Skip to main content

Table 4 Summery of DL-based studies for prediction and classification of SZ from MRI

From: Application of deep learning in detecting neurological disorders from magnetic resonance images: a survey on the detection of Alzheimer’s disease, Parkinson’s disease and schizophrenia

Ref.

Regions

DL

Pre-Proc.

Feature (count)

Dataset

Size

Accuracy

[48]

VFN, CN, DMN

3D-CNN

MC, DN, STC, SS, TF, HPF

3D-ICA (15)

COBRE

72 SZs,74 HCs

98.09%\(^{10\alpha }\)

[54]

AUD, DMN

2D-CNN

MC, SN, SS

ICA(13)

Self

42 SZs,40 HCs

slice-level DMN-72.65%\(^{5\alpha }\), AUD-78.34%\(^{5\alpha }\), subject-level DMN-91.32%\(^{5\alpha }\), AUD-98.75%\(^{5\alpha }\)

[78]

WB

DNN

ICA

FNC, SBM (10)

MRN

69 SZs, 75 HCs

94.4%

[79]

WB

DNN

 

ROI (116)

OpenfMRI

50 SZs, 49 BD, 122 HCs

76.6%\(^{\alpha }\)

[34]

WB

RNN

MC, DN, SF, TF, NM, LRg

SPF

FBIRN phase-II

87 SZs, 85 HCs

64%\(^{10\alpha }\)

[52]

WB

DNN

STC, SN, SS

FNC (116)

COBRE

72 SZs,74 HCs

95.4%\(^{5\alpha }\)

[35]

WB

DNN

 

FNC,SBM (410)

MLSP

  

[32]

WB

DBN

LR, ZN

NMF

Multisite

143 SZs,83 HCs

73.6%\(^{3\alpha }\)

[50]

WB

SAE

STC, MC,SN, SM, F

VTS

COBRE

72 SZs,74 HCs

92%\(^{10\alpha }\)

[51]

Atlas

FFBPNN

STC, MC, TF, NM, SS

FNC (20)

Hospital

39 SZs,31 HCs

79.3%\(^{10\alpha }\)

[55]

WB

DNN, LRP

MC, SN

FNC, ICA (1225)

Multisite

558 SZs, 542 HCs

84.75%\(^{10\alpha }\)

[49]

Cor., Str., Cere.

DNN

MC, NM, STC, SS, LD, TF

FNC (116)

Multisite

474 SZs,607 HCs

\(\approx\)83%\(^{10\alpha }\)

[44]

Vent.

DBN

SST, BC, SG

SV, ROI

COBRE

72 SZs,76 HCs

ROI-83.3%\(^{3\alpha }\), SV-90%\(^{3\alpha }\)

[80]

WB

MLP

 

ICA, RV

FBIRN

135 SZs,169 HCs

AUC- 0.85\(^{8\alpha }\), SD-0.05

[59]

WB

MLP

NM, SG, SS

 

Multisite

198 SZs,191 HCs

AUC-0.75\(^{10\alpha }\), SD-0.04

  1. WB whole brain, Cor. cortical, Str. striatal, Cere cerebellar, Vent. ventricle, MRN mild research network, VFN visual frontal network, AUD auditory cortex, CN cerebellar network, DMN default mode network, \(n\alpha =\)n-fold cross-validation, SPF spatial feature, NMF neuro-morphometric features, VTS voxel time series, SV segmented ventricle, Self self-generated dataset