J

o
u Inclusi
Discipline/Are Yea Esclusion/Issu Imag N. Algorith Type of Secondary
Ref. No. Author r Title AI Area one MainAim Specific Imaging Code Algorithm
a r es es Pts. m Imaging Aim
n (Y/N)
a
l
E
u
r
o
p
e
a
n
Superior skin
J Articl
cancer
o e
classification Artifici
u refers
Achim Dermatology/ 201 by the al 1144 Classifica Dermosco Improveac
(47) r Y only CNN Dermoscopic ResNet50
Hekler Oncology 9 combination Intellig 4 tion pic curacy
n to
of human ence
a image
and artificial
l s
intelligence
o
f
C
a
n
c
e
r
E
u
r
o Application
p of a deep
e learning Articl
Deep
a algorithm e
Convolut
n for detection Deep refers
Chi-Tung 201 2550 ional Detectio Radiograp FrontalPelvicradiograp High
(68) R Radiology and Learnin Y only TensorFlow 1.5.1
Cheng 9 5 Neural n hs hs (PXRs) Accuracy
a visualization g to
Network
d of hip image
(DCNN)
i fractures on s
o plain pelvic
l radiographs
o
g
y
 Predicting
cancer
Survival
outcomes
Convolut
P from
Deep ional Digital
PooyaMo N 201 histology Predicto Digital Pathology RemarkeSi
(23) Oncology Learnin Y 1061 769 Neural Pathology TensorFlow 0.12.0
badersany A 8 and utcomes images milarity
g Network images
S genomics
s
using
(SCNNs)
convolutiona
l networks
Diagnostic
Assessment
of Deep
Whether this
Learning Deep
approach has
J Algorithms Convolut
BabakEht Deep clinical utility Digital
a 201 for ional Detectio Whole-Slide Pathology Improve
(53) eshamiBej Oncology Learnin Y will require 399 Pathology GoogLeNet
m 7 Detection of Neural n images accuracy
nordi g evaluation in images
a Lymph Node Network
a clinical
Metastases (DCNN)
setting.
in Women
With Breast
Cancer
Applying
artificial
P intelligence
Deep
L to disease Not inherent
Endocrinology Convolut
O staging: Deep in Predict Fundus
Hidenori / 201 ional Improve
(58) S Deep Learnin Y radiotherapy, 9939 2740 outcome photograp Fundus photography GoogLeNet
Takahashi Ophthalmolog 7 Neural accuracy
O learning for g but still s hy
y Network
N improved worthy
(DCNN)
E staging of
diabetic
retinopathy
J
o
u
r
An
n
Algorithm Articl
a
Based on Not inherent e Recurre
l
Joon- Deep Deep in refers nt DEWS (Deep learning
o 201 5213 Detectio Improve
(42) MyoungK Cardiology Learning for Learnin Y radiotherapy, only Neural based Early Warning
f 8 1 n accuracy
won Predicting g but still to Network System)
t
In-Hospital worthy patie (RNN)
h
Cardiac nts
e
Arrest
A
m
e
ri
 c
a
n
H
e
a
r
t
A
s
s
o
c
i
a
ti
o
n
B
M
C
B
DeepM6ASe
i
q: prediction
o
and Not inherent
i Deep
characterizat Deep in
Yiqian n 201 6161 Learning Classifica M6A - containing Improve
(35) Biology ion of m6A- Learnin Y radiotherapy, DeepM6ASeq
Zhang f 8 1 Framew tion sequences knowledge
containing g but still
o ork
sequences worthy
r
using deep
m
learning
a
ti
c
s
S
c
i Artificial
e intelligence
n predicts the
ti progression
Machin
fi of diabetic
MasakiMa 201 e 4070 6405 Predicto Improveac
(61) c Endocrinology kidney Y NA EMR Multiple algoithms
kino 9 learnin 76 9 utcomes curacy
r disease
g
e using big
p data
o machine
r learning
t
s
 K
o
r
e
a
Developmen
n
t and
C
Validation of
ir
Deep-
c Deep-learning
Learning
u algorithm for ECG-
Algorithm Deep
JoonMyou l 201 5516 2276 Detectio Ecocardiography based HF identification Improveac
(40) Cardiology for Learnin Y DNN ECG
ngKwon a 9 3 5 n images (DEHF) , integrated with curacy
Electrocardi g
ti TensorFlow (the Google
ography-
o Brain Team)
Based Heart
n
Failure
J
Identificatio
o
n
u
r
n
a
l
H
Comparison
H
of Deep
S
Learning
P
With Articl
u Whole cell screens,
Multiple e
b individual proteins,
Machine Deep refers Improve
Alexandru li Pharmaceutic 201 4790 DNN, Classifica physicochemical
(84) Learning Learnin Y only DNN, SVM performan
Korotcov c s 7 32 SVM tion properties as well as a
Methods g to ces
A dataset with a
and Metrics image
c complex endpoint
Using s
c
Diverse Drug
e
Discovery
s
Data Sets
s
N
P Deep
J learning
D enables
i robust
g assessment Deep
PegahKho 201 1200 1014 Detectio Microscop Embryo time-lapse Stork Improveac
(56) it Embryology and Learnin Y DNN
sravi 9 1 8 n y images images (GoogLeNetInception) curacy
a selection of g
l human
M blastocysts
e after in vitro
d fertilization
i
 c
i
n
e

