. 1086–1104, Feb. 2020. vol. 168, pp. 799–807, Nov. 2015.
[44] D.-C. Huang, K.-D. Hung, and Y.-K. Chan, ‘‘A computer assisted method [64] D. Mahapatra, ‘‘Semi-supervised learning and graph cuts for consen-
for leukocyte nucleus segmentation and recognition in blood smear sus based medical image segmentation,’’ Pattern Recognit., vol. 63,
images,’’ J. Syst. Softw., vol. 85, no. 9, pp. 2104–2118, Sep. 2012. pp. 700–709, Mar. 2017.

10670 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

[65] Q. Zheng, H. Li, B. Fan, S. Wu, and J. Xu, ‘‘Integrating support vector [85] G. Haskins, U. Kruger, and P. Yan, ‘‘Deep learning in medical
machine and graph cuts for medical image segmentation,’’ J. Vis. Com- image registration: A survey,’’ Mach. Vis. Appl., vol. 31, nos. 1–2,
mun. Image Represent., vol. 55, pp. 157–165, Aug. 2018. Feb. 2020.
[66] X. Zheng, Y. Wang, G. Wang, and J. Liu, ‘‘Fast and robust segmentation [86] G. Litjens, T. Kooi, B. E. Bejnordi, A. A. A. Setio, F. Ciompi,
of white blood cell images by self-supervised learning,’’ Micron, vol. 107, M. Ghafoorian, J. A. W. M. van der Laak, B. van Ginneken, and
pp. 55–71, Apr. 2018. C. I. Sánchez, ‘‘A survey on deep learning in medical image analysis,’’
[67] S. Kermani, M. Amin, and A. Talebi, ‘‘Recognition of acute lymphoblas- Med. Image Anal., vol. 42, pp. 60–88, Dec. 2017.
tic leukemia cells in microscopic images using k-means clustering and [87] G. Wang, ‘‘A perspective on deep imaging,’’ IEEE Access, vol. 4,
support vector machine classifier,’’ J. Med. Signals Sensors, vol. 5, no. 1, pp. 8914–8924, 2016.
p. 49, 2015. [88] I. Ul Haq, A. Ullah, K. Muhammad, M. Y. Lee, and S. W. Baik, ‘‘Person-
[68] S. Agaian, M. Madhukar, and A. T. Chronopoulos, ‘‘A new acute alized movie summarization using deep CNN-assisted facial expression
leukaemia-automated classification system,’’ Comput. Methods Biomech. recognition,’’ Complexity, vol. 2019, May 2019, Art. no. 3581419.
Biomed. Eng., Imag. Visualizat., vol. 6, no. 3, pp. 303–314, May 2018. [89] R. B. Hegde, K. Prasad, H. Hebbar, and B. M. K. Singh, ‘‘Development
[69] J. Zhao, M. Zhang, Z. Zhou, J. Chu, and F. Cao, ‘‘Automatic detection and of a robust algorithm for detection of nuclei of white blood cells in
classification of leukocytes using convolutional neural networks,’’ Med. peripheral blood smear images,’’ Multimedia Tools Appl., vol. 78, no. 13,
Biol. Eng. Comput., vol. 55, no. 8, pp. 1287–1301, Aug. 2017. pp. 17879–17898, Jul. 2019.
[70] R. B. Hegde, K. Prasad, H. Hebbar, B. M. K. Singh, and I. Sand- [90] M.-C. Su, C.-Y. Cheng, and P.-C. Wang, ‘‘A neural-network-based
hya, ‘‘Automated decision support system for detection of leukemia approach to white blood cell classification,’’ Sci. World J., vol. 2014,
from peripheral blood smear images,’’ J. Digit. Imag., vol. 7, pp. 1–14, Jan. 2014, Art. no. 796371.
Nov. 2019. [91] H. Lee and Y.-P.-P. Chen, ‘‘Cell morphology based classification for
[71] A. Sengur, Y. Akbulut, U. Budak, and Z. Comert, ‘‘White blood cell red cells in blood smear images,’’ Pattern Recognit. Lett., vol. 49,
classification based on shape and deep features,’’ in Proc. Int. Artif. Intell. pp. 155–161, Nov. 2014.
