The Complete Guide To Artificial Intelligence in Radiology (2022)

The Complete Guide to
Artificial Intelligence
in Radiology
Calantic™ Digital Solutions The Complete Guide to Artificial Intelligence in Radiology
Contents
Introduction............................................ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 03
The case for artificial intelligence in radiology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04
Fundamentals of artificial intelligence....... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 05

Supervised learning................................................................................................................................................ 05
Unsupervised learning........................................................................................................................................... 06
Neural networks and deep learning. . ..................................................................................................................... 06
Performance evaluation......................................................................................................................................... 06
Internal and external validity................................................................................................................................ 07
Guidelines for evaluating AI research.................................................................................................................... 07
Clinical uses.. ........................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 08

Scheduling. . ............................................................................................................................................................ 08
Protocolling............................................................................................................................................................ 08
Image quality improvement and monitoring........................................................................................................ 09
Scan reading prioritization..................................................................................................................................... 10
Image interpretation.............................................................................................................................................. 10
Neurology.. .................................................................................................................................................. 10
Chest........................................................................................................................................................... 11
Breast.......................................................................................................................................................... 11
Cardiac........................................................................................................................................................ 12
Musculoskeletal.......................................................................................................................................... 12
Abdomen and pelvis.. .................................................................................................................................. 12
Obstacles to implementation. .................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Generalizability...................................................................................................................................................... 13
Risk of bias............................................................................................................................................................. 13
Data quantity, quality and variety......................................................................................................................... 14
Data protection and privacy. . ................................................................................................................................. 14
IT infrastructure..................................................................................................................................................... 14
Lack of standardization, interoperability, and integrability.................................................................................. 14
Interpretability....................................................................................................................................................... 15
Liability. . ................................................................................................................................................................. 15
Brittleness.............................................................................................................................................................. 15
Making purchasing decisions...................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Future trends........................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Conclusion.. . .. ........................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
References. . . .. .......................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2
Introduction
Artificial intelligence (AI) is playing a growing role in all

our lives and has shown promise in addressing some of
the greatest current and upcoming societal challenges
we face. The healthcare industry, though notoriously
complex and resistant to disruption, potentially has
a lot to gain from the use of AI. With an established
history of leading digital transformation in healthcare
and an urgent need for improved efficiency, radiology
has been at the forefront of harnessing AI’s potential.
This book covers how and why AI can address

challenges faced by radiology departments, provides
an overview of the fundamental concepts related
to AI, and describes some of the most promising
use cases for AI in radiology. In addition, the major
challenges associated with the adoption of AI into
routine radiological practice are discussed. The book
also covers some crucial points radiology departments
should keep in mind when deciding on which AI-based
solutions to purchase. Finally, it provides an outlook
on what new and evolving aspects of AI in radiology to
expect in the near future.
3
The case for artificial intelligence

in radiology
The healthcare industry has experienced a number With an ageing global population and a rising burden
of trends over the past few decades that demand of chronic illnesses, these issues are expected to pose
a change in the way certain things are done. These even more of a challenge to the healthcare industry in
trends are particularly salient in radiology, where the the future.
diagnostic quality of imaging scans has improved
dramatically while scan times have decreased. As a AI-based medical imaging solutions have the potential
result, the amount and complexity of medical imaging to ameliorate these challenges for several reasons.
data acquired have increased substantially over the They are particularly suited to handling large, complex
past few decades (Smith-Bindman et al., 2019; Winder datasets (Alzubaidi et al., 2021). Moreover, they are
et al., 2021) and are expected to continue to increase well suited to automate some of the tasks traditionally
(Tsao, 2020). This issue is complicated by a widespread performed by radiologists and radiographers,
global shortage of radiologists (AAMC Report Reinforces potentially freeing up time and making workflows
Mounting Physician Shortage, 2021, Clinical Radiology within radiology departments more efficient (Allen et
UK Workforce Census 2019 Report, 2019). Healthcare al., 2021; Baltruschat et al., 2021; Kalra et al., 2020;
workers, including radiologists, have an increasing O’Neill et al., 2021; van Leeuwen et al., 2021; Wong
workload (Bruls & Kwee, 2020; Levin et al., 2017) that et al., 2019). AI is also capable of detecting complex
contributes to burnout and medical errors (Harry et al., patterns in data that humans cannot necessarily find
2021). Being an essential service provider to virtually or quantify (Dance, 2021; Korteling et al., 2021; Kühl
all other hospital departments, staff shortages et al., 2020).
within radiology have significant effects that spread
throughout the hospital and to society as a whole
(England & Improvement, 2019; Sutherland et al., n.d.).
4
Fundamentals of artificial
intelligence
The term “artificial intelligence” refers to the use of One of the most common uses of machine learning is
computer systems to solve specific problems in a way in classification - assigning a piece of data a particular
that simulates human reasoning. One fundamental label. For example, a machine learning algorithm might
characteristic of AI is that, like humans, these systems be used to tell if a photo (the input) shows a dog or a
can tailor their solutions to changing circumstances. cat (the label). The algorithm can learn to do so in a
Note that, while these systems are meant to mimic on supervised or unsupervised way.
a fundamental level how humans think, their capacity
to do so (e.g. in terms of the amount of data they can Supervised learning
handle at one time, the nature and amount of patterns In supervised learning, the machine learning algorithm
they can find in the data, and the speed at which they is given data that has been labelled with the ground
do so) often exceeds that of humans. truth, in this example, photos of dogs and cats that
have been labelled as such. The process then goes
AI solutions come in the form of computer algorithms, through the following phases:
which are pieces of computer code representing
instructions to be followed to solve a specific problem. 1. Training phase: The algorithm learns the
In its most fundamental form, the algorithm takes features associated with dogs and cats using the
data as an input, performs some computation on that aforementioned data (training data).
data, and returns an output. 2. Test phase: The algorithm is then given a new
set of photos (the test data), it labels them and
An AI algorithm can be explicitly programmed to solve the performance of the algorithm on that data is
a specific task, analogous to a step-by-step recipe for assessed.
baking a cake. On the other hand, the algorithm can
be programmed to look for patterns within the data in In some cases, there is a phase in between training and
order to solve the problem. These types of algorithms test, known as the validation phase. In this phase, the
are known as machine learning algorithms. Thus, algorithm is given a new set of photos (not included
all machine learning algorithms are AI, but not all AI in either the training or test data), its performance
is machine learning. The patterns in the data that the is assessed on this data, and the model is tweaked
algorithm can be explicitly programmed to look for or and retrained on the training data. This is repeated
that it can “discover” by itself are known as features. until some predefined performance-based criterion is
An important characteristic of machine learning is that reached, and the algorithm then enters the test phase.
such algorithms learn from the data itself, and their
performance improves the more data they are given.
5
Unsupervised learning In regression, the most commonly used metrics include:

In unsupervised learning, the algorithm identifies ¡ Mean absolute error (MAE): the average difference
features within the input data that allow it to assign between the predicted values and the ground
classes to the individual data points without being told truth.
explicitly what those classes are or should be. Such ¡ Root mean square error (RMSE): the differences
algorithms can identify patterns or group data points between the predicted values and the ground
together without human intervention and include truth are squared and then averaged over the
clustering and dimensionality reduction algorithms. sample. Then the square root of the average is
Not all machine learning algorithms perform taken. Unlike the MAE, the RMSE thus gives higher
classification. Some are used to predict a continuous weight to larger differences.
metric (e.g. the temperature in four weeks’ time) instead ¡ R2: the proportion of the total variance in the
of a discrete label (e.g. cats vs dogs). These are known ground truth explained by the variance in the
as regression algorithms. predicted values. It ranges from 0 to 1.
Neural networks and deep learning The following metrics are commonly used in
A neural network is made up of an input layer and an classification tasks:
output layer, which are themselves composed of nodes. ¡ Accuracy: this is the proportion of all predictions
In simple neural networks, features that are manually that were predicted correctly. It ranges from 0 to 1.
derived from a dataset are fed into the input layer, ¡ Sensitivity: also known as the true positive rate
which performs some computations, the results of (TPR) or recall, this is the proportion of true
which are relayed to the output layer. In deep learning, positives that were predicted correctly. It ranges
multiple “hidden” layers exist between the input and from 0 to 1.
output layers. Each node of the hidden layers performs ¡ Specificity: Also known as the true negative rate
calculations using certain weights and relays the (TNR), this is the proportion of true negatives that
output to the next hidden layer until the output layer were predicted correctly. It ranges from 0 to 1.
is reached. ¡ Precision: also known as positive predictive
value (PPV), this is the proportion of positive
In the beginning, random values are assigned to the classifications that were predicted correctly. It
weights and the accuracy of the algorithm is calculated. ranges from 0 to 1.
The values of the weights are then iteratively adjusted
until a set of weight values that maximize accuracy An inherent trade-off exists between sensitivity and
is found. This iterative adjustment of the weight specificity. The relevant importance of each, as well
values is usually done by moving backwards from as their interpretation, highly depends on the specific
the output layer to the input layer, a technique called research question and classification task.
backpropagation. This entire process is done on the
training data. Importantly, although classification models are meant
to reach a binary conclusion, they are inherently
Performance evaluation probability-based. This means that these models will
Understanding how the performance of AI algorithms output a probability that a data point belongs to one
is assessed is key to interpreting the AI literature. class or another. In order to reach a conclusion on the
Several performance metrics exist for assessing how most likely class, a threshold is used. Metrics such as
well a model performs certain tasks. No single metric is accuracy, sensitivity, specificity and precision refer to
perfect, so a combination of several metrics provides a the performance of the algorithm based on a certain
fuller picture of model performance. threshold. The area under the receiver operating
6
characteristic curve (AUC) is a threshold-independent Guidelines for evaluating AI research

performance metric. The AUC can be interpreted as the Several guidelines have been developed to assess the
probability that a random positive example is ranked evidence behind AI-based interventions in healthcare
higher by the algorithm than a random negative (X. Liu et al., 2020; Mongan et al., 2020; Shelmerdine et
example. al., 2021; Weikert et al., 2021). These provide a template
for those doing AI research in healthcare and ensure
In image segmentation tasks, which are a type of that relevant information is reported transparently
classification task, the following metrics are commonly and comprehensively, but can also be used by other
used: stakeholders to assess the quality of published
¡ Dice similarity coefficient: a measure of overlap research. This helps ensure that AI-based solutions with
between two sets (e.g. two images) that is substantial potential or actual limitations, particularly
calculated as two times the number of elements those caused by poor reporting (Bozkurt et al., 2020;
common to the sets divided by the sum of the D. W. Kim et al., 2019; X. Liu et al., 2019; Nagendran
number of elements in each set. It ranges from 0 et al., 2020; Yusuf et al., 2020), are not prematurely
(no overlap) to 1 (perfect overlap). adopted (CONSORT-AI and SPIRIT-AI Steering Group,
¡ Hausdorff distance: a measure of how far two 2019). Guidelines have also been proposed for
sets (e.g. two images) within a space are far from evaluating the trustworthiness of AI-based solutions
each other. It is basically the largest distance from in terms of transparency, confidentiality, security, and
one point in one set to the closest point in the accountability (Buruk et al., 2020; Lekadir et al., 2021;
other set. Zicari et al., 2021).
Internal and external validity

Internally valid models perform well in their task on the
data being used to train and validate them. The degree
to which they are internally valid is assessed using the
performance metrics outlined above and depends on
the characteristics of the model itself and the quality of
the data that the model was trained and validated on.
Externally valid models perform well in their tasks on