I
n
t
e
r
n
a
ti
o
n
a
l
J
o
DL and multiple ML
u Deep
(random forest (RF),
r Learning/Art
support vector machine
n ificial
(SVM), linear
a Intelligence Detectio
Ray O Deep discriminant analysis
l 201 and Blood- n/ Improveac
(80) Bahado- Neurology Learnin Y 230 44 ML, DNN CpG loci (LDA), prediction
o 9 Based DNA Predicto curacy
Singh g analysis for microarrays
f Epigenomic utcomes
(PAM), and generalized
M Prediction of
linear model (GLM)—
o Cerebral
(logistic
l Palsy
regression))algorithms
e
c
u
l
a
r
S
c
i
e
n
c
e
s
E Deep Articl
L learning for e DeepLearning4j
Deep Improve
Yilong S Pharmaceutic 201 in vitro refers Predicto Pharmaceuticalformul machinelearningframe
(82) Learnin Y 276 DNN performan
Yang E s 9 prediction of only utcomes ations work(https://deeplear
g ces
V pharmaceuti to ning4j.org/)
I cal image
 E formulations s
R

A novel deep
learning Inception networks,
O
architecture VGG19
n
outperformi (http://www.robots.ox.
c
ng 'off-the- ac.uk/~vgg/research/ve
o
shelf' ry_
l
transfer deep/), DenseNet
o
learning and (https://ai-pool.com/m/
g Deep Mammogr
Eleftherio 201 feature- 1056 Detectio densenet-1556378134) Improveac
(13) y Oncology Learnin Y 2661 CNN aph Mammograph images
sTrivizakis 9 based 1 n and NASNet curacy
R g images
methods in (https://ai-pool.com/m/
e
the nasnet-1556378807).
p
automated Additionally, a custom
o
assessment end-to-end 2D CNN
r
of architecture
t
mammograp trained on the studied
s
hic breast datasets was developed
density
Deep
learning for
liver tumor
The remaining
diagnosis
missing
part II:
informations
H convolutiona Deep Improvecli
Clinton J 201 are described Classifica Radiograp
(90) H Oncology l neural Learnin Y 494 CNN MRI images nicalpractic
Wang 9 in the "Part I" tion hs
S network g ality
of this article,
interpretatio
but I can't find
n using
it
radiologic
imaging
features
Demographic
information,
treatment and
Artificial DAHF TensorFlow (the
medication,
P intelligence Google Brain
laboratory
L algorithm Team,Mountain View,
results,
Joon- O for Deep United States)
201 Predicto electrocardiography Improveac
(41) MyoungK S Cardiology predicting Learnin Y 6924 NA randomForest, glmulti,
9 utcomes (ECG) and curacy
won O mortality of g e1071, and bnlearn
echocardiography
N patients packages in R (R
findings, final
E with acute Development Core
diagnosis, clinical
heart failure Team, Vienna, Austria)
outcome during their
hospital stay, and 12-
month prognosis after
 discharge

A series of a ConvLSTM
layer and three
Convolutional (CNV)
layers, each followed by
Deploying
a Max. pooling layer.
Machine and
Subsequent to
Deep
V these layers, a global
Learning
ir average pooling (GAP)
Models for Deep CNN and
Ahmed u Infectiousdise 202 Detectio X-ray and Chest X-ray and CT acts on the fully Improveac
(6) Efficient Learnin Y 50 50 ConvLST
Sedik s ase 0 n CT images images connected layer. Finally, curacy
Data- g M
e a dense layer
Augmented
s with size of two classes
Detection of
(i.e., positive (normal)
COVID-19
and negative (infected))
Infections
is utilised for the
classification
decision in the
classification phase
J
o
u
r Automated
n Detection of
Caffe, an opensource
a Macular
CNN framework, was
l Diseases by
used for deep learning
o Optical
(http://caffe.berkeleyvi
f Coherence
sion.org/) & deep
O Tomography Deep Optical
SoichiroK Ophthalmolog 201 Detectio learning processes were Automated
(75) p and Artificial Learnin Y 1200 1200 CNN OCT coherencetomgraphy
uwayama y 9 n performed on an etection
h Intelligence g (OCT)
Amazon Web Service
t Machine
EC2 Instance with a
h Learning of
graphics processing unit
a Optical
(https://aws.amazon.co
l Coherence
m/ec2/).
m Tomography
o Images
l
o
g
 y

E
u
r
o
p
e
a
n
H
e
a
r
t
J
o
u Utility of
r machine
R/keras (see
n learning
https://lmb.informatik.
a algorithms in Detectio
Deep uni-
Gerhard- l 201 assessing n Ecocardiography Improveac
(43) Cardiology Learnin Y 199 FCNN ECG freiburg.de/people/ron
Paul Diller C 9 patients /Segmen images curacy
g neber/u-net/) & a
a with a tation
dedicated MatLab
r systemic
(Version
d right
i ventricle
o
v
a
s
c
u
l
a
r
I
m
a
g
i
n
 g