Data Process. Symp. (IDAP), Sep. 2019, pp. 1–4. [92] J. Rawat, A. Singh, H. Bhadauria, J. Virmani, and J. S. Devgun, ‘‘Appli-
[72] S. Tantikitti, S. Tumswadi, and W. Premchaiswadi, ‘‘Image processing cation of ensemble artificial neural network for the classification of white
for detection of dengue virus based on WBC classification and decision blood cells using microscopic blood images,’’ Int. J. Comput. Syst. Eng.,
tree,’’ in Proc. 13th Int. Conf. ICT Knowl. Eng., Nov. 2015, pp. 84–89. vol. 4, nos. 2–3, pp. 202–216, 2018.
[93] S. Nazlibilek, D. Karacor, T. Ercan, M. H. Sazli, O. Kalender, and
[73] J. Prinyakupt and C. Pluempitiwiriyawej, ‘‘Segmentation of white blood
Y. Ege, ‘‘Automatic segmentation, counting, size determination and
cells and comparison of cell morphology by linear and naïve Bayes
classification of white blood cells,’’ Measurement, vol. 55, no. 3,
classifiers,’’ Biomed. Eng. OnLine, vol. 14, no. 1, p. 63, Dec. 2015.
pp. 58–65, 2014.
[74] L. H. S. Vogado, R. M. S. Veras, F. H. D. Araujo, R. R. V. Silva,
[94] F. Sadeghian, Z. Seman, A. R. Ramli, B. H. A. Kahar, and M.-I. Saripan,
and K. R. T. Aires, ‘‘Leukemia diagnosis in blood slides using transfer
‘‘A framework for white blood cell segmentation in microscopic blood
learning in CNNs and SVM for classification,’’ Eng. Appl. Artif. Intell.,
images using digital image processing,’’ Biol. Procedures Online, vol. 11,
vol. 72, pp. 415–422, Jun. 2018.
no. 1, p. 196, 2009.
[75] J. Rawat, A. Singh, H. S. Bhadauria, J. Virmani, and J. S. Devgun, ‘‘Clas-
[95] W. Ullah, A. Ullah, I. U. Haq, K. Muhammad, M. Sajjad, and S. W. Baik,
sification of acute lymphoblastic leukaemia using hybrid hierarchical
‘‘CNN features with bi-directional LSTM for real-time anomaly detection
classifiers,’’ Multimedia Tools Appl., vol. 76, no. 18, pp. 19057–19085,
in surveillance networks,’’ Multimedia Tools Appl., vol. 4, pp. 1–17,
Sep. 2017.
Aug. 2020.
[76] T. Shaikhina, D. Lowe, S. Daga, D. Briggs, R. Higgins, and [96] A. Khatami, A. Khosravi, T. Nguyen, C. P. Lim, and S. Nahavandi, ‘‘Med-
N. Khovanova, ‘‘Decision tree and random forest models for ical image analysis using wavelet transform and deep belief networks,’’
outcome prediction in antibody incompatible kidney transplantation,’’ Expert Syst. Appl., vol. 86, pp. 190–198, Nov. 2017.
Biomed. Signal Process. Control, vol. 52, pp. 456–462, Jul. 2019.
[97] A. Kutlu, ‘‘A novel method for classifying liver and brain tumors using
[Online]. Available: https://creativecommons.org/licenses/by/4.0/,
convolutional neural networks, discrete wavelet transform and long short-
doi: 10.1016/j.bspc.2017.01.012.
term memory networks,’’ Sensors, vol. 19, no. 9, p. 1992, Apr. 2019.
[77] A. M. Abdeldaim, A. T. Sahlol, M. Elhoseny, and A. E. Hassanien, [98] J. Schlemper, O. Oktay, M. Schaap, M. Heinrich, B. Kainz, B. Glocker,
‘‘Computer-aided acute lymphoblastic leukemia diagnosis system based and D. Rueckert, ‘‘Attention gated networks: Learning to leverage salient
on image analysis,’’ in AAdvances in Soft Computing and Machine Learn- regions in medical images,’’ Med. Image Anal., vol. 53, pp. 197–207,
ing in Image Processing. Springer, 2018, pp. 131–147. Apr. 2019.