new data (Ramspek et al., 2021). The better the model
performs on data that differs from the data the models
were trained and validated on, the higher the external
validity. In practice, this often requires the performance
of the models to be tested on data from hospitals or
geographical areas that were not part of the model’s
training and validation datasets.
7
Clinical uses
Over the past few years, AI has shown great potential appointments thus often requires the input of someone
in addressing a broad range of tasks within a medical with domain knowledge, which stipulates that either
imaging department, including many that happen the person making the appointments must be a
before the patient is scanned. Implementations of radiologist or radiology technician, or these people will
AI to improve the efficiency of radiology workflows have to provide input regularly. In either scenario, the
prior to patient scanning are sometimes referred to as process is somewhat inefficient and can potentially
“upstream AI” (Kapoor et al., 2020; M. L. Richardson et be streamlined using AI-based algorithms that check
al., 2021). scan indications and contraindications and provide the
people scheduling the scans with information about
Scheduling scan urgency (Letourneau-Guillon et al., 2020).
One promising upstream AI application is predicting
which patients are likely to miss their scan appointments. Protocolling
Missed appointments are associated with significantly Depending on hospital or clinic policy, the decision on
increased workload and costs (Dantas et al., 2018). what exact scan protocol a patient receives is usually
Using a Gradient Boosting approach, Nelson et al. made based on the information on the referring
predicted missed hospital magnetic resonance imaging physician’s scan request and the judgement of the
(MRI) appointments in the United Kingdom’s National radiologist. This is often supplemented by direct
Health Service (NHS) with high accuracy (Nelson communication between the referring physician and
et al., 2019). Their simulations also suggested that radiologist and the radiologist’s review of the patient’s
acting on the predictions of this model by targeting medical information. This process improves patient
patients who are likely to miss their appointments care (Boland et al., 2014) but can be time-consuming
would potentially yield a net benefit of several pounds and inefficient, particularly with modalities like MRI,
per appointment across a range of model thresholds where a large number of protocol permutations exist.
and missed appointment rates (Nelson et al., 2019). In one study, protocolling alone accounted for about
Similar results were recently found in a study of a single 6 % of the radiologist’s working time (Schemmel et al.,
hospital in Singapore. For the 6-month period following 2016). Radiologists are also often interrupted by tasks
the deployment of the predictive tool they were able such as protocolling when interpreting images, despite
to significantly reduce the no show rate from 19.3 % the fact that the latter is considered a radiologist’s
to 15.9 % which translated into a potential economic primary responsibility (Balint et al., 2014; J.-P. J. Yu et
benefit of $180,000 (Chong et. al., 2020). al., 2014).
Scheduling scans in a radiology department is a Interpretation of the narrative text of the referring
challenging endeavour because, although it is largely physician’s scan request has been attempted using
an administrative task, it depends heavily on medical natural language classifiers, the same technology used
information. The task of assigning patients to specific in chatbots and virtual assistants. Natural language
8
classifiers based on deep learning have shown promise These are gradually being replaced by deep-learning-
in assigning patients to either a contrast-enhanced or based reconstruction methods, which improve image
non-enhanced MRI protocol for musculoskeletal MRI, quality while maintaining low radiation doses (Akagi
with an accuracy of 83 % (Trivedi et al., 2018) and 94 % et al., 2019; H. Chen et al., 2017; Choe et al., 2019;
(Y. H. Lee, 2018). Similar algorithms have shown an Shan et al., 2019). This reconstruction is performed
accuracy of 95 % for predicting the appropriate brain on supercomputers on the CT scanner itself or on the
MRI protocol using a combination of up to 41 different cloud. The balance between radiation dose and image
MRI sequences (Brown & Marotta, 2018). Across a wide quality can be adjusted on a protocol-specific basis to
range of body regions, a deep-learning-based natural tailor scans to individual patients and clinical scenarios
language classifier decided based on the narrative text (McLeavy et al., 2021; Willemink & Noël, 2019). Such
of the scan requests whether to automatically assign approaches have found particular use when scanning
a specific computed tomography (CT) or MRI protocol children, pregnant women, and obese patients as well
(which it did with 95 % accuracy) or, in more difficult as CT scans of the urinary tract and heart (McLeavy et
cases, recommend a list of three most appropriate al., 2021).
protocols to the radiologist (which it did with 92 %
accuracy) (Kalra et al., 2020). AI-based solutions have also been used to speed up
scans while maintaining diagnostic quality. Scan time
AI has also been used to decide whether already reduction not only improves overall efficiency but also
protocolled scans need to be extended, a decision which contributes to an overall better patient experience
has to be made in real-time while the patient is inside and compliance with imaging examination. A multi-
the scanner. One such example is in prostate MRI, where centre study of spine MRI showed that a deep-learning-
a decision on whether to administer a contrast agent is based image reconstruction algorithm that enhanced
often made after the non-contrast sequences. Hötker images using filtering and detail-preserving noise
et al. found that a convolutional neural network (CNN) reduction reduced scan times by 40 % (Bash, Johnson,
assigned 78 % of patients to the appropriate prostate et al., 2021). For T1-weighted MRI scans of the brain, a
MRI protocol (Hötker et al., 2021). The sensitivity of similar algorithm that improves image sharpness and
the CNN for the need for contrast was 94.4 % with a reduces image noise reduced scan times by 60 % while
specificity of 68.8 % and only 2 % of patients in their maintaining the accuracy of brain region volumetry
study would have had to be called back for a contrast- compared to standard scans (Bash, Wang, et al., 2021).
enhanced scan (Hötker et al., 2021).
In routine radiological practice, images often contain
Image quality improvement and artefacts that reduce their interpretability. These
monitoring artefacts are the result of characteristics of the specific
Many AI-based solutions that work in the background imaging modality or protocol used or factors intrinsic
of radiology workflows to improve image quality have to the patient being scanned, such as the presence of
recently been established. These include solutions for foreign bodies or the patient moving during the scan.
monitoring image quality, reducing image artefacts, Particularly with MRI, imaging protocols that demand
improving spatial resolution, and speeding up scans. fast scanning often introduce certain artefacts to the
reconstructed image. In one study, a deep-learning-
Such solutions are entering the radiology mainstream, based algorithm reduced banding artefacts associated
particularly for computed tomography, which for with balanced steady-state free precession MRI
decades used established but artefact-prone methods sequences of the brain and knee (K. H. Kim & Park, 2017).
for reconstructing interpretable images from the raw For real-time imaging of the heart using MRI, another
sensor data (Deák et al., 2013; Singh et al., 2010). study found that the aliasing artefacts introduced
9
by the data undersampling were reduced by using a includes the time from image acquisition to viewing
deep-learning-based approach (Hauptmann et al., by the radiologist and the time to read and report
2019). The presence of metallic foreign bodies such as the scans) was reduced from 512 to 19 minutes in an
dental, orthopaedic or vascular implants is a common outpatient setting when such a worklist prioritization
patient-related factor causing image artefacts in both was used (Arbabshirani et al., 2018). A simulation
CT and MRI (Boas & Fleischmann, 2012; Hargreaves et study using AI-based worklist prioritization based
al., 2011). Although not yet well established, several on identifying urgent findings on chest radiographs
deep-learning-based approaches for reducing these (such as pneumothorax, pleural effusions, and foreign
artefacts have been investigated (Ghani & Clem Karl, bodies) also found a substantial reduction in the time it
2019; Puvanasunthararajah et al., 2021; Zhang & Yu, took to view and report the scans compared to standard
2018). Similar approaches are being tested for reducing workflow prioritization (Baltruschat et al., 2021).
motion-related artefacts in MRI (Tamada et al., 2020;
B. Zhao et al., 2022). Image interpretation
Currently, the majority of commercially available AI-
AI-based solutions for monitoring image quality based solutions in medical imaging focus on some
potentially reduce the need to call patients back to aspect of analyzing and interpreting images (Rezazade
repeat imaging examinations, which is a common Mehrizi et al., 2021; van Leeuwen et al., 2021). This
problem (Schreiber-Zinaman & Rosenkrantz, 2017). includes segmenting parts of the image (for surgical
A deep-learning-based algorithm that identifies the or radiation therapy targeting, for example), bringing
radiographic view acquired and extracts quality-related suspicious areas to radiologists’ attention, extracting
metrics from ankle radiographs was able to predict imaging biomarkers (radiomics), comparing images
image quality with about 94 % accuracy (Mairhöfer et across time, and reaching specific imaging diagnoses.
al., 2021). Another deep-learning-based approach was
capable of predicting nondiagnostic liver MRI scans Neurology
with a negative predictive value of between 86 % and ¡ 29–38 % of commercially available AI-based
94 % (Esses et al., 2018). This real-time automated applications in radiology (Rezazade Mehrizi et al.,
quality control potentially allows radiology technicians 2021; van Leeuwen et al., 2021).
to rerun scans or run additional scans with greater
diagnostic value. Most commercially available AI-based solutions
targeted at neuroimaging data aim to detect and
Scan reading prioritization characterize ischemic stroke, intracranial haemorrhage,
With staff shortages and increasing scan numbers, dementia, and multiple sclerosis (Olthof et al., 2020).
radiologists face long reading lists. To optimize Several studies have shown excellent accuracy of AI-
efficiency and patient care, AI-based solutions have based methods for the detection and classification
been suggested as a way to prioritize which scans of intraparenchymal, subarachnoid, and subdural
radiologists read and report first, usually by screening haemorrhage on head CT (Flanders et al., 2020; Ker et
acquired images for findings that require urgent al., 2019; Kuo et al., 2019). Subsequent studies showed
intervention (O’Connor & Bhalla, 2021). This has been that, compared to radiologists, some AI-based solutions
most extensively studied in neuroradiology, where have substantially lower false positive and negative
moving CT scans that were found to have intracranial rates (Ginat, 2020; Rao et al., 2021). In ischemic
haemorrhage by an AI-based tool to the top of the stroke, AI-based solutions have largely focused on
reading list reduced the time it took radiologists to the quantification of the infarct core (Goebel et al.,
view the scans by several minutes (O’Neill et al., 2021). 2018; Maegerlein et al., 2019), the detection of large
Another study found that the time-to diagnosis (which vessel occlusion (Matsoukas et al., 2022; Morey et al.,
10
2021; Murray et al., 2020; Shlobin et al., 2022), and Outside the emergency setting, AI-based solutions have
the prediction of stroke outcomes (Bacchi et al., 2020; been widely tested and implemented for tuberculosis
Nielsen et al., 2018; Y. Yu et al., 2020, 2021). screening on chest radiographs (E. J. Hwang, Park, et
al., 2019; S. Hwang et al., 2016; Khan et al., 2020;
In multiple sclerosis, AI has been used to identify Qin et al., 2019; WHO Operational Handbook on
and segment lesions (Nair et al., 2020; S.-H. Wang et Tuberculosis Module 2: Screening – Systematic Screening
al., 2018), which can be particularly helpful for the for Tuberculosis Disease, n.d.). In addition, they have
longitudinal follow-up of patients. It has also been used been useful for lung cancer screening both in terms of
to extract imaging features associated with progressive detecting lung nodules on CT (Setio et al., 2017) and
disease and conversion from clinically isolated chest radiographs (Li et al., 2020) and by classifying
syndrome to definite multiple sclerosis (Narayana et whether nodules are likely to be malignant or benign
al., 2020; Yoo et al., 2019). Other applications of AI in (Ardila et al., 2019; Bonavita et al., 2020; Ciompi et al.,
neuroradiology include the detection of intracranial 2017; B. Wu et al., 2018). AI-based solutions also show
aneurysms (Faron et al., 2020; Nakao et al., 2018; Ueda great promise for the diagnosis of pneumonia, chronic
et al., 2019) and the segmentation of brain tumours obstructive pulmonary disease, and interstitial lung
(Kao et al., 2019; Mlynarski et al., 2019; Zhou et al., disease (F. Liu et al., 2021).
2020) as well as the prediction of brain tumour genetic
markers from imaging data (Choi et al., 2019; J. Zhao Breast
et al., 2020) ¡ 11 % of commercially available AI-based applications
in radiology (Rezazade Mehrizi et al., 2021; van
Chest Leeuwen et al., 2021).
¡ 24 %–31 % of commercially available AI-based
applications in radiology (Rezazade Mehrizi et al., So far, many of the AI-based algorithms targeting breast
2021; van Leeuwen et al., 2021). imaging aim to reduce the workload of radiologists
reading mammograms. Ways to do this have included
When interpreting chest radiographs, radiologists using AI-based algorithms to triage out negative
detected substantially more critical and urgent findings mammograms, which in one study was associated
when aided by a deep-learning-based algorithm, and with a reduction in radiologists’ workload by almost
did so much faster than without the algorithm (Nam one-fifth (Yala et al., 2019). Other studies that have
et al., 2021). Deep-learning-based image interpretation replaced second readers of mammograms with AI-
algorithms have also been found to improve radiology based algorithms have shown that this leads to fewer
residents’ sensitivity for detecting urgent findings on false positives and false negatives as well as reduces
chest radiographs from 66 % to 73 % (E. J. Hwang, Nam, the workload of the second reader by 88 % (McKinney
et al., 2019). Another study which focused on a broader et al., 2020).
range of findings on chest radiographs also found that
radiologists aided by a deep-learning-based algorithm AI-based solutions for mammography have also
had higher diagnostic accuracy than radiologists who been found to increase the diagnostic accuracy of
read the radiographs without assistance (Seah et al., radiologists (McKinney et al., 2020; Rodríguez-Ruiz et
2021). The uses of AI in chest radiology also extend al., 2019; Watanabe et al., 2019) and some have been
to cross-sectional imaging like CT. A deep learning found to be highly accurate in independently detecting
algorithm was found to detect pulmonary embolism and classifying breast lesions (Agnes et al., 2019; Al-
on CT scans with high accuracy (AUC = 0.85) (Huang, Antari et al., 2020; Rodriguez-Ruiz et al., 2019).
Kothari, et al., 2020). Moreover, a deep learning Despite this, a recent systematic review of 36 AI-
algorithm was 90
% accurate in detecting aortic based algorithms found that these studies were of
dissection on non-contrast-enhanced CT scans, similar poor methodological quality and that all algorithms
to the performance of radiologists (Hata et al., 2021). were less accurate than the consensus of two or more
11
radiologists (Freeman et al., 2021). AI-based algorithms Promising applications of AI in the assessment of
have nonetheless shown potential for extracting muscles, bones and joints include applications where
cancer-predictive features from mammograms human readers generally show poor between- and
beyond mammographic breast density (Arefan et al., within-rater reliability, such as the determination of
2020; Dembrower et al., 2020; Hinton et al., 2019). skeletal age based on bone radiographs (Halabi et
Beyond mammography, AI-based solutions have been al., 2019; Thodberg et al., 2009) and screening for
developed for detecting and classifying breast lesions osteoporosis on radiographs (Kathirvelu et al., 2019;
on ultrasound (Akkus et al., 2019; Park et al., 2019; G.- J.-S. Lee et al., 2019) and CT (Pan et al., 2020). AI-
G. Wu et al., 2019) and MRI (Herent et al., 2019). based solutions have also shown promise for detecting
fractures on radiographs and CT (Lindsey et al.,
2018; Olczak et al., 2017; Urakawa et al., 2019). One
Cardiac
systematic review of AI-based solutions for fracture
¡ 11 % of commercially available AI-based applications
detection in several different body parts showed AUCs
in radiology (Rezazade Mehrizi et al., 2021; van
ranging from 0.94 to 1.00 and accuracies of 77 % to
Leeuwen et al., 2021).
98 % (Langerhuizen et al., 2019). AI-based solutions
have also achieved accuracies similar to radiologists for
Cardiac radiology has always been particularly classification of the severity of degenerative changes
challenging because of the difficulties inherent in of the spine (Jamaludin et al., 2017) and extremity
acquiring images of a constantly moving organ. joints (F. Liu et al., 2018; Thomas et al., 2020). AI-based
Because of this, it has benefited immensely from solutions have also been developed to determine the
advances in imaging technology and seems set to origin of skeletal metastases (Lang et al., 2019) and the
benefit greatly from AI as well (Sermesant et al., 2021). classification of primary bone tumours (Do et al., 2017).
Most of the AI-based applications of the cardiovascular
system use MRI, CT or ultrasound data (Weikert et al., Abdomen and pelvis
2021). Prominent examples include the automated ¡ 4 % of commercially available AI-based applications
calculation of ejection fraction on echocardiography, in radiology (Rezazade Mehrizi et al., 2021; van
quantification of coronary artery calcification on Leeuwen et al., 2021).
cardiac CT, determination of right ventricular volume
on CT pulmonary angiography, and determination Much of the efforts in using AI in abdominal imaging
of heart chamber size and thickness on cardiac MRI have thus far concentrated on the automated
(Medical AI Evaluation, n.d., The Medical Futurist, n.d.). segmentation of organs such as the liver (Dou et al.,
AI-based solutions for the prediction of patients 2017), spleen (Moon et al., 2019), pancreas (Oktay
likely to respond favourably to cardiac interventions, et al., 2018), and kidneys (Sharma et al., 2017). In
such as cardiac resynchronization therapy, based on addition, a systematic review of 11 studies using deep
imaging and clinical parameters have also shown great learning for the detection of malignant liver masses
promise (Cikes et al., 2019; Hu et al., 2019). Changes showed accuracies of up to 97 % and AUCs of up to 0.92
in cardiac MRI not readily visible to human readers but (Azer, 2019).
potentially useful for differentiating different types
of cardiomyopathies can also be detected using AI Other applications of AI in abdominal radiology include
through texture analysis (Neisius et al., 2019; J. Wang the detection of liver fibrosis (He et al., 2019; Yasaka et
et al., 2020) and other radiomic approaches (Mancio et al., 2018), fatty liver disease, hepatic iron content, the
al., 2022). detection of free abdominal gas on CT, and automated
volumetry and segmentation of the prostate (AI for
Musculoskeletal Radiology, n.d.).
¡ 7–11 % of commercially available AI-based
applications in radiology (Rezazade Mehrizi et al.,
2021; van Leeuwen et al., 2021). 12
Obstacles to implementation
Despite the great potential of AI in medical imaging, AI solutions are often developed in a high-resource
it has yet to find widespread implementation and environment such as large technology companies
impact in routine clinical practice. This research-to- and academic medical centres in wealthy countries.
clinic translation is being hindered by several complex It is likely that findings and performance in these
and interrelated issues that directly or indirectly lower high-resource contexts will fail to generalize to lower-
the likelihood of AI-based solutions being adopted. One resource contexts such as smaller hospitals, rural areas
major way they do so is by creating a lack of trust in AI- or poorer countries (Price & Nicholson, 2019), which
based solutions by key stakeholders such as regulators, complicates the issue further.
healthcare professionals and patients (Cadario et al.,
2021; Esmaeilzadeh, 2020; J. P. Richardson et al., 2021; Risk of bias
Tucci et al., 2022). Biases can arise in AI-based solutions due to data or
human factors. The former occurs when the data used
Generalizability to train the AI solution does not adequately represent
One major challenge is to develop AI-based solutions the target population. Datasets can be unrepresentative
that continue to perform well in new, real-world when they are too small or have been collected in a way
scenarios. In a large systematic review, almost half that misrepresents a certain population category. AI
of the studied AI-based medical imaging algorithms solutions trained on unrepresentative data perpetuate
reported a greater than 0.05 decrease in the AUC when biases and perform poorly in the population categories
tested on new data (A. C. Yu et al., 2022). This lack of underrepresented or misrepresented in the training
generalizability can lead to adverse effects on how well data. The presence of such biases has been empirically
the model performs in a real-world scenario. shown in many AI-based medical imaging studies
(Larrazabal et al., 2020; Seyyed-Kalantari et al., 2021).
If a solution performs poorly when tested on a dataset
with a similar or identical distribution to the training AI-based solutions are prone to several subjective and
dataset, it is said to lack narrow generalizability and is sometimes implicitly or explicitly prejudiced decisions
often a consequence of overfitting (Eche et al., 2021). during their development by humans. These human
Potential solutions for overfitting are using larger factors include how the training data is selected, how
training datasets and reducing the model’s complexity. it is labelled, and how the decision is made to focus on
If a solution performs poorly when tested on a dataset the specific problem the AI-based solution intends to
with a different distribution to the training dataset (e.g. solve (Norori et al., 2021). Some recommendations and
a different distribution of patient ethnicities), it is said to tools are available to help minimize the risk of bias in AI
lack broad generalizability (Eche et al., 2021). Solutions research (AIF360: A Comprehensive Set of Fairness Metrics
to poor broad generalizability include stress-testing the for Datasets and Machine Learning Models, Explanations
model on datasets with different distributions from the for These Metrics, and Algorithms to Mitigate Bias in
training dataset (Eche et al., 2021). Datasets and Models, n.d., IBM Watson Studio - Model
Risk Management, n.d.; Silberg & Manyika, 2019).
13
Data quantity, quality and variety IT infrastructure