S
p
ri
n
g
e
r
O
Optical
p
Coherence
e
Tomography
n
-Based
-
Deep-
O
Learning
p
Models for
h
Classifying
t
Normal and
h Deep DCNN / Optical
NaohiroM Ophthalmolog 201 Age-Related Classifica Improveac
(9) a Learnin Y 1621 1621 Transfer OCT coherencetomography Non Indicated
otozawa y 9 Macular tion curacy
l g learning (OCT)
Degeneratio
m
n and
o
Exudative
l
and Non-
o
Exudative
g
Age-Related
y
Macular
a
Degeneratio
n
n Changes
d
T
h
e
r
a
p
y
S Beyond
p Human
ri Perception:
n Sexual Deep Make the
SehyoYun 201 1031 1031 Detectio Radiograp
(72) g Radiology Dimorphism Learnin Y DCNN Hand radiographs detectionp
e 9 8 8 n hs
e in Hand and g ossible
r Wrist GoogLeNet from Caffe
J Radiographs Zoo [Jia Y. Caffe model
o Is zoo. 2015]
 u Discernible
r by a Deep
n Learning
a Model
l
o
f
D
i
g
it
a
l
I
m
a
g
i
n
g
S
p
ri
n
g
e
r
J
o
u
r AdaptAhead
n Optimization Articl
a Algorithm e
l for Learning Deep refers
FarnazHos 201 Segment Improveac
(20) o Oncology Deep CNN Learnin Y 230 only DCNN MRI Brain MR images AdaptAhead
eini 9 ation curacy
f Applied to g to
D MRI image
i Segmentatio s
g n
it
a
l
I
m
a
g
i
n
g
 J
M
I
A Deep-
R
Learning
M
Algorithm
e
(ECG12Net)
d
for
i
Detecting
c
Hypokalemia Deep
Chin- a 202 6632 4018 Detectio Ecocardiography Improveac
(44) Cardiology and Learnin Y DCNN ECG ECG12Net
Sheng Lin lI 0 1 0 n images curacy
Hyperkalemi g
n
a by
f
Electrocardi
o
ography:
r
Algorithm
m
Developmen
a
t
ti
c
s
J
o
u
r
n
a
l
o Deep
f Learning
M With
e Electronic
d Health
i Records for
c Short-Term Deep
YasmeenA 202 1000 Predicto Radiograp Improveac
(69) a Radiology Fracture Risk Learnin Y DNN Radiographs Crystal Bone
darAlmog 0 000 utcomes hs curacy
l Identificatio g
I n: Crystal
n Bone
t Algorithm
e Developmen
r t and
n Validation
e
t
R
e
s
e
a
 r
c
h

H
H
S
Impact of
P
pre- Articl
u
analytical e
b
variables on Deep refers Digital
Andrew D li Histopatholog 201 Detectio Improveac
(52) deep Learnin Y 3000 only CNN Pathology PNG or JPG images ResNet50 , SqueezeNet
Jones c y 9 n curacy
learning g to images
A
accuracy in image
c
histopatholo s
c
gy
e
s
s
H Design and
H clinical
S validation of
P a point-of- Maximally stable
u care device extremal regions
Validate an
b for the (MSER) blob detection
Deep automated
Hyungsoo li 201 diagnosis of 1000 Detectio Microholo algorithm, faster non-
(97) Oncology Learnin Y 40 CNN, ML Microholography and low-
n Im c 8 lymphoma 0 n graphy maximum suppression
g cost
A via contrast- algorithm and
method
c enhanced convolutional neural
c microhologr network (CNN)
e aphy and
s machine
s learning
J
A Automated
M Grading of
A Age-Related
O Macular
AlexNet (University of
p Degeneratio
Deep DCNN , Retinalfun Toronto) DCNN model
Philippe h Ophthalmolog 201 n From Color 1338 Detectio Improveac
(76) Learnin Y 4613 Transfer dus Retinalfundus images & the Keras and
M Burlina t y 7 Fundus 21 n curacy
g Learning images TensorFlow DL
h Images
frameworks.
a Using Deep
l Convolution
m al Neural
o Networks
l
 o
g
y

B
M
C
M
e
d
i
c
a
l
I
n
f
Two-stage
o
framework
r Faster RCNN [Ren S, He
for optic disc
m Articl K, Girshick R, Sun J.
localization
a e Faster R-CNN: towards
Muhamm and Classifica
ti Deep refers Retinalfun real-time object
ad Ophthalmolog 201 glaucoma RCNN / tion / Improveac
(77) c Learnin Y 780 only dus Retinalfundus images detection with region
NaseerBaj y 9 classification DCNN Detectio curacy
s g to images proposal networks. In:
wa in retinal n
a image Advances in neural
fundus
n s information processing
images using
d systems; 2015. p. 91–9.]
deep
D
learning
e
c
i
s
i
o
n
M
a
k
i
n
g
P Implementin Build a
L g artificial Machin biological
Lack of
O 201 neural e Detectio neural network through neural
(98) Hu He Biology Y availableinfor ANN
S 9 networks learnin n bionic construction network
mations
O through g able to
N bionic equal a
 E construction biological
organ