[78] A. Mathur, A. Tripathi, and M. Kuse, ‘‘Scalable system for classification [99] E. Ahn, A. Kumar, M. Fulham, D. Feng, and J. Kim, ‘‘Unsupervised
of white blood cells from leishman stained blood stain images,’’ J. Pathol. domain adaptation to classify medical images using zero-bias convo-
Informat., vol. 4, no. 2, p. 15, 2013. lutional auto-encoders and context-based feature augmentation,’’ IEEE
[79] M. Sajjad, S. Zahir, A. Ullah, Z. Akhtar, and K. Muhammad, ‘‘Human Trans. Med. Imag., vol. 39, no. 7, pp. 2385–2394, Jul. 2020.
behavior understanding in big multimedia data using CNN based [100] A. Ullah, K. Muhammad, T. Hussain, and S. W. Baik, ‘‘Conflux LSTMs
facial expression recognition,’’ Mobile Netw. Appl., vol. 25, no. 4, network: A novel approach for multi-view action recognition,’’ Neuro-
pp. 1611–1621, Aug. 2020. computing, Dec. 2020.
[80] A. Ullah, K. Muhammad, T. Hussain, S. W. Baik, and [101] P. P. Banik, R. Saha, and K.-D. Kim, ‘‘An automatic nucleus segmentation
V. H. C. De Albuquerque, ‘‘Event-oriented 3D convolutional features and CNN model based classification method of white blood cell,’’ Expert
selection and hash codes generation using PCA for video retrieval,’’ Syst. Appl., vol. 149, Jul. 2020, Art. no. 113211.
IEEE Access, vol. 8, pp. 196529–196540, 2020. [102] J. W. Choi, Y. Ku, B. W. Yoo, J.-A. Kim, D. S. Lee, Y. J. Chai, H.-J. Kong,
[81] A. Sabzi Shahrebabaki, A. S. Imran, N. Olfati, and T. Svendsen, and H. C. Kim, ‘‘White blood cell differential count of maturation stages
‘‘A comparative study of deep learning techniques on frame-level speech in bone marrow smear using dual-stage convolutional neural networks,’’
data classification,’’ Circuits, Syst., Signal Process., vol. 38, no. 8, PLoS ONE, vol. 12, no. 12, Dec. 2017, Art. no. e0189259.
pp. 3501–3520, Aug. 2019. [103] M. Karthikeyan and R. Venkatesan, ‘‘Interpolative leishman-stained
[82] Z. Kastrati, A. S. Imran, and S. Y. Yayilgan, ‘‘The impact of deep learning transformation invariant deep pattern classification for white blood cells,’’
on document classification using semantically rich representations,’’ Inf. Soft Comput., vol. 5, pp. 1–11, Jan. 2020.
Process. Manage., vol. 56, no. 5, pp. 1618–1632, Sep. 2019. [104] J. Liu, A. Shahroudy, M. Perez, G. Wang, L.-Y. Duan, and A. C. Kot,
[83] F. Altaf, S. M. S. Islam, N. Akhtar, and N. K. Janjua, ‘‘Going deep ‘‘NTU RGB+D 120: A large-scale benchmark for 3D human activity
in medical image analysis: Concepts, methods, challenges, and future understanding,’’ IEEE Trans. Pattern Anal. Mach. Intell., vol. 42, no. 10,
directions,’’ IEEE Access, vol. 7, pp. 99540–99572, 2019. pp. 2684–2701, Oct. 2019.
[84] H. Greenspan, B. van Ginneken, and R. M. Summers, ‘‘Guest edito- [105] H. Fan, F. Zhang, L. Xi, Z. Li, G. Liu, and Y. Xu, ‘‘LeukocyteMask: An
rial deep learning in medical imaging: Overview and future promise of automated localization and segmentation method for leukocyte in blood
an exciting new technique,’’ IEEE Trans. Med. Imag., vol. 35, no. 5, smear images using deep neural networks,’’ J. Biophotonics, vol. 12, no. 7,
pp. 1153–1159, May 2016. Jul. 2019, Art. no. e201800488.