Problems such as bias and lack of generalizability Among hospital departments, radiology has always
can be mitigated by ensuring that training data is of been at the forefront of digitalization. AI-based solutions
sufficient quantity, quality and variety. However, this that focus on image processing and interpretation
is difficult to do because patients are often reluctant are likely to find the prerequisite infrastructure in
to share their data for commercial purposes (Aggarwal, most radiology departments, for example for linking
Farag, et al., 2021; Ghafur et al., 2020; Trinidad et al., imaging equipment to computers for analysis and for
2020), hospitals and clinics are usually not equipped archiving images and other outputs. However, most
to make this data available in a useable and secure radiology departments are likely to require significant
manner, and organizing and labelling the data is time- infrastructure upgrades for other applications of
consuming and expensive. AI, particularly those requiring the integration of
information from multiple sources and having complex
Many datasets can be used for a number of different outputs. Moreover, it is important to keep in mind that
purposes, and sharing data between companies the distribution of necessary infrastructure is highly
can help make the process of data collection and unequal across and within countries (Health Ethics &
organization more efficient, as well as increase the Governance, 2021).
amount of data available for each application. However,
developers are often reluctant to share data with each In terms of computing power, radiology departments
other, or even reveal the exact source of their data, to will either have to invest resources into the hardware
stay competitive. and personnel necessary to run these AI-based
solutions or opt for cloud-based solutions. The former
Data protection and privacy comes with an extra cost but allows data processing
The development and implementation of AI-based within the confines of the hospital or clinic’s local
solutions require that patients are explicitly informed network. Cloud-based solutions for computing (known
about, and give their consent to, the use of their data as “infrastructure as a service” or “IaaS”) are often
for a particular purpose and by certain people. This data considered the less secure and less trustworthy option,
also has to be adequately protected from data breaches but this depends on a number of factors and is thus not
and misuse. Failure to ensure this greatly undermines always true (Baccianella & Gough, n.d.). Guidelines on
the public’s trust in AI-based solutions and hinders what to consider when procuring cloud-based solutions
their adoption. While regulations governing health data in healthcare are available (Cloud Security for Healthcare
privacy state that the collection of fully anonymized Services, 2021).
data does not require explicit patient consent (General
Data Protection Regulation (GDPR) – Official Legal Text, Lack of standardization, interoperability,
2016; Office for Civil Rights (OCR), 2012) and in theory and integrability
protects from the data being misused, whether or not The problem of infrastructure becomes even more
imaging data can be fully anonymized is controversial complicated when considering how fragmented the
(Lotan et al., 2020; Murdoch, 2021). Whether consent AI medical imaging market currently is (Alexander
can be truly informed considering the complexity et al., 2020). It is therefore likely that in the near
of the data being acquired, and the resulting myriad future a single department will have several dozen
of potential future uses of the data, is also disputed AI-based solutions from different vendors running
(Vayena & Blasimme, 2017). simultaneously. Having a separate self-contained
infrastructure (e.g. a workstation or server) for each of
these would be incredibly complicated and difficult to
manage. Suggested solutions for this have included AI
14
solution “marketplaces”, similar to app stores (Advanced inherently work in a way that is independent of what
AI Solutions for Radiology, n.d., Curated Marketplace, their developers could have foreseen (Yeung, 2018).
2018, Imaging AI Marketplace - Overview, n.d., Sectra To the end-user such as the healthcare worker, the AI-
Amplifier Marketplace, 2021, The Nuance AI Marketplace based solution may be opaque and so they may not
for Diagnostic Imaging, n.d.), and development of an be able to tell when the solution is malfunctioning or
overarching vendor-neutral infrastructure (Leiner et inaccurate (Habli et al., 2020; Yeung, 2018).
al., 2021). The successful implementation of such
solutions requires close partnerships between AI Brittleness
solution developers, imaging vendors and information Despite substantial progress in their development over
technology companies. the past few years, deep learning algorithms are still
surprising brittle. This means that, when the algorithm
Interpretability faces a scenario that differs substantially from what
It is often impossible to understand exactly how AI- it faced during training, it cannot contextualize and
based solutions come to their conclusions, particularly often produces nonsensical or inaccurate results. This
with complex approaches like deep learning. This happens because, unlike humans, most algorithms
reduces how transparent the decision-making process learn to perceive things within the confines of certain
for procuring and approving these solutions can be, assumptions, but fail to generalize outside these
makes the identification of biases difficult, and makes assumptions. As an example of how this can be abused
it harder for clinicians to explain the outputs of these with malicious intent, subtle changes to medical
solutions to their patients and to determine whether images, imperceptible by humans, can render the
a solution is working properly or has malfunctioned results of disease-classifying algorithms inaccurate
(Char et al., 2018; Reddy et al., 2020; Vayena et al., (Finlayson et al., 2018). The lack of interpretability
2018; Whittlestone et al., 2019). Some have suggested of many AI-based solutions compounds this problem
that techniques that help humans understand how AI- because it makes it difficult to troubleshoot how they
based algorithms made certain decisions or predictions reached the wrong conclusion.
(“interpretable” or “explainable” AI) might help mitigate
these challenges. However, others have argued that
currently available techniques are unsuitable for
understanding individual decisions of an algorithm and
have warned against relying on them for ensuring that
algorithms work in a safe and reliable way (Ghassemi
et al., 2021).
Liability
In healthcare systems, a framework of accountability
ensures that healthcare workers and medical
institutions can be held responsible for adverse effects
resulting from their actions. The question of who
should be held accountable for the failures of an AI-
based solution is complicated. For pharmaceuticals, for
example, the accountability for inherent failures in the
product or its use often lies with either the manufacturer
or the prescriber. One key difference is that AI-based
systems are continuously evolving and learning, and so
15
Making purchasing decisions
So far, more than 100 AI-based products have gained exactly what a potential AI-based product’s scope of
conformité européenne (CE) marking or Food and Drug the solution is - i.e. what specific problem the AI-based
Adminstration (FDA) clearance. These products can be solution is designed to solve and in what specific
found in continuously updated and searchable online circumstances. This includes whether the solution
databases curated by the FDA (Center for Devices & is intended for screening, diagnosis, monitoring,
Radiological Health, n.d.), the American College of treatment recommendation or another application. It
Radiology (Assess-AI, n.d.), and others (AI for Radiology, also includes the intended users of the solution and
n.d., The Medical Futurist, n.d.; E. Wu et al., 2021). The what kind of specific qualifications or training they
increasing number of available products, the inherent are expected to have in order to be able to operate the
complexity of many of these solutions, and the fact solution and interpret its outputs. It needs to be clear
that many people who usually make purchasing to buyers whether the solution is intended to replace
decisions in hospitals are not familiar with evaluating certain tasks that would normally be performed by
such products make it important to think carefully the end-user, act as a double-reader, as a triaging
when deciding on which product to purchase. Such mechanism, or for other tasks like quality control.
decisions will need to be made after incorporating input Buyers should also understand whether the solution is
from healthcare workers, information technology (IT) intended to provide “new” information (i.e. information
professionals, as well as management, finance, legal, that would otherwise be unavailable to the user
and human resources professionals within hospitals. without the solution), improve the performance of an
existing task beyond a human’s or other non-AI-based
Deciding on whether to purchase an AI-based solution solution’s performance or if it is intended to save time
in radiology, as well as which of the increasing number or other resources.
of commercially available solutions to purchase,
includes considerations of quality, safety, and finances. Buyers should also have access to information
Over the past few years, several guidelines have that allows them to assess the potential benefits
emerged to help potential buyers make these decisions of the AI solution, and this should be backed up by
(A Buyer’s Guide to AI in Health and Care, 2020; Omoumi published scientific evidence for the efficacy and
et al., 2021; Reddy et al., 2021), and these guidelines cost-efficiency of the solution. How this is done will
are likely to evolve in the future with changing depend highly on the solution itself and the context
expectations from customers, regulatory bodies, and in which it is expected to be deployed, but guidelines
stakeholders involved in reimbursement decisions. for this are available (National Institute for Health
and Care Excellence (NICE), n.d.). Some questions to
First of all, it has to be clear to the potential buyer ask here would be: How much of an influence will
what the problem is and whether AI is the appropriate the solution have on patient management? Will it
approach to this solution, or whether alternatives improve diagnostic performance? Will it save time and
exist that are more advantageous on balance. If AI money? Will it affect patients’ quality of life? It should
is the appropriate approach, buyers should know also be clear to the buyer who exactly is expected to
16
benefit from the use of this solution (Radiologists? of the solution’s outputs), or if it can readily be
Clinicians? Patients? The healthcare system or society integrated into the existing information technology
as a whole?). infrastructure of the buyer’s organization influences
the overall cost of the solution for the buyer and is
As with any healthcare intervention, all AI-based therefore also a critical consideration. In addition,
solutions come with potential risks, and these should the degree of manual interaction required, both
be made clear to the buyer. Some of these risks might under normal circumstances and for troubleshooting,
have legal consequences, such as the potential for should be known to the buyer. All potential users of
misdiagnosis. These risks should be quantified, and the AI solution should be involved in the purchasing
potential buyers should have a framework for dealing process to ensure that they are familiar with it and
with them, including identifying a framework for that it meets their professional ethical standards and
accountability within the organizations implementing suits their needs.
these solutions. Buyers should also ensure they
clearly understand the potential negative effects on From a regulatory perspective, it should be clear to
radiologists’ training and the potential disruption to the buyer whether the solution complies with medical
radiologists’ workflows associated with the use of device and data protection regulations. Has the
these solutions. solution been approved in the buyer’s country? If so,
under which risk classification? Buyers should also
Specifics of the AI solution’s design are also relevant consider creating data flow maps that display how the
to the decision on whether or not to purchase it. These data flows in the operation of the AI-based solution,
include how robust the solution is to differences including who has access to the data.
between vendors and scanning parameters, the
circumstances under which the algorithm was trained Finally, there are other factors to consider which are
(including potential confounding factors), and the way not necessarily unique to AI-based solutions and which
that performance was assessed. It should also be clear buyers might be familiar with from purchasing other
to buyers if and how potential sources of bias were types of solutions. This includes the licensing model
accounted for during development. Because a core of the solution, how users are to be trained on using
characteristic of AI-based solutions is their ability to the solution, how the solution is maintained, how
continuously learn from new data, whether and how failures in the solution are dealt with, and whether
exactly this retraining is incorporated into the solution additional costs are to be expected when scaling up
with time should also be clear to the buyer, including the solution’s implementation (e.g. using the solution
whether or not new regulatory approval is needed for more imaging equipment or more users). This
with each iteration. This also includes whether or not allows the potential buyer to anticipate the current
retraining is required, for example, due to changes in and future costs of purchasing the solution.
imaging equipment at the buyer’s institution.
The main selling points of many AI-based solutions

are ease-of-use and improved workflows. Therefore,
potential buyers should carefully scrutinize how
these solutions are to be integrated into existing
workflows, including inter-operability with PACS and
electronic medical record systems. Whether or not
the solution requires extra hardware (e.g. graphical
processing units) or software (e.g. for visualization
17
Future trends
The past decade of increasing interest and progress 2021) despite the fact that the typical patient in a
in AI-based solutions for medical imaging has set the hospital receives multiple imaging scans during their
stage for a number of trends that are likely to appear stay (Shinagare et al., 2014). With this, it is also likely
or intensify in the near future. that more AI-based solutions will be developed that
target hitherto neglected modalities such as nuclear
Firstly, there is an increasing sentiment that, although imaging techniques and ultrasound.
AI holds a great deal of promise for interpretive
applications (such as the detection of pathology), The current market for AI-based solutions in
non-interpretive AI-based solutions might hold radiology is spread across a relatively large number
the most potential in terms of instilling efficiency of companies (Alexander et al., 2020). Potential users
into radiology workflows and improving patient are likely to expect a streamlined integration of these
experiences. This trend towards involving AI earlier in products in their workflows, which can be challenging
the patient management process is likely to extend in such a fragmented market. Improved integration
to AI increasingly acting as a clinical decision support can be achieved in several different ways, including
system to guide when and which imaging scans are with vendor-neutral marketplaces or by the gradual
performed. consolidation of providers of AI-based solutions.
For this to happen, AI needs to be integrated into With the expanding use of AI, the issue of trust between
existing clinical information systems, and the specific AI developers, healthcare professionals, regulators,
algorithms used need to be able to handle more varied and patients will become more relevant. It is therefore
data. This will likely pave the way for the development likely that efforts will intensify to take steps towards
of algorithms that are capable of integrating strengthening that trust. This will potentially include
demographic, clinical, and laboratory patient data to raising the expected standards of evidence for AI-
make recommendations about patient management based solutions (Aggarwal, Sounderajah, et al., 2021;
(Huang, Pareek, et al., 2020; Rockenbach, 2021). The X. Liu et al., 2019; van Leeuwen et al., 2021; Yusuf et
previously mentioned natural language processing al., 2020), making them more transparent through the
algorithms that have been used to interpret scan use and improvement of interpretable AI techniques
requests may be useful candidates for this. (Holzinger et al., 2017; Reyes et al., 2020; “Towards
Trustable Machine Learning,” 2018), and enhancing
In addition, we are likely to see AI algorithms that techniques for maintaining patient data privacy (G.
can interpret multiple different types of imaging data Kaissis et al., 2021; G. A. Kaissis et al., 2020).
from the same patient. Currently, less than 5 % of
commercially available AI-based solutions in medical Furthermore, while most existing regulations stipulate
imaging work with more than one imaging modality that AI-based algorithms cannot be modified after
(Rezazade Mehrizi et al., 2021; van Leeuwen et al., regulatory approval, this is likely to change in the
18
future. The potential for these algorithms to learn from

data acquired after approval and adapt to changing
circumstances is a major advantage of AI. Still,
frameworks for doing so have thus far been lacking in
the healthcare sector. However, promising ideas have
recently emerged, including adapting existing hospital
quality assurance and improvement frameworks to
monitor AI-based algorithms’ performance and the
data they are trained on and update the algorithms
accordingly (Feng et al., 2022). This will likely require
the development of multidisciplinary teams within
hospitals consisting of clinicians, IT professionals, and
biostatisticians who closely collaborate with model
developers and regulators (Feng et al., 2022).
While the obstacles discussed in previous sections

might slow down the adoption of AI in radiology
somewhat, the fear of AI potentially replacing
radiologists is unlikely to be one of them. A recent
survey from Europe showed that most radiologists
did not perceive a reduction in their clinical workload
after adopting AI-based solutions (European Society
of Radiology (ESR), 2022), likely because, at the same
time, demand for radiologists’ services has been
continuously rising. Studies from around the world
have shown that radiology professionals, particularly
those with AI exposure and experience, are generally
optimistic about the role of AI in their practice (Y. Chen
et al., 2021; Huisman et al., 2021; Ooi et al., 2021;
Santomartino & Yi, 2022; Scott et al., 2021).
19
Conclusion
AI has shown promise in positively impacting

