Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site FMRI data

Weizheng Yan, V. Calhoun, Ming Song, Yue Cui, Hao Yan, Shengfeng Liu, Lingzhong Fan, Nianming Zuo, Zhengyi Yang, Kaibin Xu, Jun Yan, Luxian Lv, Jun Chen, Yunchun Chen, Hua Guo, Peng Li, Lin Lu, Ping Wan, Huaning Wang, Huiling WangYongfeng Yang, Hongxing Zhang, Dai Zhang, Tianzi Jiang, Jing Sui

Research output: Contribution to journalArticle

Abstract

Background: Current fMRI-based classification approaches mostly use functional connectivity or spatial maps as input, instead of exploring the dynamic time courses directly, which does not leverage the full temporal information. Methods: Motivated by the ability of recurrent neural networks (RNN) in capturing dynamic information of time sequences, we propose a multi-scale RNN model, which enables classification between 558 schizophrenia and 542 healthy controls by using time courses of fMRI independent components (ICs) directly. To increase interpretability, we also propose a leave-one-IC-out looping strategy for estimating the top contributing ICs. Findings: Accuracies of 83·2% and 80·2% were obtained respectively for the multi-site pooling and leave-one-site-out transfer classification. Subsequently, dorsal striatum and cerebellum components contribute the top two group-discriminative time courses, which is true even when adopting different brain atlases to extract time series. Interpretation: This is the first attempt to apply a multi-scale RNN model directly on fMRI time courses for classification of mental disorders, and shows the potential for multi-scale RNN-based neuroimaging classifications. Fund: Natural Science Foundation of China, the Strategic Priority Research Program of the Chinese Academy of Sciences, National Institutes of Health Grants, National Science Foundation.

Original languageEnglish (US)
JournalEBioMedicine
DOIs
StateAccepted/In press - Jan 1 2019
Externally publishedYes

Fingerprint

Recurrent neural networks
Schizophrenia
Neural Networks (Computer)
Magnetic Resonance Imaging
Neuroimaging
Natural sciences
Natural Science Disciplines
Aptitude
Time series
Brain
Organized Financing
Atlases
National Institutes of Health (U.S.)
Health
Mental Disorders
Cerebellum
China
Research

Keywords

  • Cerebellum
  • Deep learning
  • fMRI
  • Multi-site classification
  • Recurrent neural network (RNN)
  • Schizophrenia
  • Striatum

ASJC Scopus subject areas

  • Biochemistry, Genetics and Molecular Biology(all)

Cite this

Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site FMRI data. / Yan, Weizheng; Calhoun, V.; Song, Ming; Cui, Yue; Yan, Hao; Liu, Shengfeng; Fan, Lingzhong; Zuo, Nianming; Yang, Zhengyi; Xu, Kaibin; Yan, Jun; Lv, Luxian; Chen, Jun; Chen, Yunchun; Guo, Hua; Li, Peng; Lu, Lin; Wan, Ping; Wang, Huaning; Wang, Huiling; Yang, Yongfeng; Zhang, Hongxing; Zhang, Dai; Jiang, Tianzi; Sui, Jing.

In: EBioMedicine, 01.01.2019.

Research output: Contribution to journalArticle

Yan, W, Calhoun, V, Song, M, Cui, Y, Yan, H, Liu, S, Fan, L, Zuo, N, Yang, Z, Xu, K, Yan, J, Lv, L, Chen, J, Chen, Y, Guo, H, Li, P, Lu, L, Wan, P, Wang, H, Wang, H, Yang, Y, Zhang, H, Zhang, D, Jiang, T & Sui, J 2019, 'Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site FMRI data', EBioMedicine. https://doi.org/10.1016/j.ebiom.2019.08.023
Yan, Weizheng ; Calhoun, V. ; Song, Ming ; Cui, Yue ; Yan, Hao ; Liu, Shengfeng ; Fan, Lingzhong ; Zuo, Nianming ; Yang, Zhengyi ; Xu, Kaibin ; Yan, Jun ; Lv, Luxian ; Chen, Jun ; Chen, Yunchun ; Guo, Hua ; Li, Peng ; Lu, Lin ; Wan, Ping ; Wang, Huaning ; Wang, Huiling ; Yang, Yongfeng ; Zhang, Hongxing ; Zhang, Dai ; Jiang, Tianzi ; Sui, Jing. / Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site FMRI data. In: EBioMedicine. 2019.
@article{0550cac62d9248d9aed0a1158d517961,
title = "Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site FMRI data",
abstract = "Background: Current fMRI-based classification approaches mostly use functional connectivity or spatial maps as input, instead of exploring the dynamic time courses directly, which does not leverage the full temporal information. Methods: Motivated by the ability of recurrent neural networks (RNN) in capturing dynamic information of time sequences, we propose a multi-scale RNN model, which enables classification between 558 schizophrenia and 542 healthy controls by using time courses of fMRI independent components (ICs) directly. To increase interpretability, we also propose a leave-one-IC-out looping strategy for estimating the top contributing ICs. Findings: Accuracies of 83·2{\%} and 80·2{\%} were obtained respectively for the multi-site pooling and leave-one-site-out transfer classification. Subsequently, dorsal striatum and cerebellum components contribute the top two group-discriminative time courses, which is true even when adopting different brain atlases to extract time series. Interpretation: This is the first attempt to apply a multi-scale RNN model directly on fMRI time courses for classification of mental disorders, and shows the potential for multi-scale RNN-based neuroimaging classifications. Fund: Natural Science Foundation of China, the Strategic Priority Research Program of the Chinese Academy of Sciences, National Institutes of Health Grants, National Science Foundation.",
keywords = "Cerebellum, Deep learning, fMRI, Multi-site classification, Recurrent neural network (RNN), Schizophrenia, Striatum",
author = "Weizheng Yan and V. Calhoun and Ming Song and Yue Cui and Hao Yan and Shengfeng Liu and Lingzhong Fan and Nianming Zuo and Zhengyi Yang and Kaibin Xu and Jun Yan and Luxian Lv and Jun Chen and Yunchun Chen and Hua Guo and Peng Li and Lin Lu and Ping Wan and Huaning Wang and Huiling Wang and Yongfeng Yang and Hongxing Zhang and Dai Zhang and Tianzi Jiang and Jing Sui",
year = "2019",
month = "1",
day = "1",
doi = "10.1016/j.ebiom.2019.08.023",
language = "English (US)",
journal = "EBioMedicine",
issn = "2352-3964",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Discriminating schizophrenia using recurrent neural network applied on time courses of multi-site FMRI data