VOLUME 9, 2021 10671

 S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

[106] R. B. Hegde, K. Prasad, H. Hebbar, and B. M. K. Singh, ‘‘Feature [126] T. Hussain, K. Muhammad, A. Ullah, J. Del Ser, A. H. Gandomi,
extraction using traditional image processing and convolutional neural M. Sajjad, S. W. Baik, and V. H. C. de Albuquerque, ‘‘Multi-view sum-
network methods to classify white blood cells: A study,’’ Australas. Phys. marization and activity recognition meet edge computing in IoT envi-
Eng. Sci. Med., vol. 42, no. 2, pp. 627–638, Jun. 2019. ronments,’’ IEEE Internet Things J., early access, Sep. 29, 2020, doi:
[107] A. Acevedo, S. Alférez, A. Merino, L. Puigví, and J. Rodellar, ‘‘Recog- 10.1109/JIOT.2020.3027483.
nition of peripheral blood cell images using convolutional neural net- [127] T. Han, K. Muhammad, T. Hussain, J. Lloret, and S. W. Baik, ‘‘An
works,’’ Comput. Methods Programs Biomed., vol. 180, Oct. 2019, efficient deep learning framework for intelligent energy management in
Art. no. 105020. IoT networks,’’ IEEE Internet Things J., early access, Aug. 3, 2020, doi:
[108] F. Qin, N. Gao, Y. Peng, Z. Wu, S. Shen, and A. Grudtsin, ‘‘Fine-grained 10.1109/JIOT.2020.3013306.
leukocyte classification with deep residual learning for microscopic [128] K. Muhammad, T. Hussain, and S. L. Baik, ‘‘Efficient CNN based
images,’’ Comput. Methods Programs Biomed., vol. 162, pp. 243–252, summarization of surveillance videos for resource-constrained devices,’’
Aug. 2018. Pattern Recognit. Lett., vol. 130, pp. 370–375, Feb. 2020.
[109] P. Tiwari, J. Qian, Q. Li, B. Wang, D. Gupta, A. Khanna, [129] C. Zhang, S. Bengio, M. Hardt, B. Recht, and O. Vinyals, ‘‘Under-
J. J. P. C. Rodrigues, and V. H. C. de Albuquerque, ‘‘Detection of standing deep learning requires rethinking generalization,’’ 2016,
subtype blood cells using deep learning,’’ Cognit. Syst. Res., vol. 52, arXiv:1611.03530. [Online]. Available: http://arxiv.org/abs/1611.03530
pp. 1036–1044, Dec. 2018. [130] I. Mehmood, T. Hussain, I. Hassan, S. Rho, and M. Sajjad, ‘‘Let the deaf
[110] J. Hung and A. Carpenter, ‘‘Applying faster R-CNN for object detection understand: Mainstreaming the marginalized in context with personalized
on malaria images,’’ in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. digital media services and social needs,’’ in Proc. IEEE Int. Conf. Multi-
Workshops, vol. 2017, pp. 56–61. media Expo Workshops (ICMEW), Jul. 2017, pp. 220–225.
[111] M. Habibzadeh, M. Jannesari, Z. Rezaei, H. Baharvand, and M. Totonchi, [131] B. J. Erickson, P. Korfiatis, Z. Akkus, T. Kline, and K. Philbrick, ‘‘Toolk-
‘‘Automatic white blood cell classification using pre-trained deep learn- its and libraries for deep learning,’’ J. Digit. Imag., vol. 30, no. 4,
ing models: Resnet and inception,’’ in Proc. 10th Int. Conf. Mach. Vis. pp. 400–405, Aug. 2017.
(ICMV), 2018, vol. 10696, Art. no. 1069612. [132] K. Muhammad, T. Hussain, M. Tanveer, G. Sannino, and
[112] Y. Song, L. Zhang, S. Chen, D. Ni, B. Li, Y. Zhou, B. Lei, and T. Wang, V. H. C. de Albuquerque, ‘‘Cost-effective video summarization using
‘‘A deep learning based framework for accurate segmentation of cervical deep CNN with hierarchical weighted fusion for IoT surveillance
cytoplasm and nuclei,’’ in Proc. 36th Annu. Int. Conf. IEEE Eng. Med. networks,’’ IEEE Internet Things J., vol. 7, no. 5, May 2020,
Biol. Soc., Aug. 2014, pp. 2903–2906. doi: 10.1109/JIOT.2019.2950469.
[133] C. Marzahl, M. Aubreville, J. Voigt, and A. Maier, ‘‘Classification of
[113] F. Özyurt, ‘‘A fused CNN model for WBC detection with MRMR fea-
leukemic b-lymphoblast cells from blood smear microscopic images
ture selection and extreme learning machine,’’ Soft Comput., vol. 24,
with an attention-based deep learning method and advanced augmenta-
pp. 8163–8172, Jun. 2019.
tion techniques,’’ in ISBI C-NMC Challenge: Classification Cancer Cell
[114] A. Rehman, N. Abbas, T. Saba, S. I. U. Rahman, Z. Mehmood, and
Imaging. Springer, 2019, pp. 13–22.
H. Kolivand, ‘‘Classification of acute lymphoblastic leukemia using
[134] P. Huang, J. Wang, J. Zhang, Y. Shen, C. Liu, W. Song, S. Wu, Y. Zuo,
deep learning,’’ Microsc. Res. Techn., vol. 81, no. 11, pp. 1310–1317,
Z. Lu, and D. Li, ‘‘Attention-aware residual network based manifold
Nov. 2018.