virtually every facet of a radiology department’s
work - from scheduling and protocolling patient
scans to interpreting images and reaching diagnoses.
Promising research on AI-based tools in radiology has
not yet been widely translated to adoption in routine
practice, however, because of a number of complex,
partially intertwined issues. Potential solutions exist
for many of these challenges, but many of these
solutions require further refinement and testing.
In the meantime, guidelines are emerging to help
potential users of AI-based solutions in radiology
navigate the increasing number of commercial
products. This encourages their adoption in real-world
scenarios, thus allowing their true potential to be
uncovered, as well as their weaknesses to be identified
and addressed in a safe and effective way. As these
incremental improvements are made, these tools
will likely evolve to handle more varied data, become
integrated into consolidated workflows, become more
transparent, and ultimately more useful for increasing
efficiency and improving patient care.
20
References
AAMC Report Reinforces Mounting Physician Shortage. Akkus, Z., Cai, J., Boonrod, A., Zeinoddini, A., Weston, A. D.,
(2021). AAMC. https://www.aamc.org/news-insights/press- Philbrick, K. A., & Erickson, B. J. (2019). A Survey of Deep-
releases/aamc-report-reinforces-mounting-physician-shortage. Learning Applications in Ultrasound: Artificial Intelligence-
Accessed 17/11/2022 Powered Ultrasound for Improving Clinical Workflow. Journal of
the American College of Radiology: JACR, 16(9 Pt B), 1318–1328.
A buyer’s guide to AI in health and care. (2020). NHS https://doi.org/10.1016/j.jacr.2019.06.004
Transformation Directorate. https://www.nhsx.nhs.uk/ai-lab/
explore-all-resources/adopt-ai/a-buyers-guide-to-ai-in-health- Al-Antari, M. A., Al-Masni, M. A., & Kim, T.-S. (2020). Deep
and-care/. Accessed 17/11/2022 Learning Computer-Aided Diagnosis for Breast Lesion in Digital
Mammogram. Advances in Experimental Medicine and Biology,
Advanced AI solutions for radiology. (n.d.). Calantic Website. 1213, 59–72. https://doi.org/10.1007/978-3-030-33128-3_4
Retrieved July 3, 2022, from https://aivisions.calantic.com/
Alexander, A., Jiang, A., Ferreira, C., & Zurkiya, D. (2020).
Aggarwal, R., Farag, S., Martin, G., Ashrafian, H., & Darzi, A. An Intelligent Future for Medical Imaging: A Market Outlook
(2021). Patient Perceptions on Data Sharing and Applying on Artificial Intelligence for Medical Imaging. Journal of the
Artificial Intelligence to Health Care Data: Cross-sectional American College of Radiology: JACR, 17(1 Pt B), 165–170. https://
Survey. Journal of Medical Internet Research, 23(8), e26162. doi.org/10.1016/j.jacr.2019.07.019
https://doi.org/10.2196/26162
Allen, B., Agarwal, S., Coombs, L., Wald, C., & Dreyer, K.
Aggarwal, R., Sounderajah, V., Martin, G., Ting, D. S. W., (2021). 2020 ACR Data Science Institute Artificial Intelligence
Karthikesalingam, A., King, D., Ashrafian, H., & Darzi, Journal of the American College of Radiology: JACR. 2021
A. (2021). Diagnostic accuracy of deep learning in medical Aug;18(8):1153-1159.
imaging: a systematic review and meta-analysis. NPJ Digital
Medicine, 4(1), 65. https://doi.org/10.1038/s41746-021-00438-z Alzubaidi, L., Zhang, J., Humaidi, A. J., Al-Dujaili, A., Duan, Y.,
Al-Shamma, O., Santamaría, J., Fadhel, M. A., Al-Amidie, M.,
Agnes, S. A., Anitha, J., Pandian, S. I. A., & Peter, J. D. (2019). & Farhan, L. (2021). Review of deep learning: concepts, CNN
Classification of Mammogram Images Using Multiscale all architectures, challenges, applications, future directions. J Big
Convolutional Neural Network (MA-CNN). Journal of Medical Data. 2021;8(1):53.
Systems, 44(1), 30. https://doi.org/10.1007/s10916-019-1494-z
Arbabshirani, M. R., Fornwalt, B. K., Mongelluzzo, G. J.,
AIF360: A comprehensive set of fairness metrics for datasets Suever, J. D., Geise, B. D., Patel, A. A., & Moore, G. J. (2018).
and machine learning models, explanations for these Advanced machine learning in action: identification of
metrics, and algorithms to mitigate bias in datasets and intracranial hemorrhage on computed tomography scans of the
models. (n.d.). Github. Retrieved June 11, 2022, from https:// head with clinical workflow integration. NPJ Digital Medicine, 1,
github.com/Trusted-AI/AIF360 9. https://doi.org/10.1038/s41746-017-0015-z
AI for radiology. (n.d.). Retrieved June 26, Ardila, D., Kiraly, A. P., Bharadwaj, S., Choi, B., Reicher, J. J.,
2022, from https://grand-challenge.org/ Peng, L., Tse, D., Etemadi, M., Ye, W., Corrado, G., Naidich, D.
aiforradiology/?subspeciality=Abdomen&modality=All&ce_ P., & Shetty, S. (2019). End-to-end lung cancer screening with
under=All&ce_class=All&fda_class=All&sort_by=last %20 three-dimensional deep learning on low-dose chest computed
modified&search= tomography. Nature Medicine, 25(6), 954–961. https://doi.
org/10.1038/s41591-019-0447-x
Akagi, M., Nakamura, Y., Higaki, T., Narita, K., Honda, Y.,
Zhou, J., Yu, Z., Akino, N., & Awai, K. (2019). Deep learning Arefan, D., Mohamed, A. A., Berg, W. A., Zuley, M. L., Sumkin,
reconstruction improves image quality of abdominal ultra-high- J. H., & Wu, S. (2020). Deep learning modeling using normal
resolution CT. European Radiology, 29(11), 6163–6171. https:// mammograms for predicting breast cancer risk. Medical Physics,
doi.org/10.1007/s00330-019-06170-3 47(1), 110–118. https://doi.org/10.1002/mp.13886 21
Assess-AI. (n.d.). Retrieved July 2, 2022, from https://www. Bonavita, I., Rafael-Palou, X., Ceresa, M., Piella, G.,
acrdsi.org/DSI-Services/Assess-AI Ribas, V., & González Ballester, M. A. (2020). Integration
of convolutional neural networks for pulmonary nodule
Azer, S. A. (2019). Deep learning with convolutional neural malignancy assessment in a lung cancer classification pipeline.
networks for identification of liver masses and hepatocellular Computer Methods and Programs in Biomedicine, 185, 105172.
carcinoma: A systematic review. World Journal of Gastrointestinal https://doi.org/10.1016/j.cmpb.2019.105172
Oncology, 11(12), 1218–1230. https://doi.org/10.4251/wjgo.v11.
i12.1218 Bozkurt, S., Cahan, E. M., Seneviratne, M. G., Sun, R., Lossio-
Ventura, J. A., Ioannidis, J. P. A., & Hernandez-Boussard, T.
Bacchi, S., Zerner, T., Oakden-Rayner, L., Kleinig, T., Patel, (2020). Reporting of demographic data and representativeness
S., & Jannes, J. (2020). Deep Learning in the Prediction of in machine learning models using electronic health records.
Ischaemic Stroke Thrombolysis Functional Outcomes: A Journal of the American Medical Informatics Association: JAMIA,
Pilot Study. Academic Radiology, 27(2), e19–e23. https://doi. 27(12), 1878–1884. https://doi.org/10.1093/jamia/ocaa164
org/10.1016/j.acra.2019.03.015
Brown, A. D., & Marotta, T. R. (2018). Using machine learning
Baccianella, S., & Gough, T. (n.d.). Why cloud computing is the for sequence-level automated MRI protocol selection in
best option for hospitals adopting AI. Retrieved June 11, 2022, neuroradiology. Journal of the American Medical Informatics
from https://www.aidence.com/articles/cloud-best-option- Association: JAMIA, 25(5), 568–571. https://doi.org/10.1093/
imaging-ai/ jamia/ocx125
Balint, B. J., Steenburg, S. D., Lin, H., Shen, C., Steele, J. L., & Bruls, R. J. M., & Kwee, R. M. (2020). Workload for radiologists
Gunderman, R. B. (2014). Do telephone call interruptions have during on-call hours: dramatic increase in the past 15 years.
an impact on radiology resident diagnostic accuracy? Academic Insights into Imaging, 11(1), 121. https://doi.org/10.1186/
Radiology, 21(12), 1623–1628. https://doi.org/10.1016/j. s13244-020-00925-z
acra.2014.08.001
Buruk, B., Ekmekci, P. E., & Arda, B. (2020). A critical
Baltruschat, I., Steinmeister, L., Nickisch, H., Saalbach, A., perspective on guidelines for responsible and trustworthy
Grass, M., Adam, G., Knopp, T., & Ittrich, H. (2021). Smart artificial intelligence. Medicine, Health Care, and Philosophy,
chest X-ray worklist prioritization using artificial intelligence: 23(3), 387–399. https://doi.org/10.1007/s11019-020-09948-1
a clinical workflow simulation. European Radiology, 31(6), 3837–
3845. https://doi.org/10.1007/s00330-020-07480-7 Cadario, R., Longoni, C., & Morewedge, C. K. (2021).
Understanding, explaining, and utilizing medical artificial
Bash, S., Johnson, B., Gibbs, W., Zhang, T., intelligence. Nature Human Behaviour, 5(12), 1636–1642.
Shankaranarayanan, A., & Tanenbaum, L. N. (2021). https://doi.org/10.1038/s41562-021-01146-0
Deep Learning Image Processing Enables 40 % Faster Spinal
MR Scans Which Match or Exceed Quality of Standard of Center for Devices, & Radiological Health. (n.d.). Artificial
Care : A Prospective Multicenter Multireader Study. Clinical Intelligence and Machine Learning (AI/ML)-Enabled Medical
Neuroradiology. https://doi.org/10.1007/s00062-021-01121-2 Devices. U.S. Food and Drug Administration; FDA. Retrieved July
2, 2022, from https://www.fda.gov/medical-devices/software-
Bash, S., Wang, L., Airriess, C., Zaharchuk, G., Gong, E., medical-device-samd/artificial-intelligence-and-machine-
Shankaranarayanan, A., & Tanenbaum, L. N. (2021). Deep learning-aiml-enabled-medical-devices
Learning Enables 60 % Accelerated Volumetric Brain MRI While
Preserving Quantitative Performance: A Prospective, Multicenter, Char, D. S., Shah, N. H., & Magnus, D. (2018). Implementing
Multireader Trial. AJNR. American Journal of Neuroradiology, Machine Learning in Health Care - Addressing Ethical
42(12), 2130–2137. https://doi.org/10.3174/ajnr.A7358 Challenges. The New England Journal of Medicine, 378(11),
981–983. https://doi.org/10.1056/NEJMp1714229
Boas, F. E., & Fleischmann, D. (2012). CT artifacts: causes
and reduction techniques. Imaging in Medicine, 4(2), 229–240. Chen, H., Zhang, Y., Kalra, M. K., Lin, F., Chen, Y., Liao, P.,
https://doi.org/10.2217/iim.12.13 Zhou, J., & Wang, G. (2017). Low-Dose CT With a Residual
Encoder-Decoder Convolutional Neural Network. IEEE
Boland, G. W., Duszak, R., Jr, & Kalra, M. (2014). Protocol Transactions on Medical Imaging, 36(12), 2524–2535. https://doi.
design and optimization. Journal of the American College of org/10.1109/TMI.2017.2715284
Radiology: JACR, 11(5), 440–441. https://doi.org/10.1016/j.
jacr.2014.01.021
22
Chen, Y., Stavropoulou, C., Narasinkan, R., Baker, A., Dantas, L. F., Fleck, J. L., Cyrino Oliveira, F. L., & Hamacher,
& Scarbrough, H. (2021). Professionals’ responses to S. (2018). No-shows in appointment scheduling - a systematic
the introduction of AI innovations in radiology and their literature review. Health Policy, 122(4), 412–421. https://doi.
implications for future adoption: a qualitative study. BMC org/10.1016/j.healthpol.2018.02.002
Health Services Research, 21(1), 813. https://doi.org/10.1186/
s12913-021-06861-y Deák, Z., Grimm, J. M., Treitl, M., Geyer, L. L., Linsenmaier,
U., Körner, M., Reiser, M. F., & Wirth, S. (2013). Filtered back
Choe, J., Lee, S. M., Do, K.-H., Lee, G., Lee, J.-G., Lee, S. M., projection, adaptive statistical iterative reconstruction, and
& Seo, J. B. (2019). Deep Learning-based Image Conversion of a model-based iterative reconstruction in abdominal CT: an
CT Reconstruction Kernels Improves Radiomics Reproducibility experimental clinical study. Radiology, 266(1), 197–206. https://
for Pulmonary Nodules or Masses. Radiology, 292(2), 365–373. doi.org/10.1148/radiol.12112707
https://doi.org/10.1148/radiol.2019181960
Dembrower, K., Liu, Y., Azizpour, H., Eklund, M., Smith,
Choi, K. S., Choi, S. H., & Jeong, B. (2019). Prediction of IDH K., Lindholm, P., & Strand, F. (2020). Comparison of a Deep
genotype in gliomas with dynamic susceptibility contrast Learning Risk Score and Standard Mammographic Density Score
perfusion MR imaging using an explainable recurrent neural for Breast Cancer Risk Prediction. Radiology, 294(2), 265–272.
network. Neuro-Oncology, 21(9), 1197–1209. https://doi. https://doi.org/10.1148/radiol.2019190872
org/10.1093/neuonc/noz095
Do, B. H., Langlotz, C., & Beaulieu, C. F. (2017). Bone Tumor
Chong, L. R., Tsai, K. T., Lee, L. L., Foo, S. G., & Chang, P. Diagnosis Using a Naïve Bayesian Model of Demographic
C. (2020). Artificial Intelligence Predictive Analytics in the and Radiographic Features. Journal of Digital Imaging, 30(5),
Management of Outpatient MRI Appointment No-Shows. AJR. 640–647. https://doi.org/10.1007/s10278-017-0001-7
American Journal of Roentgenology, 215(5), 1155–1162. https://
doi.org/10.2214/AJR.19.22594 Dou, Q., Yu, L., Chen, H., Jin, Y., Yang, X., Qin, J., & Heng,
P.-A. (2017). 3D deeply supervised network for automated
Cikes, M., Sanchez-Martinez, S., Claggett, B., Duchateau, N., segmentation of volumetric medical images. Medical
Piella, G., Butakoff, C., Pouleur, A. C., Knappe, D., Biering- Image Analysis, 41, 40–54. https://doi.org/10.1016/j.
Sørensen, T., Kutyifa, V., Moss, A., Stein, K., Solomon, S. D., media.2017.05.001
& Bijnens, B. (2019). Machine learning-based phenogrouping in
heart failure to identify responders to cardiac resynchronization Eche, T., Schwartz, L. H., Mokrane, F.-Z., & Dercle, L.
therapy. European Journal of Heart Failure, 21(1), 74–85. https:// (2021). Toward Generalizability in the Deployment of Artificial
doi.org/10.1002/ejhf.1333 Intelligence in Radiology: Role of Computation Stress Testing to
Overcome Underspecification. Radiology. Artificial Intelligence,
Ciompi, F., Chung, K., van Riel, S. J., Setio, A. A. A., Gerke, 3(6), e210097. https://doi.org/10.1148/ryai.2021210097
P. K., Jacobs, C., Scholten, E. T., Schaefer-Prokop, C., Wille,
M. M. W., Marchianò, A., Pastorino, U., Prokop, M., & van England, N. H. S., & Improvement, N. H. S. (2019). NHS
Ginneken, B. (2017). Towards automatic pulmonary nodule diagnostic waiting times and activity data. NHS. https://www.
management in lung cancer screening with deep learning. england.nhs.uk/statistics/wp-content/uploads/sites/2/2021/12/
Scientific Reports, 7, 46479. https://doi.org/10.1038/srep46479 DWTA-Report-October-2021_M43D4.pdf. Accessed 17/11/2022
Clinical radiology UK workforce census 2019 report. (2019). Esmaeilzadeh, P. (2020). Use of AI-based tools for healthcare
https://www.rcr.ac.uk/publication/clinical-radiology-uk- purposes: a survey study from consumers’ perspectives. BMC
workforce-census-2019-report. Accessed 17/11/2022 Medical Informatics and Decision Making, 20(1), 170. https://doi.
org/10.1186/s12911-020-01191-1
Cloud security for healthcare services. (2021, January
14). ENISA. https://www.enisa.europa.eu/publications/cloud- Esses, S. J., Lu, X., Zhao, T., Shanbhogue, K., Dane, B., Bruno,
security-for-healthcare-services/ M., & Chandarana, H. (2018). Automated image quality
evaluation of T2 -weighted liver MRI utilizing deep learning
CONSORT-AI and SPIRIT-AI Steering Group. (2019). Reporting architecture. Journal of Magnetic Resonance Imaging: JMRI, 47(3),
guidelines for clinical trials evaluating artificial intelligence 723–728. https://doi.org/10.1002/jmri.25779
interventions are needed. Nature Medicine, 25(10), 1467–1468.
https://doi.org/10.1038/s41591-019-0603-3 European Society of Radiology (ESR). (2022). Current practical
experience with artificial intelligence in clinical radiology:
Curated marketplace. (2018, May 22). Blackford. https://www. a survey of the European Society of Radiology. Insights into
blackfordanalysis.com/applications/ Imaging, 13(1), 107. https://doi.org/10.1186/s13244-022-
01247-y
Dance, A. (2021). AI spots cell structures that humans can’t.
Nature. 592 (7852), 154–155.
23
Faron, A., Sichtermann, T., Teichert, N., Luetkens, J. A., Ginat, D. T. (2020). Analysis of head CT scans flagged by
Keulers, A., Nikoubashman, O., Freiherr, J., Mpotsaris, A., & deep learning software for acute intracranial hemorrhage.
Wiesmann, M. (2020). Performance of a Deep-Learning Neural Neuroradiology, 62(3), 335–340. https://doi.org/10.1007/
Network to Detect Intracranial Aneurysms from 3D TOF-MRA s00234-019-02330-w
Compared to Human Readers. Clinical Neuroradiology, 30(3),
591–598. https://doi.org/10.1007/s00062-019-00809-w Goebel, J., Stenzel, E., Guberina, N., Wanke, I., Koehrmann,
M., Kleinschnitz, C., Umutlu, L., Forsting, M., Moenninghoff,
Feng, J., Phillips, R. V., Malenica, I., Bishara, A., Hubbard, C., & Radbruch, A. (2018). Automated ASPECT rating:
A. E., Celi, L. A., & Pirracchio, R. (2022). Clinical artificial comparison between the Frontier ASPECT Score software and the
intelligence quality improvement: towards continual Brainomix software. Neuroradiology, 60(12), 1267–1272. https://
monitoring and updating of AI algorithms in healthcare. NPJ doi.org/10.1007/s00234-018-2098-x
Digital Medicine, 5(1), 66. https://doi.org/10.1038/s41746-022-
00611-y Habli, I., Lawton, T., & Porter, Z. (2020). Artificial intelligence
in health care: accountability and safety. Bulletin of the World
Finlayson, S. G., Chung, H. W., Kohane, I. S., & Beam, A. L. Health Organization, 98(4), 251–256. https://doi.org/10.2471/
(2018). Adversarial Attacks Against Medical Deep Learning BLT.19.237487
Systems. In arXiv [cs.CR]. arXiv. https://doi.org/10.1145/nnnnnnn.
nnnnnnn. Accessed 17/11/2022 Halabi, S. S., Prevedello, L. M., Kalpathy-Cramer, J.,
Mamonov, A. B., Bilbily, A., Cicero, M., Pan, I., Pereira, L. A.,
Flanders, A. E., Prevedello, L. M., Shih, G., Halabi, S. S., Sousa, R. T., Abdala, N., Kitamura, F. C., Thodberg, H. H.,
Kalpathy-Cramer, J., Ball, R., Mongan, J. T., Stein, A., Chen, L., Shih, G., Andriole, K., Kohli, M. D., Erickson, B. J., &
Kitamura, F. C., Lungren, M. P., Choudhary, G., Cala, L., Flanders, A. E. (2019). The RSNA Pediatric Bone Age Machine
Coelho, L., Mogensen, M., Morón, F., Miller, E., Ikuta, I., Learning Challenge. Radiology, 290(2), 498–503. https://doi.
Zohrabian, V., McDonnell, O., … RSNA-ASNR 2019 Brain org/10.1148/radiol.2018180736
Hemorrhage CT Annotators. (2020). Construction of a Machine
Learning Dataset through Collaboration: The RSNA 2019 Brain Hargreaves, B. A., Worters, P. W., Pauly, K. B., Pauly, J. M.,
CT Hemorrhage Challenge. Radiology. Artificial Intelligence, 2(3), Koch, K. M., & Gold, G. E. (2011). Metal-induced artifacts in
e190211. https://doi.org/10.1148/ryai.2020190211 MRI. AJR. American Journal of Roentgenology, 197(3), 547–555.
https://doi.org/10.2214/AJR.11.7364
Freeman, K., Geppert, J., Stinton, C., Todkill, D., Johnson,
S., Clarke, A., & Taylor-Phillips, S. (2021). Use of artificial Harry, E., Sinsky, C., Dyrbye, L. N., Makowski, M. S., Trockel,
intelligence for image analysis in breast cancer screening M., Tutty, M., Carlasare, L. E., West, C. P., & Shanafelt, T. D.
programmes: systematic review of test accuracy. BMJ , 374, (2021). Physician Task Load and the Risk of Burnout Among
n1872. https://doi.org/10.1136/bmj.n1872 US Physicians in a National Survey. Joint Commission Journal on
Quality and Patient Safety / Joint Commission Resources, 47(2),
General Data Protection Regulation (GDPR) – Official Legal 76–85. https://doi.org/10.1016/j.jcjq.2020.09.011
Text. (2016, July 13). General Data Protection Regulation
(GDPR). https://gdpr-info.eu/ Hata, A., Yanagawa, M., Yamagata, K., Suzuki, Y., Kido, S.,
Kawata, A., Doi, S., Yoshida, Y., Miyata, T., Tsubamoto, M.,
Ghafur, S., Van Dael, J., Leis, M., Darzi, A., & Sheikh, A. Kikuchi, N., & Tomiyama, N. (2021). Deep learning algorithm
(2020). Public perceptions on data sharing: key insights from for detection of aortic dissection on non-contrast-enhanced CT.
the UK and the USA. The Lancet. Digital Health, 2(9), e444–e446. European Radiology, 31(2), 1151–1159. https://doi.org/10.1007/
https://doi.org/10.1016/S2589-7500(20)30161-8 s00330-020-07213-w
Ghani, M. U., & Clem Karl, W. (2019). Fast Enhanced CT Metal Hauptmann, A., Arridge, S., Lucka, F., Muthurangu, V.,
Artifact Reduction using Data Domain Deep Learning. In arXiv & Steeden, J. A. (2019). Real-time cardiovascular MR with
[cs.CV]. arXiv. http://arxiv.org/abs/1904.04691. Accessed spatio-temporal artifact suppression using deep learning-proof
17/11/2022 of concept in congenital heart disease. Magnetic Resonance in
Medicine: Official Journal of the Society of Magnetic Resonance
Ghassemi, M., Oakden-Rayner, L., & Beam, A. L. (2021). in Medicine / Society of Magnetic Resonance in Medicine, 81(2),
The false hope of current approaches to explainable artificial 1143–1156. https://doi.org/10.1002/mrm.27480
intelligence in health care. The Lancet. Digital Health, 3(11),
e745–e750. https://doi.org/10.1016/S2589-7500(21)00208-9
24
Health Ethics & Governance. (2021, June 28). Ethics and Huisman, M., Ranschaert, E., Parker, W., Mastrodicasa, D.,
governance of artificial intelligence for health. World Koci, M., Pinto de Santos, D., Coppola, F., Morozov, S., Zins,
Health Organization. https://www.who.int/publications/i/ M., Bohyn, C., Koç, U., Wu, J., Veean, S., Fleischmann, D.,
item/9789240029200 Leiner, T., & Willemink, M. J. (2021). An international survey
on AI in radiology in 1,041 radiologists and radiology residents
He, L., Li, H., Dudley, J. A., Maloney, T. C., Brady, S. L., part 1: fear of replacement, knowledge, and attitude. European
Somasundaram, E., Trout, A. T., & Dillman, J. R. (2019). Radiology, 31(9), 7058–7066. https://doi.org/10.1007/s00330-
Machine Learning Prediction of Liver Stiffness Using Clinical 021-07781-5
and T2-Weighted MRI Radiomic Data. AJR. American Journal