F
r
o
n
ti
e A Novel
r Transfer
s Learning
Deep
i Approach to
Transfer Improve
n Enhance
Deep Lack of Learning accuracy,
N 201 Deep Neural Classifica
(21) Hailong Li Neurology Learnin Y availableinfor 60 Neural Brain MRI Brain MRI DTL-NN framework specificity
e 8 Network tion
g mations Network and
u Classification
(DTL - sensitivity
r of Brain
NN)
o Functional
s Connectome
c s
i
e
n
c
e
Matlab - DeepSCN
Fear Level
S [Deep Stochastic
Classification
e Configuration
Based on Machin Develop a
n Networks. Available
Oana 201 Emotional e Classifica system for
(99) s Sociology Y 6192 32 ML, DNN EEG EEG online:
Balan 9 Dimensions learnin tion treating
o http://www.deepscn.co
and Machine g phobias
r m/index.php (accessed
Learning
s on
Techniques
21 March 2019).].
Deep neural
networks
show an
equivalent
P Articl
and often
L e
superior
O Deep refers
SeungSeo 201 performance 5092 Detectio Nail Improveac
(50) S Dermatology Learnin Y only RCNN Nail images ResNet - 152 , VGG - 19
g Han 8 to 5 n images curacy
O g to
dermatologi
N image
sts in
E s
onychomyco
sis diagnosis:
Automatic
construction
 of
onychomyco
sis datasets
by region-
based
convolutiona
l deep
neural
network
W
a Automated
n Assessment Articl
f for Bone Age e
a of Left Wrist Deep refers
201 Classifica Radiograp Improveac
(70) T H Hu n Radiology Joint in Learnin Y only 472 CNN Left wrist X-ray films AlexNet
8 tion hs curacy
g Uyghur g to
d Teenagers patie
a by Deep nts
t Learning
a
S
p
ri
n
g
e
r
Automatic
J
Determinati
o
on of the
u
Need for
r
Intravenous Articl
n
Contrast in e
a Machin
Musculoskel refers
HariTrived l 201 e Classifica Improveeff
(63) Radiology etal MRI Y 1520 only ML MRI Musculoskeletal MRI IBM Watson
i o 8 learnin tion iciency
Examination to
f g
s Using IBM image
D
Watson's s
i
Natural
g
Language
it
Processing
a
Algorithm
l
I
m
a
g
i
n
 g

F
r
o
n
ti
e
r
s
i
n
C
e
ll
a
n
d
AI-Assisted runSVM.m&StaticKinet
D Machin Classifica
Josh Forward MATLAB application
e 201 e tion / Microscop Fluorescentkinetochor Improveeff
(32) Lawrimor Biology Modeling of Y CNN (https://github.com/Blo
v 9 learnin Detectio y images e images iciency
e Biological omLabYeast/KineticBut
e g n
Structures Shakeless)
l
o
p
m
e
n
t
a
l
B
i
o
l
o
g
y
Evaluation of
J Combined Artifici multiple including
Thomas a 202 Artificial al 1442 8558 Detectio Radiograp Screening Therapixel model, Improveac
(64) Radiology Y DCNN
Schaffter m 0 Intelligence Intellig 31 0 n hs mammograms Faster R-CNN, curacy
a and ence DeepHealth’s model
Radiologist
 Assessment
to Interpret
Screening
Mammogra
ms
C
o
m
p
u
t
a
ti
o
n
a
l
a
n
d
M
a Gait
t Biomarkers
h Classification
e by
Classifica WEKA tool(available
Eddy m Combining Deep
201 tion / from Improveac
(62) Sánchez- a Endocrinology Assembled Learnin Y 15 ML photo photo
9 Detectio http://www.weka.org) curacy
DelaCruz ti Algorithms g
n & Dl4jMLPClassifier
c and Deep
a Learning:
l Results of a
M Local Study
e
t
h
o
d
s
i
n
M
e
d
i
c
i
n
e
 Noninvasive
assessment
of dofetilide
plasma
concentratio
P n using a Articl
L deep e
O learning Deep refers
Zachi I 201 Predicto ECG , plasma Improveeff
(45) S Cardiology (neural Learnin Y only 42 DCNN EEG Adam optimizer
Attia 8 utcomes drugconcentrations iciency
O network) g to
N analysis of patie
E the surface nts
electrocardi
ogram: A
proof of
concept
study
B
M
C
M
e
d
i
c
a
l
I
n EEG-based
f image
o classification
r via a region-
Deep Improve
Ahmed m 201 level stacked LTSM Classifica
(29) Neurology Learnin Y 2000 EEG EEG Bidirectional LSTM performan
Fares a 9 bi- Network tion
g ces
ti directional
c deep
s learning
a framework
n
d
D
e
c
i
s
i
o
n
M
 a
k
i
n
g
S
p
ri
n
g
e
r
O
p
e
n
Improve
-
communic
E
ation
u
between
r
radiologists
o Deep
and
p learning to
Articl clinicians
e convert
e without
a unstructured Computed convolutional neural
Adam Deep refers loss of
n 201 CT Detectio tomography network (CNN)-based
(100) Spandorfe Radiology Learnin Y only 875 CNN CT productivit
R 9 pulmonary n pulmonary text classification
r g to y and
a angiography angiography (CTPA) algorithm
patie provide
d reports into
nts more
i structured
structured
o reports
data for
l
research/d
o
ata mining
g
application
y
s
E
x
p
e
ri
m
e
n
t
a
l
 Comparative
H effectivenes
H s of
S convolutiona
P l neural
Articl
u network
e
b (CNN) and
Deep refers
ImonBane li 201 recurrent CNN / Classifica CNN Word - Glove , DPA Compariso
(101) Radiology Learnin Y 7370 only CT Thoracic CT
rjee c 9 neural RNN tion - HNN n
g to
A network
image
c (RNN)
s
c architecture
e s for
s radiology
s text report
classification
E
L
S
E
V
I
E Automated
R Deep
- Learning-
S Based
t System to
Deep Fluorescen phase-contrast and High
Dai e 201 Identify Detectio
(33) Biology Learnin Y 800 CNN tkinetocho immunofluorescent AlexNet&LeNet performan
Kusumoto m 8 Endothelial n
g re images images ce
Cells Derived
C from
e Induced
ll Pluripotent
R Stem Cells
e
p
o
r
t
s
E Dermoscopy
L diagnosis of
S cancerous
Dermosco Dermoscopic and
E lesions Deep
BN Dermatology/ 201 Detectio pic and context images (i.e. Improveac
(48) V utilizing dual Learnin Y 5161 4499 CNN Inception V2
Walker Oncology 9 n context non - dermoscopic curacy
I deep g
images regular photos)
E learning
R algorithms
- via visual
 E and audio
B (sonification
i ) outputs:
o Laboratory
M and
e prospective
d observationa
i l studies
c
i
n
e
J
o
u
r
n
a
l
o
f
B
i
o
m
Combining
e
deep
d
learning and
i
coherent
c
anti-Stokes
a Deep Coherent anti-Stokes
Sheng 201 Raman Classifica Spectrosco Improveac
(92) l Oncology Learnin Y 7926 388 DCNN Raman scattering GoogleNet
Weng 7 scattering tion pic curacy
O g images
imaging for
p
automated
ti
differential
c
diagnosis of
s
lung cancer
(
S
P
I
E
D
i
g
it
a
l
L
i
 b
r
a
r
y
)
B
M
C
J
o
u
r
n
a
l
o
f
O
t
o
l
a
r
y Building an
n Otoscopic
Develop an
Devon g screening Deep
Otolaryngolog 201 Classifica TensorFlow frame work automatic
(86) Livingston o prototype Learnin Y 734 DCNN OCT Otoscopic images
y 9 tion in Python identifier
e l tool using g
software
o deep
g learning
y
-
H
e
a
d
&