learning for white blood cells classification,’’ IEEE J. Biomed. Health
[115] D. Bani-Hani, N. Khan, F. Alsultan, S. Karanjkar, and N. Nagarur, ‘‘Clas- Informat., early access, Jul. 29, 2020, doi: 10.1109/JBHI.2020.3012711.
sification of leucocytes using convolutional neural network optimized
through genetic algorithm,’’ in Proc. 7th Annu. World Conf. Soc. Ind. Syst.
Eng., Oct. 2018, pp. 1–6. SIRAJ KHAN received the master’s degree in com-
[116] C. Di Ruberto, A. Loddo, and G. Puglisi, ‘‘Blob detection and deep puter vision and machine learning from Islamia
learning for leukemic blood image analysis,’’ Appl. Sci., vol. 10, no. 3, College University Peshawar, Pakistan, in 2017.
p. 1176, Feb. 2020. He is currently pursuing the Ph.D. degree with the
[117] M. Loey, M. Naman, and H. Zayed, ‘‘Deep transfer learning in diagnosing Digital Image Processing Laboratory, Department
leukemia in blood cells,’’ Computers, vol. 9, no. 2, p. 29, Apr. 2020. of Computer Science, Islamia College University
[118] L. Ma, R. Shuai, X. Ran, W. Liu, and C. Ye, ‘‘Combining DC-GAN with Peshawar. He is currently an Active Member with
ResNet for blood cell image classification,’’ Med. Biol. Eng. Comput., the Digital Image Processing Laboratory. He is
vol. 4, pp. 1–14, Mar. 2020.
also working as a Lecturer of Computer Science
[119] Y. Y. Baydilli and Ü. Atila, ‘‘Classification of white blood cells using
with the Government College of Management Sci-
capsule networks,’’ Comput. Med. Imag. Graph., vol. 80, Mar. 2020,
ence Khar, Bajaur, (Higher Education Department) Pakistan. His major
Art. no. 101699.
research interests include image processing, image segmentation, image
[120] R. R. Tobias, L. Carlo De Jesus, M. E. Mital, S. Lauguico, M. Guillermo,
R. R. Vicerra, A. Bandala, and E. Dadios, ‘‘Faster R-CNN model with
classification, healthcare services, and deep learning for medical image
momentum optimizer for RBC and WBC variants classification,’’ in analysis.
Proc. IEEE 2nd Global Conf. Life Sci. Technol. (LifeTech), Mar. 2020,
pp. 235–239. MUHAMMAD SAJJAD received the master’s
[121] X. Wang, T. Xu, J. Zhang, S. Chen, and Y. Zhang, ‘‘SO-YOLO based degree from the Department of Computer Sci-
WBC detection with Fourier ptychographic microscopy,’’ IEEE Access, ence, College of Signals, National University of
vol. 6, pp. 51566–51576, 2018. Sciences and Technology, Rawalpindi, Pakistan,
[122] L. Alzubaidi, M. A. Fadhel, O. Al-Shamma, J. Zhang, and Y. Duan, and the Ph.D. degree in digital contents from
‘‘Deep learning models for classification of red blood cells in microscopy Sejong University, Seoul, South Korea. He is cur-
images to aid in sickle cell anemia diagnosis,’’ Electronics, vol. 9, no. 3, rently an Assistant Professor with the Department
p. 427, Mar. 2020.
of Computer Science, Islamia College University
[123] S. Hosseinzadeh Kassani, P. Hosseinzadeh kassani, M. J. Wesolowski,
Peshawar, Pakistan. He is also the Head of the
K. A. Schneider, and R. Deters, ‘‘A hybrid deep learning architecture
Digital Image Processing Laboratory with Islamia
for leukemic B-lymphoblast classification,’’ 2019, arXiv:1909.11866.
[Online]. Available: http://arxiv.org/abs/1909.11866 College University Peshawar, where students are involved in research
[124] M. I. Razzak, S. Naz, and A. Zaib, ‘‘Deep learning for medical image pro- projects under his supervision, such as social data analysis, medical image
cessing: Overview, challenges and the future,’’ in Classification BioApps. analysis, multi-modal data mining and summarization, image/video prior-
Ithaca, NY, USA: Cornell Univ., 2018, pp. 323–350. itization and ranking, fog computing, the Internet of Things, virtual real-
[125] K. Muhammad, T. Hussain, J. Del Ser, V. Palade, and ity, and image/video retrieval. His primary research interests include com-
V. H. C. de Albuquerque, ‘‘DeepReS: A deep learning-based video puter vision, image understanding, pattern recognition, robot vision, and
summarization strategy for resource-constrained industrial surveillance multimedia applications, with current emphasis on raspberry-pi and deep
scenarios,’’ IEEE Trans. Ind. Informat., vol. 16, no. 9, pp. 5938–5947, learning-based bioinformatics, video scene understanding, activity analysis,
Sep. 2020. fog computing, the Internet of Things, and real-time tracking.