of Roentgenology, 213(3), 592–601. https://doi.org/10.2214/ Hu, S.-Y., Santus, E., Forsyth, A. W., Malhotra, D., Haimson,
AJR.19.21082 J., Chatterjee, N. A., Kramer, D. B., Barzilay, R., Tulsky,
J. A., & Lindvall, C. (2019). Can machine learning improve
Herent, P., Schmauch, B., Jehanno, P., Dehaene, O., Saillard, patient selection for cardiac resynchronization therapy?
C., Balleyguier, C., Arfi-Rouche, J., & Jégou, S. (2019). PloS One, 14(10), e0222397. https://doi.org/10.1371/journal.
Detection and characterization of MRI breast lesions using pone.0222397
deep learning. Diagnostic and Interventional Imaging, 100(4),
219–225. https://doi.org/10.1016/j.diii.2019.02.008 Hwang, E. J., Nam, J. G., Lim, W. H., Park, S. J., Jeong, Y. S.,
Kang, J. H., Hong, E. K., Kim, T. M., Goo, J. M., Park, S., Kim,
Hinton, B., Ma, L., Mahmoudzadeh, A. P., Malkov, S., Fan, B., K. H., & Park, C. M. (2019). Deep Learning for Chest Radiograph
Greenwood, H., Joe, B., Lee, V., Kerlikowske, K., & Shepherd, Diagnosis in the Emergency Department. Radiology, 293(3),
J. (2019). Deep learning networks find unique mammographic 573–580. https://doi.org/10.1148/radiol.2019191225
differences in previous negative mammograms between interval
and screen-detected cancers: a case-case study. Cancer Imaging: Hwang, E. J., Park, S., Jin, K.-N., Kim, J. I., Choi, S. Y., Lee, J.
The Official Publication of the International Cancer Imaging H., Goo, J. M., Aum, J., Yim, J.-J., Park, C. M., & Deep Learning-
Society, 19(1), 41. https://doi.org/10.1186/s40644-019-0227-3 Based Automatic Detection Algorithm Development and
Evaluation Group. (2019). Development and Validation of
Holzinger, A., Biemann, C., Pattichis, C. S., & Kell, D. B. a Deep Learning-based Automatic Detection Algorithm for
(2017). What do we need to build explainable AI systems for Active Pulmonary Tuberculosis on Chest Radiographs. Clinical
the medical domain? In arXiv [cs.AI]. arXiv. http://arxiv.org/ Infectious Diseases: An Official Publication of the Infectious
abs/1712.09923. Accessed 17/11/2022 Diseases Society of America, 69(5), 739–747. https://doi.
org/10.1093/cid/ciy967
Hötker, A. M., Da Mutten, R., Tiessen, A., Konukoglu, E.,
& Donati, O. F. (2021). Improving workflow in prostate MRI: Hwang, S., Kim, H.-E., Jeong, J., & Kim, H.-J. (2016). A
AI-based decision-making on biparametric or multiparametric novel approach for tuberculosis screening based on deep
MRI. Insights into Imaging, 12(1), 112. https://doi.org/10.1186/ convolutional neural networks. In G. D. Tourassi & S. G. Armato
s13244-021-01058-7 (Eds.), Medical Imaging 2016: Computer-Aided Diagnosis. SPIE.
https://doi.org/10.1117/12.2216198
Huang, S.-C., Kothari, T., Banerjee, I., Chute, C., Ball, R.
L., Borus, N., Huang, A., Patel, B. N., Rajpurkar, P., Irvin, IBM Watson Studio - Model Risk Management. (n.d.).
J., Dunnmon, J., Bledsoe, J., Shpanskaya, K., Dhaliwal, A., Retrieved June 11, 2022, from https://www.ibm.com/cloud/
Zamanian, R., Ng, A. Y., & Lungren, M. P. (2020). PENet-a watson-studio/model-risk-management
scalable deep-learning model for automated diagnosis of
pulmonary embolism using volumetric CT imaging. NPJ Digital Imaging AI Marketplace - overview. (n.d.). Retrieved June
Medicine, 3, 61. https://doi.org/10.1038/s41746-020-0266-y 11, 2022, from https://www.ibm.com/products/imaging-ai-
marketplace
Huang, S.-C., Pareek, A., Seyyedi, S., Banerjee, I., & Lungren,
M. P. (2020). Fusion of medical imaging and electronic
health records using deep learning: a systematic review and
implementation guidelines. NPJ Digital Medicine, 3, 136. https://
doi.org/10.1038/s41746-020-00341-z
25
Jamaludin, A., Lootus, M., Kadir, T., Zisserman, A., Urban, J., Khan, F. A., Majidulla, A., Tavaziva, G., Nazish, A., Abidi, S.
Battié, M. C., Fairbank, J., McCall, I., & Genodisc Consortium. K., Benedetti, A., Menzies, D., Johnston, J. C., Khan, A. J.,
(2017). ISSLS PRIZE IN BIOENGINEERING SCIENCE 2017: & Saeed, S. (2020). Chest x-ray analysis with deep learning-
Automation of reading of radiological features from magnetic based software as a triage test for pulmonary tuberculosis: a
resonance images (MRIs) of the lumbar spine without human prospective study of diagnostic accuracy for culture-confirmed
intervention is comparable with an expert radiologist. European disease. The Lancet. Digital Health, 2(11), e573–e581. https://doi.
Spine Journal: Official Publication of the European Spine Society, org/10.1016/S2589-7500(20)30221-1
the European Spinal Deformity Society, and the European Section
of the Cervical Spine Research Society, 26(5), 1374–1383. https:// Kim, D. W., Jang, H. Y., Kim, K. W., Shin, Y., & Park, S.
doi.org/10.1007/s00586-017-4956-3 H. (2019). Design Characteristics of Studies Reporting the
Performance of Artificial Intelligence Algorithms for Diagnostic
Kaissis, G. A., Makowski, M. R., Rückert, D., & Braren, R. Analysis of Medical Images: Results from Recently Published
F. (2020). Secure, privacy-preserving and federated machine Papers. Korean Journal of Radiology: Official Journal of the Korean
learning in medical imaging. Nature Machine Intelligence, 2(6), Radiological Society, 20(3), 405–410. https://doi.org/10.3348/
305–311. https://doi.org/10.1038/s42256-020-0186-1 kjr.2019.0025
Kaissis, G., Ziller, A., Passerat-Palmbach, J., Ryffel, T., Kim, K. H., & Park, S.-H. (2017). Artificial neural network for
Usynin, D., Trask, A., Lima, I., Mancuso, J., Jungmann, F., suppression of banding artifacts in balanced steady-state free
Steinborn, M.-M., Saleh, A., Makowski, M., Rueckert, D., & precession MRI. Magnetic Resonance Imaging, 37, 139–146.
Braren, R. (2021). End-to-end privacy preserving deep learning https://doi.org/10.1016/j.mri.2016.11.020
on multi-institutional medical imaging. Nature Machine
Intelligence, 3(6), 473–484. https://doi.org/10.1038/s42256- Korteling, J. E. H., van de Boer-Visschedijk, G. C.,
021-00337-8 Blankendaal, R. A. M., Boonekamp, R. C., & Eikelboom,
A. R. (2021). Human- versus Artificial Intelligence. Frontiers
Kalra, A., Chakraborty, A., Fine, B., & Reicher, J. (2020). in Artificial Intelligence 4, 622364. https://doi.org/10.3389/
Machine Learning for Automation of Radiology Protocols for frai.2021.622364
Quality and Efficiency Improvement. Journal of the American
College of Radiology: JACR, 17(9), 1149–1158. https://doi. Kühl, N., Goutier, M., Baier, L., Wolff, C., & Martin, D. (2020).
org/10.1016/j.jacr.2020.03.012 Human vs. supervised machine learning: Who learns patterns
faster? In arXiv [cs.AI] arXiv. http://arxiv.org/abs/2012.03661.
Kao, P.-Y., Chen, J. W., & Manjunath, B. S. (2019). Improving Accessed 17/11/2022
3D U-Net for Brain Tumor Segmentation by Utilizing Lesion
Prior. In arXiv [cs.CV]. arXiv. http://arxiv.org/abs/1907.00281. Kuo, W., Häne, C., Mukherjee, P., Malik, J., & Yuh, E. L. (2019).
Accessed 17/11/2022. Expert-level detection of acute intracranial hemorrhage on head
computed tomography using deep learning. Proceedings of the
Kapoor, N., Lacson, R., & Khorasani, R. (2020). Workflow National Academy of Sciences of the United States of America,
Applications of Artificial Intelligence in Radiology and an 116(45), 22737–22745. https://doi.org/10.1073/
Overview of Available Tools. Journal of the American College of pnas.1908021116
Radiology: JACR, 17(11), 1363–1370. https://doi.org/10.1016/j.
jacr.2020.08.016 Langerhuizen, D. W. G., Janssen, S. J., Mallee, W. H., van den
Bekerom, M. P. J., Ring, D., Kerkhoffs, G. M. M. J., Jaarsma, R.
Kathirvelu, D., Vinupritha, P., & Kalpana, V. (2019). A L., & Doornberg, J. N. (2019). What Are the Applications and
computer aided diagnosis system for measurement of Limitations of Artificial Intelligence for Fracture Detection and
mandibular cortical thickness on dental panoramic radiographs Classification in Orthopaedic Trauma Imaging? A Systematic
in prediction of women with low bone mineral density. Journal of Review. Clinical Orthopaedics and Related Research, 477(11),
Medical Systems, 43(6), 148. https://doi.org/10.1007/ 2482–2491. https://doi.org/10.1097/CORR.0000000000000848
s10916-019-1268-7
Lang, N., Zhang, Y., Zhang, E., Zhang, J., Chow, D., Chang, P.,
Ker, J., Singh, S. P., Bai, Y., Rao, J., Lim, T., & Wang, L. (2019). Yu, H. J., Yuan, H., & Su, M.-Y. (2019). Differentiation of spinal
Image Thresholding Improves 3-Dimensional Convolutional metastases originated from lung and other cancers using
Neural Network Diagnosis of Different Acute Brain Hemorrhages radiomics and deep learning based on DCE-MRI. Magnetic
on Computed Tomography Scans. Sensors, 19(9). https://doi. Resonance Imaging, 64, 4–12. https://doi.org/10.1016/j.
org/10.3390/s19092167 mri.2019.02.013
26
Larrazabal, A. J., Nieto, N., Peterson, V., Milone, D. H., & Liu, F., Zhou, Z., Samsonov, A., Blankenbaker, D., Larison, W.,
Ferrante, E. (2020). Gender imbalance in medical imaging Kanarek, A., Lian, K., Kambhampati, S., & Kijowski, R. (2018).
datasets produces biased classifiers for computer-aided Deep Learning Approach for Evaluating Knee MR Images:
diagnosis. Proceedings of the National Academy of Sciences of Achieving High Diagnostic Performance for Cartilage Lesion
the United States of America, 117(23), 12592–12594. https://doi. Detection. Radiology, 289(1), 160–169. https://doi.org/10.1148/
org/10.1073/pnas.1919012117 radiol.2018172986
Lee, J.-S., Adhikari, S., Liu, L., Jeong, H.-G., Kim, H., & Yoon, Liu, X., Cruz Rivera, S., Moher, D., Calvert, M. J., Denniston,
S.-J. (2019). Osteoporosis detection in panoramic radiographs A. K., & SPIRIT-AI and CONSORT-AI Working Group. (2020).
using a deep convolutional neural network-based computer- Reporting guidelines for clinical trial reports for interventions
assisted diagnosis system: a preliminary study. Dento Maxillo involving artificial intelligence: the CONSORT-AI extension.
Facial Radiology, 48(1), 20170344. https://doi.org/10.1259/ Nature Medicine, 26(9), 1364–1374. https://doi.org/10.1038/
dmfr.20170344 s41591-020-1034-x
Lee, Y. H. (2018). Efficiency Improvement in a Busy Radiology Liu, X., Faes, L., Kale, A. U., Wagner, S. K., Fu, D. J.,
Practice: Determination of Musculoskeletal Magnetic Resonance Bruynseels, A., Mahendiran, T., Moraes, G., Shamdas, M.,
Imaging Protocol Using Deep-Learning Convolutional Neural Kern, C., Ledsam, J. R., Schmid, M. K., Balaskas, K., Topol,
Networks. Journal of Digital Imaging, 31(5), 604–610. https://doi. E. J., Bachmann, L. M., Keane, P. A., & Denniston, A. K.
org/10.1007/s10278-018-0066-y (2019). A comparison of deep learning performance against
health-care professionals in detecting diseases from medical
Leiner, T., Bennink, E., Mol, C. P., Kuijf, H. J., & Veldhuis, W. B. imaging: a systematic review and meta-analysis. The Lancet.
(2021). Bringing AI to the clinic: blueprint for a vendor-neutral Digital Health, 1(6), e271–e297. https://doi.org/10.1016/S2589-
AI deployment infrastructure. Insights into Imaging, 12(1), 11. 7500(19)30123-2
https://doi.org/10.1186/s13244-020-00931-1
Li, X., Shen, L., Xie, X., Huang, S., Xie, Z., Hong, X., & Yu,
Lekadir, K., Osuala, R., Gallin, C., Lazrak, N., Kushibar, J. (2020). Multi-resolution convolutional networks for chest
K., Tsakou, G., Aussó, S., Alberich, L. C., Marias, K., X-ray radiograph based lung nodule detection. Artificial
Tsiknakis, M., Colantonio, S., Papanikolaou, N., Salahuddin, Intelligence in Medicine, 103, 101744. https://doi.org/10.1016/j.
Z., Woodruff, H. C., Lambin, P., & Martí-Bonmatí, L. artmed.2019.101744
(2021). FUTURE-AI: Guiding Principles and Consensus
Recommendations for Trustworthy Artificial Intelligence Lotan, E., Tschider, C., Sodickson, D. K., Caplan, A. L., Bruno,
in Medical Imaging. In arXiv [cs.CV]. arXiv. http://arxiv.org/ M., Zhang, B., & Lui, Y. W. (2020). Medical Imaging and Privacy
abs/2109.09658. Accessed 17/11/2022 in the Era of Artificial Intelligence: Myth, Fallacy, and the
Future. Journal of the American College of Radiology: JACR, 17(9),
Letourneau-Guillon, L., Camirand, D., Guilbert, F., & 1159–1162. https://doi.org/10.1016/j.jacr.2020.04.007
Forghani, R. (2020). Artificial Intelligence Applications for
Workflow, Process Optimization and Predictive Analytics. Maegerlein, C., Fischer, J., Mönch, S., Berndt, M.,
Neuroimaging Clinics of North America, 30(4), e1–e15. https:// Wunderlich, S., Seifert, C. L., Lehm, M., Boeckh-Behrens,
doi.org/10.1016/j.nic.2020.08.008 T., Zimmer, C., & Friedrich, B. (2019). Automated Calculation
of the Alberta Stroke Program Early CT Score: Feasibility and
Levin, D. C., Parker, L., & Rao, V. M. (2017). Recent Trends Reliability. Radiology, 291(1), 141–148. https://doi.org/10.1148/
in Imaging Use in Hospital Settings: Implications for Future radiol.2019181228
Planning. Journal of the American College of Radiology: JACR,
14(3), 331–336. https://doi.org/10.1016/j.jacr.2016.08.025 Mairhöfer, D., Laufer, M., Simon, P. M., Sieren, M., Bischof,
A., Käster, T., Barth, E., Barkhausen, J., & Martinetz,
Lindsey, R., Daluiski, A., Chopra, S., Lachapelle, A., Mozer, T. (2021). An AI-based Framework for Diagnostic Quality
M., Sicular, S., Hanel, D., Gardner, M., Gupta, A., Hotchkiss, Assessment of Ankle Radiographs. https://openreview.net/pdf?
R., & Potter, H. (2018). Deep neural network improves fracture id=bj04hJss_xZ. Accessed 17/11/2022
detection by clinicians. Proceedings of the National Academy of
Sciences of the United States of America, 115(45), 11591–11596. Mancio, J., Pashakhanloo, F., El-Rewaidy, H., Jang, J., Joshi,
https://doi.org/10.1073/pnas.1806905115 G., Csecs, I., Ngo, L., Rowin, E., Manning, W., Maron, M., &
Nezafat, R. (2022). Machine learning phenotyping of scarred
Liu, F., Tang, J., Ma, J., Wang, C., Ha, Q., Yu, Y., & Zhou, Z. myocardium from cine in hypertrophic cardiomyopathy.
(2021). The application of artificial intelligence to chest medical European Heart Journal Cardiovascular Imaging, 23(4), 532–542.
image analysis. Intelligent Medicine, 1(3), 104–117. https://doi. https://doi.org/10.1093/ehjci/jeab056
org/10.1016/j.imed.2021.06.004
27
Matsoukas, S., Morey, J., Lock, G., Chada, D., Shigematsu, Murray, N. M., Unberath, M., Hager, G. D., & Hui, F. K.
T., Marayati, N. F., Delman, B. N., Doshi, A., Majidi, S., (2020). Artificial intelligence to diagnose ischemic stroke and
De Leacy, R., Kellner, C. P., & Fifi, J. T. (2022). AI software identify large vessel occlusions: a systematic review. Journal
detection of large vessel occlusion stroke on CT angiography: of Neurointerventional Surgery, 12(2), 156–164. https://doi.
a real-world prospective diagnostic test accuracy study. org/10.1136/neurintsurg-2019-015135
Journal of Neurointerventional Surgery. https://doi.org/10.1136/
neurintsurg-2021-018391 Nagendran, M., Chen, Y., Lovejoy, C. A., Gordon, A. C.,
Komorowski, M., Harvey, H., Topol, E. J., Ioannidis, J. P. A.,
McKinney, S. M., Sieniek, M., Godbole, V., Godwin, Collins, G. S., & Maruthappu, M. (2020). Artificial intelligence
J., Antropova, N., Ashrafian, H., Back, T., Chesus, M., versus clinicians: systematic review of design, reporting
Corrado, G. S., Darzi, A., Etemadi, M., Garcia-Vicente, F., standards, and claims of deep learning studies. BMJ, 368.
Gilbert, F. J., Halling-Brown, M., Hassabis, D., Jansen, S., https://doi.org/10.1136/bmj.m689
Karthikesalingam, A., Kelly, C. J., King, D., … Shetty, S.
(2020). International evaluation of an AI system for breast Nair, T., Precup, D., Arnold, D. L., & Arbel, T. (2020). Exploring
cancer screening. Nature, 577(7788), 89–94. https://doi. uncertainty measures in deep networks for Multiple sclerosis
org/10.1038/s41586-019-1799-6 lesion detection and segmentation. Medical Image Analysis, 59,
101557. https://doi.org/10.1016/j.media.2019.101557
McLeavy, C. M., Chunara, M. H., Gravell, R. J., Rauf, A.,
Cushnie, A., Staley Talbot, C., & Hawkins, R. M. (2021). The Nakao, T., Hanaoka, S., Nomura, Y., Sato, I., Nemoto, M.,
future of CT: deep learning reconstruction. Clinical Radiology, Miki, S., Maeda, E., Yoshikawa, T., Hayashi, N., & Abe,
76(6), 407–415. https://doi.org/10.1016/j.crad.2021.01.010 O. (2018). Deep neural network-based computer-assisted
detection of cerebral aneurysms in MR angiography. Journal of
Medical AI evaluation. (n.d.). Retrieved June 26, 2022, from Magnetic Resonance Imaging: JMRI, 47(4), 948–953. https://doi.
https://ericwu09.github.io/medical-ai-evaluation/ org/10.1002/jmri.25842
Mlynarski, P., Delingette, H., Criminisi, A., & Ayache, N. Nam, J. G., Kim, M., Park, J., Hwang, E. J., Lee, J. H.,
(2019). Deep learning with mixed supervision for brain tumor Hong, J. H., Goo, J. M., & Park, C. M. (2021). Development
segmentation. Journal of Medical Imaging (Bellingham, Wash.), and validation of a deep learning algorithm detecting
6(3), 034002. https://doi.org/10.1117/1.JMI.6.3.034002 10 common abnormalities on chest radiographs. The
European Respiratory Journal: Official Journal of the European
Mongan, J., Moy, L., & Kahn, C. E., Jr. (2020). Checklist for Society for Clinical Respiratory Physiology, 57(5). https://doi.
Artificial Intelligence in Medical Imaging (CLAIM): A Guide for org/10.1183/13993003.03061-2020
Authors and Reviewers. Radiology. Artificial Intelligence, 2(2),
e200029. https://doi.org/10.1148/ryai.2020200029 Narayana, P. A., Coronado, I., Sujit, S. J., Wolinsky, J. S.,
Lublin, F. D., & Gabr, R. E. (2020). Deep Learning for Predicting
Moon, H., Huo, Y., Abramson, R. G., Peters, R. A., Assad, Enhancing Lesions in Multiple Sclerosis from Noncontrast
A., Moyo, T. K., Savona, M. R., & Landman, B. A. (2019). MRI. Radiology, 294(2), 398–404. https://doi.org/10.1148/
Acceleration of spleen segmentation with end-to-end deep radiol.2019191061
learning method and automated pipeline. Computers in
Biology and Medicine, 107, 109–117. https://doi.org/10.1016/j. National Institute for Health and Care Excellence (NICE).
compbiomed.2019.01.018 (n.d.). Evidence standards framework for digital health
technologies. Retrieved June 10, 2022, from https://www.nice.
Morey, J. R., Zhang, X., Yaeger, K. A., Fiano, E., Marayati, N. org.uk/corporate/ecd7
F., Kellner, C. P., De Leacy, R. A., Doshi, A., Tuhrim, S., & Fifi,
J. T. (2021). Real-World Experience with Artificial Intelligence- Neisius, U., El-Rewaidy, H., Nakamori, S., Rodriguez, J.,
Based Triage in Transferred Large Vessel Occlusion Stroke Manning, W. J., & Nezafat, R. (2019). Radiomic Analysis
Patients. Cerebrovascular Diseases, 50(4), 450–455. https://doi. of Myocardial Native T1 Imaging Discriminates Between
org/10.1159/000515320 Hypertensive Heart Disease and Hypertrophic Cardiomyopathy.
JACC. Cardiovascular Imaging, 12(10), 1946–1954. https://doi.
Murdoch, B. (2021). Privacy and artificial intelligence: org/10.1016/j.jcmg.2018.11.024
challenges for protecting health information in a new era. BMC
Medical Ethics, 22(1), 122. https://doi.org/10.1186/s12910-021- Nelson, A., Herron, D., Rees, G., & Nachev, P. (2019).
00687-3 Predicting scheduled hospital attendance with artificial
intelligence. Npj Digital Medicine, 2(1), 26. https://doi.
org/10.1038/s41746-019-0103-3
28
Nielsen, A., Hansen, M. B., Tietze, A., & Mouridsen, K. (2018). Ooi, S. K. G., Makmur, A., Soon, A. Y. Q., Fook-Chong, S.,
Prediction of Tissue Outcome and Assessment of Treatment Liew, C., Sia, S. Y., Ting, Y. H., & Lim, C. Y. (2021). Attitudes
Effect in Acute Ischemic Stroke Using Deep Learning. Stroke; a toward artificial intelligence in radiology with learner needs
Journal of Cerebral Circulation, STROKEAHA.117.019740. https:// assessment within radiology residency programmes: a national
doi.org/10.1161/STROKEAHA.117.019740 multi-programme survey. Singapore Medical Journal, 62(3),
126–134. https://doi.org/10.11622/smedj.2019141
Norori, N., Hu, Q., Aellen, F. M., Faraci, F. D., & Tzovara, A.
(2021). Addressing bias in big data and AI for health care: A call Pan, Y., Shi, D., Wang, H., Chen, T., Cui, D., Cheng, X., & Lu, Y.
for open science. Patterns (New York, N.Y.), 2(10), 100347. (2020). Automatic opportunistic osteoporosis screening using
https://doi.org/10.1016/j.patter.2021.100347. Accessed low-dose chest computed tomography scans obtained for lung
17/11/2022 cancer screening. European Radiology, 30(7), 4107–4116. https://
doi.org/10.1007/s00330-020-06679-y
O’Connor, S. D., & Bhalla, M. (2021). Should Artificial
Intelligence Tell Radiologists Which Study to Read Next? [Review Park, H. J., Kim, S. M., La Yun, B., Jang, M., Kim, B., Jang, J.
of Should Artificial Intelligence Tell Radiologists Which Study to Y., Lee, J. Y., & Lee, S. H. (2019). A computer-aided diagnosis
Read Next?]. Radiology. Artificial Intelligence, 3(2), e210009. system using artificial intelligence for the diagnosis and
https://doi.org/10.1148/ryai.2021210009 characterization of breast masses on ultrasound: Added
value for the inexperienced breast radiologist. Medicine, 98(3),
Office for Civil Rights (OCR). (2012, September 7). Guidance e14146. https://doi.org/10.1097/MD.0000000000014146
Regarding Methods for De-identification of Protected Health
Information in Accordance with the Health Insurance Portability Price, I. I., & Nicholson, W. (2019). Medical AI and Contextual
and Accountability Act (HIPAA) Privacy Rule. HHS.gov; US Bias. https://papers.ssrn.com/abstract=3347890
Department of Health and Human Services. https://www.
hhs.gov/hipaa/for-professionals/privacy/special-topics/de- Puvanasunthararajah, S., Fontanarosa, D., Wille, M.-L., &