N
e
c
k
S
u
r
g
e
r
 y

W
o
lt
e
r
s Retrospectiv
K e imaging
Numb
l studies of Numb
er of
u gastric er of Classifica deep CNN with Improve
Deep patie
Zixing w 202 cancer: image tion / Depthwise Convolution performan
(102) Oncology Learnin Y nts DCNN CT CT images
Huang e 0 Study s not Detectio blocks and Fully ces and
g not
r protocol specif n Connection accurancy
specif
M clinical trial ied
ied
e (SPIRIT
d Compliant)
i
c
i
n
e
J
o
u
r
n
a
l Pixel-Level
o Deep
f Segmentatio
D n: Artificial
i Intelligence
Deep Caffe % fully
Hyunkwa g 201 Quantifies Segment Improvequi
(65) Radiology Learnin Y 400 400 FCNN CT CT images convolutional network
ng Lee it 7 Muscle on ation ckness
g (FCN)
a Computed
l Tomography
I for Body
m Morphometr
a ic Analysis
g
i
n
g
(
S
 p
ri
n
g
e
r
)
E
L
S
E
V
I
Deep
E
Learning
R
Neural
S
Networks
t
Highly
e Deep ResNet50-SA, Adam,
Ariel 201 Predict Very Classifica Microscop Transmitted light Improveac
(103) m Biology Learnin Y 2120 CNN Adamax, Adagrad,
Waisman 9 Early Onset tion y images microscopy images curacy
g Nadam and RMSprop
of
C
Pluripotent
e
Stem Cell
ll
Differentiati
R
on
e
p
o
r
t
s
J
o
u
r
n Caffe (Jia Y, Shelhamer
Machine
a E, Donahue J, Karayev S,
learning-
l Long J, Girshick R, et al.
based
o Caffe:convolutional
prediction of Machin Detectio Provide a
f architecture for fast
Mengxua Pharmaceutic 201 adverse drug e n/ LFP new
(83) P Y 70 ML Alveus LFP images feature embedding.
n Gao s 7 effects: An learnin Predicto images detection
h arXiv. 2014;140: 5093.)
example of g utcomes platform
a & custom-made
seizure-
r routines in Matlab (The
inducing
m MathWorks, Natick,
compounds
a MA, USA).
c
o
l
o
 g
i
c
a
l
S
c
i
e
n
c
e
s
J
o
u
r
n
a
l
o
f
P Classification
a of
t Melanocytic Hematoxylin and eosin
h Lesions in (H and E) slides each
Deep Whole - Improve
Steven N o 201 Selected and Classifica containing
(46) Dermatology Learnin Y 300 CNN slide TensorFlow performan
Hart l 9 Whole-Slide tion conventional (n =
g images ces
o Images via 150) or Spitz nevi [n =
g Convolution 150, Figure 1]
y al Neural
I Networks
n
f
o
r
m
a
ti
c
s
B Prospective
ri evaluation
ti Endocrinology of an Classifica Improveac
Deep Digital
Peter s / 202 artificial 3040 3040 tion / Digital curacy,
(59) Learnin Y NA DiabeticRetinopathy EyeArt v2.1
Heydon h Ophthalmolog 0 intelligence- 5 5 Detectio images performan
g images
J y enabled n ces
o algorithm
u for
 r automated
n diabetic
a retinopathy
l screening of
o 30000
f patients
O
p
h
t
h
a
l
m
o
l
o
g
y
F
r
o
DeepExplain toolbox
n
(Ancona M, Ceolini E,
ti
Öztireli C, Gross M.
e Using Deep
Towards Better
r Learning
Understanding of
s Algorithms
Gradient-Based
i to Grade Deep
Lauren C 202 Classifica Ultrasoun Renalultrasound Attribution Methods for High
(88) n Nephrology Hydronephr Learnin Y 2420 673 DCNN
Smail 0 tion d images images Deep Neural Networks accuracy
P osis Severity: g
(2017).
e Toward a
Available online at:
d Clinical
http://arxiv.org/abs/17
i Adjunct
11.06104 -
a
accessedSeptember 20,
t
2018)
ri
c
s
A novel
YOLOv3-arch
P
model for
L
identifying Classifica
O Deep
Shanchen 201 cholelithiasis 2238 tion / Spine, liver and gall CT Improveac
(66) S Epatology Learnin Y 1369 FCNN CT images Yolov3-arch
Pang 9 and 46 Detectio images curacy
O g
classifying n
N
gallstones
E
on CT
images
 B
M
C De novo
B profile
i generation
Articl
o based on
e Predicto
i sequence
Deep refers utcomes Improve
Kazunori n 201 context 1603 LSTM
(36) Biology Learnin Y only / Aminoacidssequences LSTM network performan
D Yamada f 8 specificity 667 network
g to Detectio ces
o with the
image n
r long short-
s
m term
a memory
ti network
c
s
S
p
ri
n
g
e
r
J
o
u
r
n Fully
a Automated
Segment
l Deep Deep
Hyunkwa General 201 ation / Radiograp Digital X-ray Improveac
(71) o Learning Learnin Y 8325 8325 CNN ImageNet
ng Lee medicine 7 Predicto hs radiogrammetry curacy
f System for g
utcomes
D Bone Age
i Assessment
g
it
a
l
I
m
a
g
i
n
g
F Diagnosis of
Deep T1-MR and FDG-PET
Yechong r 201 Alzheimer's Classifica TensorFlow (Abadi et Improveac
(30) Neurology Learnin Y 5722 2145 CNN images of the
Huang o 9 Disease via tion al., 2016). curacy
g hippocampal area
n Multi-
 ti Modality 3D
e Convolution
r al Neural
s Network
i
n
N
e
u
r
o
s
c
i
e
n
c
e
N
a
t
u
r
e
r
e
s
e
Efficient
a
Deep Adam optimizer
r
Network (Kingma, D. P. & Ba, J.
c
Architecture Adam: A Method for
h Deep
201 s for Fast Classifica Radiograp Stochastic Improveeff
(67) F Pasa S Radiology Learnin Y 1104 1104 CNN Chest X - Rays
9 Chest X-Ray tion hs Optimization. 1–15. iciency
c g
Tuberculosis https://doi.org/10.1145
i
Screening /1830483.1830503
e
and (2014))
n
Visualization
ti
fi
c
R
e
p
o
r
t
s
 N
a
t
u
r
e
r
e
s
Predictive
e Med2Vec33 & a set of
Modeling of
a deep learning models,
the Hospital
r specifically
Readmission
c Convolutional Neural
Risk from Machin
h Networks (CNN)13, Improve
201 Patients' e 1119 Predicto Medicalcla Patients'
(104) Xu Min S Pneumology Y NA Recurrent Neural performan
9 Claims Data learnin 92 utcomes ims medicalclaims
c Networks (RNN)18 , ces
Using g
i Long-Short Term
Machine
e Memory (LSTM)29,
Learning: A
n Gated Recurrent Unit
Case Study
ti (GRU)30).
on COPD
fi
c
R
e
p
o
r
t
s
N
P
J
D
Impact of a
i
deep
g
learning
it
assistant on CAM generation, freely
a Deep Hematoxylin and
Amirhosse 202 the 2500 Classifica available at Improveac
(22) l Oncology Learnin Y NA eosin-stained whole-
inKiani 0 histopatholo 0 tion https://github.com/sta curacy
M g slide images (WSI)
gic nfordmlgroup/lca-code.
e
classification
d
of liver
i
cancer
c
i
n
e
 I
n
d
i
a
n
J
Validation of
o
Deep
u
Convolution
r
al Neural
n
Network-
a
based
l Endocrinology
algorithm Deep Macula-
Payal o / 202 1124 Detectio Improveac
(60) for detection Learnin Y 675 DCNN centeredfundus 17-layer CNN
Shah f Ophthalmolog 0 89 n curacy
of diabetic g images
O y
retinopathy -
p
Artificial
h
intelligence
t
versus
h
clinician for
a
screening
l
m
o
l
o
g
y
A
R
V
O

T
Effects of
r
Study
a
Population,
n
Labeling and
Mark s Deep Improve
Ophthalmolog 202 Training on 14.82 Detectio
(78) Christoph l Learnin Y 2329 NA OCT Fundusphotographs ResNet50 & ResNet34 performan
y 0 Glaucoma 2 n
er a g ces
Detection
ti
Using Deep
o
Learning
n
Algorithms
a
l
V
i
s
i
 o
n
S
c
i
e
n
c
e
&