10672 VOLUME 9, 2021

S. Khan et al.: Review on TML and DL Models for WBCs Classification in Blood Smear Images

TANVEER HUSSAIN (Student Member, IEEE) ALI SHARIQ IMRAN (Member, IEEE) received
received the bachelor’s degree (Hons.) in com- the master’s degree in software engineering and
puter science from Islamia College University computing from the National University of Sci-
Peshawar, Peshawar, Pakistan, in 2017. He is ence and Technology (NUST), Pakistan, in 2008,
currently pursuing the joint master’s and Ph.D. and the Ph.D. degree in computer science from the
degrees with Sejong University, Seoul, South University of Oslo (UiO), Norway, in 2013. He is
Korea. He is currently serving as a Research Assis- currently an Associate Professor with the Depart-
tant with Intelligent Media Laboratory (IM Lab), ment of Computer Science, Norwegian Univer-
Sejong University. His major research domains sity of Science and Technology (NTNU), Norway.
are features extraction (learned and low-level fea- He has over 65 peer-reviewed journals and confer-
tures), video analytics, image processing, pattern recognition, medical ence publications to his name. He specializes in applied research focusing
image analysis, multimedia data retrieval, deep learning for multimedia on deep learning technology and its application to signal processing, natural
data understanding, single/multi-view video summarization, IoT, IIoT, and language processing, and the semantic Web. He is a member of the Nor-
resource-constrained programming. He has filed/published several patents wegian Colour and Visual Computing Laboratory (Colourlab) and the IEEE
and articles in peer-reviewed journals and conferences in reputed venues, Norway Section. He has served as a reviewer for many reputed journals over
including IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, IEEE INTERNET the years.
OF THINGS JOURNAL, IEEE Network Magazine, Pattern Recognition (Else-
vier), Neurocomputing, Pattern Recognition Letters, International Journal
of Energy Research (Wiley), International Journal of Distributed Sensors
Networks, and Multimedia Tools and Applications (Springer). He pro-
viding professional review services in various reputed journals, such as
IEEE TRANSACTIONS ON CYBERNETICS and IEEE TRANSACTIONS ON INDUSTRIAL
INFORMATICS. He is serving as an Associate Editor for Journal of Biomedical
and Journal of Biological Sciences and an Editorial Board Member for
the Journal of Artificial Intelligence and Systems. For further activities and
implementations, visit: https://github.com/tanveer-hussain

AMIN ULLAH (Member, IEEE) received the

Ph.D. degree in digital contents from Sejong Uni-
versity, South Korea. He is currently working as a
Postdoctoral Researcher with the Intelligent Media
Laboratory, Department of Software, Sejong Uni-
versity. His major research focus is on human
action and activity recognition, sequence learning,
image and video analytics, content-based indexing
and retrieval, IoT and smart cities, and deep learn-
ing for multimedia understanding. He has pub-
lished several papers in reputed peer-reviewed international journals and
conferences, including IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, IEEE
TRANSACTIONS ON INDUSTRIAL INFORMATICS, IEEE TRANSACTIONS ON INTELLIGENT
TRANSPORTATION SYSTEMS, IEEE INTERNET OF THINGS JOURNAL, IEEE ACCESS,
Future Generation Computer Systems (Elsevier), MDPI Sensors, Multime-
dia Tools and Applications (Springer), Mobile Networks and Applications
(Springer), and IEEE Joint Conference on Neural Networks.

VOLUME 9, 2021 10673