AU - Yan, Weizheng

AU - Calhoun, V.

AU - Song, Ming

AU - Cui, Yue

AU - Yan, Hao

AU - Liu, Shengfeng

AU - Fan, Lingzhong

AU - Zuo, Nianming

AU - Yang, Zhengyi

AU - Xu, Kaibin

AU - Yan, Jun

AU - Lv, Luxian

AU - Chen, Jun

AU - Chen, Yunchun

AU - Guo, Hua

AU - Li, Peng

AU - Lu, Lin

AU - Wan, Ping

AU - Wang, Huaning

AU - Wang, Huiling

AU - Yang, Yongfeng

AU - Zhang, Hongxing

AU - Zhang, Dai

AU - Jiang, Tianzi

AU - Sui, Jing

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Background: Current fMRI-based classification approaches mostly use functional connectivity or spatial maps as input, instead of exploring the dynamic time courses directly, which does not leverage the full temporal information. Methods: Motivated by the ability of recurrent neural networks (RNN) in capturing dynamic information of time sequences, we propose a multi-scale RNN model, which enables classification between 558 schizophrenia and 542 healthy controls by using time courses of fMRI independent components (ICs) directly. To increase interpretability, we also propose a leave-one-IC-out looping strategy for estimating the top contributing ICs. Findings: Accuracies of 83·2% and 80·2% were obtained respectively for the multi-site pooling and leave-one-site-out transfer classification. Subsequently, dorsal striatum and cerebellum components contribute the top two group-discriminative time courses, which is true even when adopting different brain atlases to extract time series. Interpretation: This is the first attempt to apply a multi-scale RNN model directly on fMRI time courses for classification of mental disorders, and shows the potential for multi-scale RNN-based neuroimaging classifications. Fund: Natural Science Foundation of China, the Strategic Priority Research Program of the Chinese Academy of Sciences, National Institutes of Health Grants, National Science Foundation.

AB - Background: Current fMRI-based classification approaches mostly use functional connectivity or spatial maps as input, instead of exploring the dynamic time courses directly, which does not leverage the full temporal information. Methods: Motivated by the ability of recurrent neural networks (RNN) in capturing dynamic information of time sequences, we propose a multi-scale RNN model, which enables classification between 558 schizophrenia and 542 healthy controls by using time courses of fMRI independent components (ICs) directly. To increase interpretability, we also propose a leave-one-IC-out looping strategy for estimating the top contributing ICs. Findings: Accuracies of 83·2% and 80·2% were obtained respectively for the multi-site pooling and leave-one-site-out transfer classification. Subsequently, dorsal striatum and cerebellum components contribute the top two group-discriminative time courses, which is true even when adopting different brain atlases to extract time series. Interpretation: This is the first attempt to apply a multi-scale RNN model directly on fMRI time courses for classification of mental disorders, and shows the potential for multi-scale RNN-based neuroimaging classifications. Fund: Natural Science Foundation of China, the Strategic Priority Research Program of the Chinese Academy of Sciences, National Institutes of Health Grants, National Science Foundation.

KW - Cerebellum

KW - Deep learning

KW - fMRI

KW - Multi-site classification

KW - Recurrent neural network (RNN)

KW - Schizophrenia

KW - Striatum

UR - http://www.scopus.com/inward/record.url?scp=85070497886&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85070497886&partnerID=8YFLogxK

U2 - 10.1016/j.ebiom.2019.08.023

DO - 10.1016/j.ebiom.2019.08.023

M3 - Article

C2 - 31420302

AN - SCOPUS:85070497886

JO - EBioMedicine

JF - EBioMedicine

SN - 2352-3964

ER -