T
e
c
h
n
o
l
o
g
y
N
a
t
u
r
e
r
e
s
e
a Relevance of
r deep
c learning to Classifica multiple cells per
Michel E Deep
h 201 facilitate the tion / images annotated in Improveac
(55) Vandenbe Oncology Learnin Y 71 71 ML, CNN ConvNets
S 7 diagnosis of Detectio Breast tumor curacy
rghe g
c HER2 status n resection samples
i in breast
e cancer
n
ti
fi
c
R
e
p
o
r
t
 s

Comparing
B biological
M information
C contained in
Artifici Deep
Johannes C mRNA and Illumina HiSeq 2000 - Rprop algorithm &
201 al 1236 Belief Classifica Improveac
(39) Smolande a Oncology non-coding Y 124 generated RNA - seq Restricted Boltzmann
9 Intellig 0 Network tion curacy
r n RNAs for data Machine
ence s
c classification
e of lung
r cancer
patients
P
L
O
S
C
o
m
p
u Evaluating
t reproducibili
a ty of AI Deep DAPPER code (at
Andrea 201 Classifica Improveac
(51) ti Health care algorithms in Learnin Y 787 CNN Whole Slide Images gitlab.fbk.eu/mpba-
Bizzego 9 tion curacy
o digital g histology/dapper)
n pathology
a with DAPPER
l
B
i
o
l
o
g
y
Local
Interpretabl The
S Generate
e Model- suddivision in
e explanatio
Agnostic training/test/ Classifica
IamPalatn n Deep LIME algorithm and ns on how
201 Explanations validation is 2200 tion / Histologywhole slide
(54) ik de s Oncology Learnin Y CNN SLIC, quickshift, and a CNN
9 for the same of 26 Detectio images
Sousa o g FHA algorithms detects
Classification the n
r tumor
of Lymph Camelyon16
s tissue
Node (not found)
Metastases
 A
Comparison
S
Study of
e
Classifier Machin
n
201 Algorithms e Classifica Remarkesi
(85) Yago Saez s Health care Y 9 CNN Physical activity data Google'sTensorFlow
6 for Cross- learnin tion milarities
o
Person g
r
Physical
s
Activity
Recognition
B
M
C
B
OffsampleAI:
i
artificial
o
intelligence
i Machin METASPACE available
Katja approach to Improve
n 202 e Informationsn 2323 Classifica Mass spectrometry at:
(25) Ovchinnik Biology recognize Y CNN performan
f 0 learnin otavaliable 8 tion images https://github.com/met
ova off-sample ces
o g aspace2020/offsample.
mass
r
spectrometr
m
y images
a
ti
c
s
KymoButler,
a deep Automate
e
Maximilia learning Deep tracking,
L 201 Classifica
(105) n Ah Biology software for Learnin Y 787 CNN Kymographs KymoButler improve
if 9 tion
Jakobs automated g performan
e
kymograph ces
analysis
H
H
S
P DeepNAT:
u Deep
b convolutiona
Christian Deep Improve
li 201 l neural 3300 Segment T1-weighted magnetic
(106) Wachinge Neurology Learnin Y DCNN DeepNAT performan
c 8 network for 0 ation resonance images
r g ces
A segmenting
c neuroanato
c my
e
s
s
 J
o
u
r
n
a
l
o
f
M
Artificial
e
Intelligence
d
and Its Effect
i
on
c
Dermatologi Improveac
a
sts' Accuracy Deep curacy,
Roman C l Dermatology/ 202 Classifica Dermoscopic images
(49) in Learnin Y 1200 CNN improve
Maron I Oncology 0 tion of melanoma and nevi
Dermoscopic g performan
n
Melanoma ces
t
Image
e
Classification
r
: Web-Based
n
Survey Study
e
t
R
e
s
e
a
r
c
h ResNet-50
O
Developmen
x
t of an
f
artificial
o
intelligence-
r
based Residual
d
assessment Network
J Predicto
model for s,
o Deep utcomes Images captured by
M 202 prediction of Densely non-invasive Life Improveac
(57) u Embryology Learnin Y 8886 / optical light
VerMilyea 0 embryo Connect Whisperer AI model curacy
r g Classifica microscopy
viability ed
n Articl tion
using static Network
a e
images s
l refers
captured by
s only
optical light
H to
microscopy
u patie
during IVF
m nts
 a
n
R
e
p
r
o
d
u
c
ti
o
n
F
r
o
n
ti
e
r
Crowdsourci
s
ng the
i Automate
creation of IDSIA, BlackEagles, MLL-
n the