identification/index.html. Accessed 17/11/2022 Camps, S. M. (2021). The application of metal artifact reduction
methods on computed tomography scans for radiotherapy
Oktay, O., Schlemper, J., Le Folgoc, L., Lee, M., Heinrich, applications: A literature review. Journal of Applied Clinical
M., Misawa, K., Mori, K., McDonagh, S., Hammerla, N. Y., Medical Physics / American College of Medical Physics, 22(6),
Kainz, B., Glocker, B., & Rueckert, D. (2018). Attention U-Net: 198–223. https://doi.org/10.1002/acm2.13255
Learning Where to Look for the Pancreas. In arXiv [cs.CV]. arXiv.
http://arxiv.org/abs/1804.03999. Accessed 17/11/2022 Qin, Z. Z., Sander, M. S., Rai, B., Titahong, C. N., Sudrungrot,
S., Laah, S. N., Adhikari, L. M., Carter, E. J., Puri, L., Codlin,
Olczak, J., Fahlberg, N., Maki, A., Razavian, A. S., Jilert, A., A. J., & Creswell, J. (2019). Using artificial intelligence to
Stark, A., Sköldenberg, O., & Gordon, M. (2017). Artificial read chest radiographs for tuberculosis detection: A multi-site
intelligence for analyzing orthopedic trauma radiographs. Acta evaluation of the diagnostic accuracy of three deep learning
Orthopaedica, 88(6), 581–586. https://doi.org/10.1080/1745367 systems. Scientific Reports, 9(1), 15000. https://doi.org/10.1038/
4.2017.1344459 s41598-019-51503-3
Olthof, A. W., van Ooijen, P. M. A., & Rezazade Mehrizi, M. H. Ramspek, C. L., Jager, K. J., Dekker, F. W., Zoccali, C., & van
(2020). Promises of artificial intelligence in neuroradiology: a Diepen, M. (2021). External validation of prognostic models:
systematic technographic review. Neuroradiology, 62(10), 1265– what, why, how, when and where? Clin Kidney J. 2020 Nov
1278. https://doi.org/10.1007/s00234-020-02424-w 24;14(1):49-58.
Omoumi, P., Ducarouge, A., Tournier, A., Harvey, H., Kahn, C. Rao, B., Zohrabian, V., Cedeno, P., Saha, A., Pahade, J., &
E., Jr, Louvet-de Verchère, F., Pinto Dos Santos, D., Kober, T., Davis, M. A. (2021). Utility of Artificial Intelligence Tool as a
& Richiardi, J. (2021). To buy or not to buy-evaluating Prospective Radiology Peer Reviewer - Detection of Unreported
commercial AI solutions in radiology (the ECLAIR guidelines). Intracranial Hemorrhage. Academic Radiology, 28(1), 85–93.
European Radiology, 31(6), 3786–3796. https://doi.org/10.1007/ https://doi.org/10.1016/j.acra.2020.01.035
s00330-020-07684-x
Reddy, S., Allan, S., Coghlan, S., & Cooper, P. (2020). A
O’Neill, T. J., Xi, Y., Stehel, E., Browning, T., Ng, Y. S., Baker, governance model for the application of AI in health care.
C., & Peshock, R. M. (2021). Active Reprioritization of the Journal of the American Medical Informatics Association: JAMIA,
Reading Worklist Using Artificial Intelligence Has a Beneficial 27(3), 491–497. https://doi.org/10.1093/jamia/ocz192
Effect on the Turnaround Time for Interpretation of Head CT
with Intracranial Hemorrhage. Radiology. Artificial Intelligence,
3(2), e200024. https://doi.org/10.1148/ryai.2020200024
29
Reddy, S., Rogers, W., Makinen, V.-P., Coiera, E., Brown, Schemmel, A., Lee, M., Hanley, T., Pooler, B. D., Kennedy,
P., Wenzel, M., Weicken, E., Ansari, S., Mathur, P., Casey, T., Field, A., Wiegmann, D., & Yu, J.-P. J. (2016). Radiology
A., & Kelly, B. (2021). Evaluation framework to guide Workflow Disruptors: A Detailed Analysis. Journal of the
implementation of AI systems into healthcare settings. BMJ American College of Radiology: JACR, 13(10), 1210–1214. https://
Health & Care Informatics, 28(1). https://doi.org/10.1136/ doi.org/10.1016/j.jacr.2016.04.009
bmjhci-2021-100444
Schreiber-Zinaman, J., & Rosenkrantz, A. B. (2017). Frequency
Reyes, M., Meier, R., Pereira, S., Silva, C. A., Dahlweid, F.-M., and reasons for extra sequences in clinical abdominal MRI
von Tengg-Kobligk, H., Summers, R. M., & Wiest, R. (2020). examinations. Abdominal Radiology (New York), 42(1), 306–311.
On the Interpretability of Artificial Intelligence in Radiology: https://doi.org/10.1007/s00261-016-0877-6
Challenges and Opportunities. Radiology. Artificial Intelligence,
2(3), e190043. https://doi.org/10.1148/ryai.2020190043 Scott, I. A., Carter, S. M., & Coiera, E. (2021). Exploring
stakeholder attitudes towards AI in clinical practice. BMJ
Rezazade Mehrizi, M. H., van Ooijen, P., & Homan, M. (2021). Health & Care Informatics, 28(1). https://doi.org/10.1136/
Applications of artificial intelligence (AI) in diagnostic radiology: bmjhci-2021-100450
a technography study. European Radiology, 31(4), 1805–1811.
https://doi.org/10.1007/s00330-020-07230-9 Seah, J. C. Y., Tang, C. H. M., Buchlak, Q. D., Holt, X. G.,
Wardman, J. B., Aimoldin, A., Esmaili, N., Ahmad, H., Pham,
Richardson, J. P., Smith, C., Curtis, S., Watson, S., Zhu, H., Lambert, J. F., Hachey, B., Hogg, S. J. F., Johnston, B. P.,
X., Barry, B., & Sharp, R. R. (2021). Patient apprehensions Bennett, C., Oakden-Rayner, L., Brotchie, P., & Jones, C. M.
about the use of artificial intelligence in healthcare. NPJ Digital (2021). Effect of a comprehensive deep-learning model on
Medicine, 4(1), 140. https://doi.org/10.1038/s41746-021-00509-1 the accuracy of chest x-ray interpretation by radiologists: a
retrospective, multireader multicase study. The Lancet. Digital
Richardson, M. L., Garwood, E. R., Lee, Y., Li, M. D., Lo, H. Health, 3(8), e496–e506. https://doi.org/10.1016/S2589-
S., Nagaraju, A., Nguyen, X. V., Probyn, L., Rajiah, P., Sin, J., 7500(21)00106-0
Wasnik, A. P., & Xu, K. (2021). Noninterpretive Uses of Artificial
Intelligence in Radiology. Academic Radiology, 28(9), 1225– Sectra Amplifier Marketplace. (2021, July 5). Sectra
1235. https://doi.org/10.1016/j.acra.2020.01.012 Medical. https://medical.sectra.com/product/sectra-amplifier-
marketplace/
Rockenbach, M. A. B. (2021, June 13). Multimodal AI in
healthcare: Closing the gaps. CodeX. https://medium.com/codex/ Sermesant, M., Delingette, H., Cochet, H., Jaïs, P., & Ayache,
multimodal-ai-in-healthcare-1f5152e83be2 N. (2021). Applications of artificial intelligence in cardiovascular
imaging. Nature Reviews. Cardiology, 18(8), 600–609. https://doi.
Rodríguez-Ruiz, A., Krupinski, E., Mordang, J.-J., Schilling, K., org/10.1038/s41569-021-00527-2
Heywang-Köbrunner, S. H., Sechopoulos, I., & Mann, R. M.
(2019). Detection of Breast Cancer with Mammography: Effect Setio, A. A. A., Traverso, A., de Bel, T., Berens, M. S. N., van
of an Artificial Intelligence Support System. Radiology, 290(2), den Bogaard, C., Cerello, P., Chen, H., Dou, Q., Fantacci, M.
305–314. https://doi.org/10.1148/radiol.2018181371 E., Geurts, B., Gugten, R. van der, Heng, P. A., Jansen, B.,
de Kaste, M. M. J., Kotov, V., Lin, J. Y.-H., Manders, J. T. M.
Rodriguez-Ruiz, A., Lång, K., Gubern-Merida, A., Broeders, C., Sóñora-Mengana, A., García-Naranjo, J. C., … Jacobs, C.
M., Gennaro, G., Clauser, P., Helbich, T. H., Chevalier, (2017). Validation, comparison, and combination of algorithms
M., Tan, T., Mertelmeier, T., Wallis, M. G., Andersson, I., for automatic detection of pulmonary nodules in computed
Zackrisson, S., Mann, R. M., & Sechopoulos, I. (2019). Stand- tomography images: The LUNA16 challenge. Medical Image
Alone Artificial Intelligence for Breast Cancer Detection in Analysis, 42, 1–13. https://doi.org/10.1016/j.media.2017.06.015
Mammography: Comparison With 101 Radiologists. Journal
of the National Cancer Institute, 111(9), 916–922. https://doi. Seyyed-Kalantari, L., Zhang, H., McDermott, M. B. A., Chen,
org/10.1093/jnci/djy222 I. Y., & Ghassemi, M. (2021). Underdiagnosis bias of artificial
intelligence algorithms applied to chest radiographs in under-
Santomartino, S. M., & Yi, P. H. (2022). Systematic Review served patient populations. Nature Medicine, 27(12), 2176–
of Radiologist and Medical Student Attitudes on the Role and 2182. https://doi.org/10.1038/s41591-021-01595-0
Impact of AI in Radiology. Academic Radiology. https://doi.
org/10.1016/j.acra.2021.12.032
30
Shan, H., Padole, A., Homayounieh, F., Kruger, U., Khera, Tamada, D., Kromrey, M.-L., Ichikawa, S., Onishi, H., &
R. D., Nitiwarangkul, C., Kalra, M. K., & Wang, G. (2019). Motosugi, U. (2020). Motion Artifact Reduction Using a
Competitive performance of a modularized deep neural network Convolutional Neural Network for Dynamic Contrast Enhanced
compared to commercial algorithms for low-dose CT image MR Imaging of the Liver. Magnetic Resonance in Medical
reconstruction. Nature Machine Intelligence, 1(6), 269–276. Sciences: MRMS: An Official Journal of Japan Society of Magnetic
https://doi.org/10.1038/s42256-019-0057-9 Resonance in Medicine, 19(1), 64–76. https://doi.org/10.2463/
mrms.mp.2018-0156
Sharma, K., Rupprecht, C., Caroli, A., Aparicio, M. C.,
Remuzzi, A., Baust, M., & Navab, N. (2017). Automatic The Medical Futurist. (n.d.). The Medical Futurist. Retrieved
Segmentation of Kidneys using Deep Learning for Total Kidney February 23, 2022, from https://medicalfuturist.com/fda-
Volume Quantification in Autosomal Dominant Polycystic approved-ai-based-algorithms/
Kidney Disease. Scientific Reports, 7(1), 2049. https://doi.
org/10.1038/s41598-017-01779-0 The Nuance AI Marketplace for Diagnostic Imaging. (n.d.).
https://www.nuance.com/content/dam/nuance/en_us/collateral/
Shelmerdine, S. C., Arthurs, O. J., Denniston, A., & Sebire, healthcare/data-sheet/ds-ai-marketplace-for-diagnostic-
N. J. (2021). Review of study reporting guidelines for clinical imaging-en-us.pdf. Accessed 17/11/2022
studies using artificial intelligence in healthcare. BMJ
Health & Care Informatics, 28(1). https://doi.org/10.1136/ Thodberg, H. H., Kreiborg, S., Juul, A., & Pedersen, K. D.
bmjhci-2021-100385 (2009). The BoneXpert method for automated determination of
skeletal maturity. IEEE Transactions on Medical Imaging, 28(1),
Shinagare, A. B., Ip, I. K., Abbett, S. K., Hanson, R., Seltzer, S. 52–66. https://doi.org/10.1109/TMI.2008.926067
E., & Khorasani, R. (2014). Inpatient imaging utilization: trends
of the past decade. AJR. American Journal of Roentgenology, Thomas, K. A., Kidziński, Ł., Halilaj, E., Fleming, S. L.,
202(3), W277–W283. https://doi.org/10.2214/AJR.13.10986 Venkataraman, G. R., Oei, E. H. G., Gold, G. E., & Delp, S.
L. (2020). Automated Classification of Radiographic Knee
Shlobin, N. A., Baig, A. A., Waqas, M., Patel, T. R., Dossani, Osteoarthritis Severity Using Deep Neural Networks. Radiology.
R. H., Wilson, M., Cappuzzo, J. M., Siddiqui, A. H., Tutino, V. Artificial Intelligence, 2(2), e190065. https://doi.org/10.1148/
M., & Levy, E. I. (2022). Artificial Intelligence for Large-Vessel ryai.2020190065
Occlusion Stroke: A Systematic Review. World Neurosurgery, 159,
207–220.e1. https://doi.org/10.1016/j.wneu.2021.12.004 Towards trustable machine learning. (2018). Nature
Biomedical Engineering, 2(10), 709–710. https://doi.org/10.1038/
Silberg, J., & Manyika, J. (2019, June 6). Tackling bias in s41551-018-0315-x
artificial intelligence (and in humans). McKinsey & Company.
https://www.mckinsey.com/featured-insights/artificial- Trinidad, M. G., Platt, J., & Kardia, S. L. R. (2020). The public’s
intelligence/tackling-bias-in-artificial-intelligence-and-in- comfort with sharing health data with third-party commercial
humans companies. Humanities and Social Sciences Communications,
7(1), 1–10. https://doi.org/10.1057/s41599-020-00641-5
Singh, S., Kalra, M. K., Hsieh, J., Licato, P. E., Do, S., Pien,
H. H., & Blake, M. A. (2010). Abdominal CT: comparison Trivedi, H., Mesterhazy, J., Laguna, B., Vu, T., & Sohn, J. H.
of adaptive statistical iterative and filtered back projection (2018). Automatic Determination of the Need for Intravenous
reconstruction techniques. Radiology, 257(2), 373–383. https:// Contrast in Musculoskeletal MRI Examinations Using IBM
doi.org/10.1148/radiol.10092212 Watson’s Natural Language Processing Algorithm. Journal
of Digital Imaging, 31(2), 245–251. https://doi.org/10.1007/
Smith-Bindman, R., Kwan, M. L., Marlow, E. C., Theis, M. s10278-017-0021-3
K., Bolch, W., Cheng, S. Y., Bowles, E. J. A., Duncan, J. R.,
Greenlee, R. T., Kushi, L. H., Pole, J. D., Rahm, A. K., Stout, Tsao, D. N. (2020, July 27). AI in medical diagnostics 2020-
N. K., Weinmann, S., & Miglioretti, D. L. (2019). Trends in 2030: Image recognition, players, clinical applications, forecasts:
Use of Medical Imaging in US Health Care Systems and in IDTechEx. https://www.idtechex.com/en/research-report/ai-in-
Ontario, Canada, 2000-2016. JAMA: The Journal of the American medical-diagnostics-2020-2030-image-recognition-players-
Medical Association, 322(9), 843–856. https://doi.org/10.1001/ clinical-applications-forecasts/766
jama.2019.11456
Tucci, V., Saary, J., & Doyle, T. E. (2022). Factors influencing
Sutherland, G., Russell, N., Gibbard, R., & Dobrescu, A. trust in medical artificial intelligence for healthcare
(n.d.). The value of radiology, part II. https://car.ca/wp-content/ professionals: a narrative review. Journal of Medical Artificial
uploads/2019/07/value-of-radiology-part-2-en.pdf. Accessed Intelligence, 5, 4–4. https://doi.org/10.21037/jmai-21-25
17/11/2022
31
Ueda, D., Yamamoto, A., Nishimori, M., Shimono, T., Whittlestone, J., Nyrup, R., Alexandrova, A., Dihal, K., &
Doishita, S., Shimazaki, A., Katayama, Y., Fukumoto, S., Cave, S. (2019). Ethical and societal implications of algorithms,
Choppin, A., Shimahara, Y., & Miki, Y. (2019). Deep Learning data, and artificial intelligence: a roadmap for research.
for MR Angiography: Automated Detection of Cerebral London: Nuffield Foundation. https://www.nuffieldfoundation.
Aneurysms. Radiology, 290(1), 187–194. https://doi.org/10.1148/ org/sites/default/files/files/Ethical-and-Societal-Implications-of-
radiol.2018180901 Data-and-AI-report-Nuffield-Foundat.pdf. Accessed 17/11/2022
Urakawa, T., Tanaka, Y., Goto, S., Matsuzawa, H., Watanabe, WHO operational handbook on tuberculosis Module 2:
K., & Endo, N. (2019). Detecting intertrochanteric hip fractures Screening – Systematic screening for tuberculosis disease.
with orthopedist-level accuracy using a deep convolutional (n.d.). Retrieved June 19, 2022, from https://www.who.int/
neural network. Skeletal Radiology, 48(2), 239–244. https://doi. publications-detail-redirect/9789240022614
org/10.1007/s00256-018-3016-3
Willemink, M. J., & Noël, P. B. (2019). The evolution of image
van Leeuwen, K. G., Schalekamp, S., Rutten, M. J. C. M., reconstruction for CT-from filtered back projection to artificial
van Ginneken, B., & de Rooij, M. (2021). Artificial intelligence intelligence. European Radiology, 29(5), 2185–2195. https://doi.
in radiology: 100 commercially available products and their org/10.1007/s00330-018-5810-7
scientific evidence. European Radiology, 31(6), 3797–3804.
https://doi.org/10.1007/s00330-021-07892-z Winder, M., Owczarek, A. J., Chudek, J., Pilch-Kowalczyk, J.,
& Baron, J. (2021). Are We Overdoing It? Changes in Diagnostic
Vayena, E., & Blasimme, A. (2017). Biomedical Big Data: New Imaging Workload during the Years 2010-2020 including
Models of Control Over Access, Use and Governance. Journal the Impact of the SARS-CoV-2 Pandemic. Healthcare (Basel,
of Bioethical Inquiry, 14(4), 501–513. https://doi.org/10.1007/ Switzerland), 9(11). https://doi.org/10.3390/healthcare9111557
s11673-017-9809-6
Wong, T. T., Kazam, J. K., & Rasiej, M. J. (2019). Effect
Vayena, E., Blasimme, A., & Cohen, I. G. (2018). Machine of Analytics-Driven Worklists on Musculoskeletal MRI
learning in medicine: Addressing ethical challenges. PLoS Interpretation Times in an Academic Setting. AJR. American
Medicine, 15(11), e1002689. https://doi.org/10.1371/journal. Journal of Roentgenology, 1–5. https://doi.org/10.2214/
pmed.1002689 AJR.18.20434
Wang, J., Yang, F., Liu, W., Sun, J., Han, Y., Li, D., Gkoutos, G. Wu, B., Zhou, Z., Wang, J., & Wang, Y. (2018). Joint learning for
V., Zhu, Y., & Chen, Y. (2020). Radiomic Analysis of Native T1 pulmonary nodule segmentation, attributes and malignancy
Mapping Images Discriminates Between MYH7 and MYBPC3- prediction. 2018 IEEE 15th International Symposium on
Related Hypertrophic Cardiomyopathy. Journal of Magnetic Biomedical Imaging (ISBI 2018), 1109–1113. https://doi.
Resonance Imaging: JMRI, 52(6), 1714–1721. https://doi. org/10.1109/ISBI.2018.8363765
org/10.1002/jmri.27209
Wu, E., Wu, K., Daneshjou, R., Ouyang, D., Ho, D. E., & Zou, J.
Wang, S.-H., Tang, C., Sun, J., Yang, J., Huang, C., Phillips, (2021). How medical AI devices are evaluated: limitations and
P., & Zhang, Y.-D. (2018). Multiple Sclerosis Identification recommendations from an analysis of FDA approvals. Nature
by 14-Layer Convolutional Neural Network With Batch Medicine, 27(4), 582–584. https://doi.org/10.1038/s41591-021-
Normalization, Dropout, and Stochastic Pooling. Frontiers in 01312-x
Neuroscience, 12, 818. https://doi.org/10.3389/fnins.2018.00818
Wu, G.-G., Zhou, L.-Q., Xu, J.-W., Wang, J.-Y., Wei, Q., Deng,
Watanabe, A. T., Lim, V., Vu, H. X., Chim, R., Weise, E., Y.-B., Cui, X.-W., & Dietrich, C. F. (2019). Artificial intelligence
Liu, J., Bradley, W. G., & Comstock, C. E. (2019). Improved in breast ultrasound. World Journal of Radiology, 11(2), 19–26.
Cancer Detection Using Artificial Intelligence: a Retrospective https://doi.org/10.4329/wjr.v11.i2.19
Evaluation of Missed Cancers on Mammography. Journal
of Digital Imaging, 32(4), 625–637. https://doi.org/10.1007/ Yala, A., Schuster, T., Miles, R., Barzilay, R., & Lehman,
s10278-019-00192-5 C. (2019). A Deep Learning Model to Triage Screening
Mammograms: A Simulation Study. Radiology, 293(1), 38–46.
Weikert, T., Francone, M., Abbara, S., Baessler, B., Choi, B. https://doi.org/10.1148/radiol.2019182908
W., Gutberlet, M., Hecht, E. M., Loewe, C., Mousseaux, E.,
Natale, L., Nikolaou, K., Ordovas, K. G., Peebles, C., Prieto, Yasaka, K., Akai, H., Kunimatsu, A., Abe, O., & Kiryu, S.
C., Salgado, R., Velthuis, B., Vliegenthart, R., Bremerich, J., & (2018). Liver Fibrosis: Deep Convolutional Neural Network for
Leiner, T. (2021). Machine learning in cardiovascular radiology: Staging by Using Gadoxetic Acid-enhanced Hepatobiliary Phase
ESCR position statement on design requirements, quality MR Images. Radiology, 287(1), 146–155. https://doi.org/10.1148/
assessment, current applications, opportunities, and challenges. radiol.2017171928
European Radiology, 31(6), 3909–3922. https://doi.org/10.1007/
s00330-020-07417-0
32
Yeung, K. (2018). A Study of the Implications of Advanced Zhao, J., Huang, Y., Song, Y., Xie, D., Hu, M., Qiu, H., & Chu,
Digital Technologies (Including AI Systems) for the Concept of J. (2020). Diagnostic accuracy and potential covariates for
Responsibility Within a Human Rights Framework. https://papers. machine learning to identify IDH mutations in glioma patients:
ssrn.com/abstract=3286027 evidence from a meta-analysis. European Radiology, 30(8),
4664–4674. https://doi.org/10.1007/s00330-020-06717-9
Yoo, Y., Tang, L. Y. W., Li, D. K. B., Metz, L., Kolind, S.,
Traboulsee, A. L., & Tam, R. C. (2019). Deep learning of brain Zhou, C., Ding, C., Wang, X., Lu, Z., & Tao, D. (2020). One-pass
lesion patterns and user-defined clinical and MRI features Multi-task Networks with Cross-task Guided Attention for Brain
for predicting conversion to multiple sclerosis from clinically Tumor Segmentation. IEEE Transactions on Image Processing:
isolated syndrome. Computer Methods in Biomechanics and A Publication of the IEEE Signal Processing Society. https://doi.
Biomedical Engineering: Imaging & Visualization, 7(3), 250–259. org/10.1109/TIP.2020.2973510
https://doi.org/10.1080/21681163.2017.1356750
Zicari, R. V., Brodersen, J., Brusseau, J., Düdder, B., Eichhorn,
Yu, A. C., Mohajer, B., & Eng, J. (2022). External Validation T., Ivanov, T., Kararigas, G., Kringen, P., McCullough, M.,
of Deep Learning Algorithms for Radiologic Diagnosis: A Möslein, F., Mushtaq, N., Roig, G., Stürtz, N., Tolle, K., Tithi,
Systematic Review. Radiology. Artificial Intelligence, 4(3), J. J., van Halem, I., & Westerlund, M. (2021). Z-Inspection®: A
e210064. https://doi.org/10.1148/ryai.210064 Process to Assess Trustworthy AI. IEEE Transactions on Technology
and Society, 1–1. https://doi.org/10.1109/TTS.2021.3066209
Yu, J.-P. J., Kansagra, A. P., & Mongan, J. (2014). The
radiologist’s workflow environment: evaluation of disruptors
and potential implications. Journal of the American College
of Radiology: JACR, 11(6), 589–593. https://doi.org/10.1016/j.
jacr.2013.12.026
Yusuf, M., Atal, I., Li, J., Smith, P., Ravaud, P., Fergie, M.,
Callaghan, M., & Selfe, J. (2020). Reporting quality of studies
using machine learning models for medical diagnosis: a
systematic review. BMJ Open, 10(3), e034568. https://doi.
org/10.1136/bmjopen-2019-034568
Yu, Y., Xie, Y., Thamm, T., Gong, E., Ouyang, J., Christensen,
S., Marks, M. P., Lansberg, M. G., Albers, G. W., & Zaharchuk,
G. (2021). Tissue at Risk and Ischemic Core Estimation Using
Deep Learning in Acute Stroke. AJNR. American Journal of
Neuroradiology, 42(6), 1030–1037. https://doi.org/10.3174/ajnr.
A7081
Yu, Y., Xie, Y., Thamm, T., Gong, E., Ouyang, J., Huang, C.,
Christensen, S., Marks, M. P., Lansberg, M. G., Albers, G.
W., & Zaharchuk, G. (2020). Use of Deep Learning to Predict
Final Ischemic Stroke Lesions From Initial Magnetic Resonance
Imaging. JAMA Network Open, 3(3), e200772. https://doi.
org/10.1001/jamanetworkopen.2020.0772
Zhang, Y., & Yu, H. (2018). Convolutional Neural Network Based

