Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 7

Appendix 4

Summary of all the data extracted from included studies

First author, Machine Study Bipolar type Data Types Accuracy Other Data
year learning Settings statistics for Availability
models Validation
Poletti et al. Not Existing Not specified immune- 86.0% Private
(2021) mentioned inflammatory AUC = 97%
signature
Idemoto et al. Existing Not specified Not Private
(2021) Blood samples mentioned sensitivity of
Not (Serum) 51.6% and
Mentioned the specificity
of 65.3%

Suen et al. Elastic Net Existing BD type I, EMR (clinical 78.0% AUC 84% Private
(2021) algorithm BD Type II & and sensitivity
BD not demographic 75%
specified variables) specificity
81%
Sawalha et SVM Existing chronic BD Neuropsycholo 77.0% sensitivity Private
al. (2021) and first- gical Test 76%
episode BD specificity is
77%
Parker et al. Ensemble Existing bipolar I, Survey 91.8% sensitivity Private
(2021) Model bipolar II assessing 90.9%
manic/hypoma specificity of
nic symptoms 96.6%
Linke et al. Gaussian Existing Not specified fractional 75.0% Private
(2020) process anisotropy sensitivity
classifiers (FA), axial 66.67%
diffusivity (AD), specificity
and radial of 84.21%
diffusivity (RD)
Sonkurt et al. Not Novel Bipolar I neurocognitive 78.0% AUC 78% Private
(2021) mentioned tests sensitivity 80
%
specificity
76.2
Mwangi et al. RVM Existing BD type I, MRI 70.3% 66.4% Private
(2016) BD Type II sensitivity
74.2%
specificity
Cho et al. Random Existing BD type I, Passive digital 64% and Private
(2019) forest to BD Type II phenotypes 65%
develop (Heart ryhthms) AUC 67%,
mood and 67%
prediction
algorithm
Ma et al. Random Existing Not specified bipolarity index 96.0% Private
(2019) forest (BPx) and
algorithm, Affective
support Disorder
vector Evaluation AUC 92.1%
regression scale (ADE)
(SVR), and
logistic
regression
Li et al. SVM Existing Not specified MRI 87.5% sensitivity Private
(2020) 86.4%
specificity
88.9%
Han et al. Not existing Not specified Not mentioned Not Not Not
(2018) mentioned mentioned mentioned mentioned
Jakobsen et Random Existing Not specified activity 84.0% Public
al. (2020) Forrest, monitoring
Deep
Neural 82%
Network sensitivity
and 84%
Convolution specificity
al Neural
Network
algorithms
Laksshman DeepBipola Novel Not specified Genomic data Not public
AUC 65%
et al. (2017) r mentioned
Chandran et NLP existing Not specified EHR Not Not Public
al. (2019) algorithm (searching key mentioned mentioned
words like
YBOCS, OCD,
etc.)
Perez et al. Not Existing Not specified daily mood 75.0% Not private
(2018) mentioned ratings Survey mentioned
Achalia et al. SVM Existing BD type I MRI, rs-fMRI, 87.6% Not
(2020) diffusion tensor sensitivity of mentioned
images 82.3 %
(neuropsycholo specificity of
gical 92.7 %
measures)
Schwarz et random existing Not specified structural and 76.0% private
al. (2019) forest functional MRI
machine AUC 74%
learning
and SVM
Gong et al. Neural existing BD type II rs-fMRI Not Not private
(2019) networks mentioned mentioned
Osuch et al. SVM Existing Not specified fMRI 92.4% sensitivity Private
(2018) 87.5%
specificity
97.1%
Palaniyappan MVAR existing psychotic rs-fMRI 96.2% Private
et al. (2019) bipolar
disorder
Shafquat et Not N/A Not specified phenotypes Not Public
al. (2020) mentioned (GWAS mentioned
datasets) AUC 69.7%
Frangou et al. Support Existing BD type I fMRI 83.5% Public
(2017) vector
machines
(SVM) and
Gaussian sensitivity of
Process 84.6%
Classifier specificity of
(GPC) 92.3%

Deng et al. SVM Existing BD type I, MRI- Diffusion 68.3% Private


Not
(2018) BD Type II tensor images
mentioned
(DTI)
Wu et al. Not Existing Euthymic CANTAB 71.0% Private
(2016) mentioned subjects with cognitive
BD types I or scores- Not
II measure mentioned
cognitive
performance
Mothi et al. Unsupervis Existing Not specified MRI Not Public
(2019) ed machine mentioned Not
learning- mentioned
clustering
Jo et al. Not Existing Not PGBI-10M Not 83% Private
(2018) mentioned mentioned manic symptom mentioned sensitivity
data 89%
specificity
Chung et al. Baseline Existing Not Images of Not Private
(2018) logistic mentioned faces mentioned
regression AUC 87.8%
and RNN
classifier
El Gohary et SVM Existing Not specified EEG data 98.0% Not Not
al. (2016) mentioned mentioned
Saylan et al. 1) K Existing Not specified Microarray 86.0% Public
(2016) Nearest expression Not
Neighbour data set mentioned
Classificatio
n Algorithm
(K-NN). 2)
Decision
Tree
Algorithm.
3) Naive
Bayes
algorithm
Liu et al. Logistic Existing BD type II sMRI and fMRI 83%- 89% Public
(2018) regression AUC 95%
model
Chuang et al. random Existing Not large-scale 85.2% Sensitivity Public
(2017) forest (RF) mentioned genome-wide 77.7%
association Specificity
(GWA) data 85.4%
Erguzel et al. Artificial Existing Not Electroencepha 89.89 % Private
(2016) neural mentioned lography (EEG)
networks -
Sensitivity
particle
83.87 %
swarm
optimization
(ANN–PSO)

Abbreviations: SVM: Support Vector Machine; RF: Random Forest; LR: Logistic Regression; ANN-PSO: Artificial neural networks-particle
swarm optimization; GPC: Gaussian Process Classifier ; FA: Fractional anisotropy; AD: Axial diffusivity; RD: Radial diffusivity ; EEG:
Electroencephalography; GWA: Large-scale genome-wide association data; DTI: Diffusion tensor images; EHR: Electronic Health record;
NLP: Natural Language processing; MRI: Magnetic resonance imaging; fMRI: Functional magnetic resonance imaging;
rs-fMRI: Resting state functional magnetic resonance imaging; CANTAB: Cambridge neurophysiological test automated battery; OCD:
Obsessive compulsive disorder; YBOCS: Yellow-Brown Obsessive-Compulsive Disorder

You might also like