Ignacio image Deep ETH, SCI, CellProfiler,
N 201 Segment Electron microscopic reconstruct
(107) Arganda- Neurology segmentatio Learnin Y 30 NA Harvard, CoMPLEX,
e 5 ation brain images ion of
Carreras n algorithms g UCL, TSC+PP, IMMI,
u neural
for CLP, Freiburg, NIST
r circuits
connectomic
o
s
a
n
a
t
o
m
y
F
e
d
e
Using Articl
r
Artificial e
a Artifici Classifica
Intelligence refers Improve
Andrew A l 202 al tion / Radiograp
(26) Radiology for COVID- Y 1133 only CNN Chest X-ray Microsoft CustomVision performan
Borkowski P 0 Intellig Detectio hs
19 Chest X- to ces
r ence n
ray image
a
Diagnosis s
c
ti
ti
o
 n
e
r
novel our algorithm + a
patch-based 3D Dense-
Unet (Jegou S, Drozdzal
M, Vazquez D, Romero
T
A, Bengio Y. The one
o A Fully
Articl hundred layers
m Automated
Chandan e Multipara tiramisu:fully
o Deep
Ganesh Deep refers metricmag Brain multiparametric convolutional
g 202 Learning Segment Improveac
(73) Bangalore Oncology Learnin Y only 449 CNN neticreson magnetic resonance densenets for semantic
r 0 Network for ation curacy
Yoganand g to ance images segmentation. In: 2017
a Brain Tumor
a patie images IEEE conference on
p Segmentatio
nts computer vision and
h n
pattern recognition
y
workshops (CVPRW).
IEEE:
Honolulu, HI;
2017:1175–1183.).
Machine
Learning
Model for
B Articl
Identifying
i e
Antioxidant Machin Classifica Deeplearing4j or Eclipse
Luu Ho o refers
202 Proteins e tion / Deeplearning4 Improveac
(37) Thanh l Biology Y 2020 only DNN Proteinsequencies
0 Using learnin Detectio integrated with WEKA curacy
Lam o to
Features g n tool
g image
Calculated
y s
from
Primary
Sequences
E
L
S
E MitoSegNet:
V Easy-to-use
Articl Fluorescence
I Deep
e microscopy images of
E Learning
Deep refers mitoGFP expressing
Christian R 202 Segmentatio Segment Improveeff
(34) Biology Learnin Y 3840 only CNN mitochondria in wild- MitoSegNet
A Fischer i 0 n for ation iciency
g to type and catp-6
S Analyzing
image ATP13A2 mutant C.
c Mitochondri
s elegans adults
i al
e Morphology
n
c
e
 T
r
a
n
s
l To develop
a an artificial
ti intelligence
o (AI)–based
n structure-
a function
l (SF) map
v relating
i retinal
s nerve fiber
i Artificial layer
o Intelligence Artifici (RNFL)
Eduardo B Spectral domain
n Ophthalmolog 202 Mapping of al 2649 Detectio damage on
(79) Mariotton Y 8878 CNN optical coherence Keras (Phyton)
s y 0 Structure to Intellig 9 n spectral
i tomography (SDOCT)
c Function in ence domain
i Glaucoma optical
e coherence
n tomograph
c y (SDOCT)
e to
& functional
t loss on
e standard
c automated
h perimetry
n (SAP)
o
l
o
g
y
 J
o
u
r
n
a
Face++ (Jung SG, An J,
l
Kwak H, Salminen J,
o
Jansen BJ. Inferring
f
social media users’
M
demographics from
e
profile pictures: A
d
Diagnosing Face++ analysis on
i
Parkinson Twitter users.
c
Disease Proceedings of the 17th
a Machin
Through International
l 202 e Detectio Videos of Improveac
(81) Bo Jin Neurology Facial Y 176 64 CNN Conference
I 0 learnin n facialexpressions curacy
Expression on Electronic Business;
n g
Recognition: August, 2015; Seoul,
t
Video Republic of Korea. New
e
Analysis York, NY, United States:
r
Association for
n
Computing Machinery;
e
2017. pp. 140–145.
t
https://aisel.aisnet.org/
R
iceb2017/22/
e
[Google Scholar])
s
e
a
r
c
h
J
o
u
r
n CTRL - a
a label-free
l artificial
Artifici
o intelligence U-Net & CTRL: Cell
202 al Segment Microscop Improveac
(38) KaiYao f Biology method for Y 1890 CNN microscopy images Topography
0 Intellig ation y images curacy
C dynamic Reconstruction Learner
ence
e measureme
ll nt of single-
S cell volume
c
i
e
n
 c
e