Metal Artifact Reduction in X-Ray Computed Tomography. IEEE
Transactions on Medical Imaging, 37(6), 1370–1381. https://doi.
org/10.1109/TMI.2018.2823083
Zhao, B., Liu, Z., Ding, S., Liu, G., Cao, C., & Wu, H.
(2022). Motion artifact correction for MR images based on
convolutional neural network. Optoelectronics Letters, 18(1),
54–58. https://doi.org/10.1007/s11801-022-1084-z
33
Bayer, the Bayer Cross, and Calantic are trademarks owned by and/or registered to Bayer in the U.S. and/or other countries.
Other trademarks and company names mentioned herein are properties of their respective owners and are used herein solely for
informational purposes. No relationship or endorsement should be inferred or implied.
This is intended to provide information for healthcare professionals and other relevant decision makers in the US, European Union,
United Kingdom and Switzerland. For information specific to your country please contact your local Bayer representative.
You are encouraged to report adverse events; reporting information can be found on https://www.bayer.com/en/pharma/report-side-
effect
© 2022 Bayer. This material may not be reproduced, displayed, modified or distributed without the express prior written consent of
Bayer.
PP-CALA-DE-0084-1 November 2022
PP-CALA-ALL-0233-1 November 2022
PP-CALA-ES-0010-1 November 2022
PP-CALA-AT-0015-1 December 2022

PP-CALA-GB-0109 November 2022
Bayer AG
13353 Berlin, Germany
More information on
www.calantic.com

The Complete Guide To Artificial Intelligence in Radiology (2022)

Uploaded by

Copyright:

Available Formats

The Complete Guide To Artificial Intelligence in Radiology (2022)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Complete Guide To Artificial Intelligence in Radiology (2022)

Uploaded by

Copyright:

Available Formats

The Complete Guide to

The case for artificial intelligence in radiology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04

Fundamentals of artificial intelligence....... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 05

Clinical uses.. ........................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 08

Obstacles to implementation. .................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Making purchasing decisions...................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Artificial intelligence (AI) is playing a growing role in all

This book covers how and why AI can address

The case for artificial intelligence

Unsupervised learning In regression, the most commonly used metrics include:

characteristic curve (AUC) is a threshold-independent Guidelines for evaluating AI research

Internal and external validity

Externally valid models perform well in their tasks on

Data quantity, quality and variety IT infrastructure

Making purchasing decisions

The main selling points of many AI-based solutions

future. The potential for these algorithms to learn from

While the obstacles discussed in previous sections

AI has shown promise in positively impacting

Zhang, Y., & Yu, H. (2018). Convolutional Neural Network Based

PP-CALA-ES-0010-1 November 2022

PP-CALA-AT-0015-1 December 2022

You might also like