E
L
S
E
V
I
E
R
P
u
b Application
li of deep
c learning
H technique to
e manage
AlexNet, VGG-16, VGG-
a COVID-19 in
19, SqueezeNet,
Ali lt routine Deep Improve
Infectiousdise 202 Classifica GoogleNet, MobileNet-
(74) Abbasian h clinical Learnin Y 1200 194 CNN CT images Thoracic CT images performan
ase 0 tion V2, ResNet-18, ResNet-
Ardakani E practice g ces
50, ResNet-101,
m using CT
Xception
e images:
r Results of 10
g convolutiona
e l neural
n networks
c
y
C
o
ll
e
c
ti
o
n
N A deep Support vector
a learning machine, decision trees,
t model and and ensemble
u machine algorithms, DCNN
NourEldee Deep
r Infectiousdise 202 learning Classifica Digital (three main Improveac
(87) n M Learnin Y 3750 DCNN Digital cell images
e ase 0 methods for tion images convolutional layers curacy
Khalifa g
P the with window
u classification size 3 × 3 pixels, three
b of potential ReLU layers, and three
li coronavirus pooling
 c treatments layers)
H on a single
e human cell
a
lt
h
E
m
e
r
g
e
n
c
y
C
o
ll
e
c
ti
o
n
A
Provide a
d
Generalizabl generalizab
v
e, le, and
a
Reproducibl neuroscien
n
e, and tifically
c 3D
Neuroscienti interpretab
e Deep attentio
202 fically Classifica le imaging
(31) Dan Jin d Neurology Learnin Y 1832 n MRI ResNet-18
0 Interpretabl tion biomarker
S g network
e Imaging that can
c (3DAN)
Biomarkers support
i
for clinicians in
e
Alzheimer's the early
n
Disease diagnosis
c
of AD
e

Table S1 : Included papers.