State-space model with deep learning for functional dynamics estimation in resting-state fMRI

Research output: Contribution to journalArticle

83 Citations (Scopus)

Abstract

Studies on resting-state functional Magnetic Resonance Imaging (rs-fMRI) have shown that different brain regions still actively interact with each other while a subject is at rest, and such functional interaction is not stationary but changes over time. In terms of a large-scale brain network, in this paper, we focus on time-varying patterns of functional networks, i.e., functional dynamics, inherent in rs-fMRI, which is one of the emerging issues along with the network modelling. Specifically, we propose a novel methodological architecture that combines deep learning and state-space modelling, and apply it to rs-fMRI based Mild Cognitive Impairment (MCI) diagnosis. We first devise a Deep Auto-Encoder (DAE) to discover hierarchical non-linear functional relations among regions, by which we transform the regional features into an embedding space, whose bases are complex functional networks. Given the embedded functional features, we then use a Hidden Markov Model (HMM) to estimate dynamic characteristics of functional networks inherent in rs-fMRI via internal states, which are unobservable but can be inferred from observations statistically. By building a generative model with an HMM, we estimate the likelihood of the input features of rs-fMRI as belonging to the corresponding status, i.e., MCI or normal healthy control, based on which we identify the clinical label of a testing subject. In order to validate the effectiveness of the proposed method, we performed experiments on two different datasets and compared with state-of-the-art methods in the literature. We also analyzed the functional networks learned by DAE, estimated the functional connectivities by decoding hidden states in HMM, and investigated the estimated functional connectivities by means of a graph-theoretic approach.

Original languageEnglish
Pages (from-to)292-307
Number of pages16
JournalNeuroImage
Volume129
DOIs
Publication statusPublished - 2016 Apr 1

Fingerprint

Space Simulation
Magnetic Resonance Imaging
Learning
Brain

Keywords

  • Deep learning
  • Dynamic functional connectivity
  • Hidden Markov model
  • Mild cognitive impairment
  • Resting-state functional magnetic resonance imaging

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

State-space model with deep learning for functional dynamics estimation in resting-state fMRI. / Suk, Heung-Il; Wee, Chong Yaw; Lee, Seong Whan; Shen, Dinggang.

In: NeuroImage, Vol. 129, 01.04.2016, p. 292-307.

Research output: Contribution to journalArticle

@article{ea2491c20a4e4c4abbed55d9442fe9c8,
title = "State-space model with deep learning for functional dynamics estimation in resting-state fMRI",
abstract = "Studies on resting-state functional Magnetic Resonance Imaging (rs-fMRI) have shown that different brain regions still actively interact with each other while a subject is at rest, and such functional interaction is not stationary but changes over time. In terms of a large-scale brain network, in this paper, we focus on time-varying patterns of functional networks, i.e., functional dynamics, inherent in rs-fMRI, which is one of the emerging issues along with the network modelling. Specifically, we propose a novel methodological architecture that combines deep learning and state-space modelling, and apply it to rs-fMRI based Mild Cognitive Impairment (MCI) diagnosis. We first devise a Deep Auto-Encoder (DAE) to discover hierarchical non-linear functional relations among regions, by which we transform the regional features into an embedding space, whose bases are complex functional networks. Given the embedded functional features, we then use a Hidden Markov Model (HMM) to estimate dynamic characteristics of functional networks inherent in rs-fMRI via internal states, which are unobservable but can be inferred from observations statistically. By building a generative model with an HMM, we estimate the likelihood of the input features of rs-fMRI as belonging to the corresponding status, i.e., MCI or normal healthy control, based on which we identify the clinical label of a testing subject. In order to validate the effectiveness of the proposed method, we performed experiments on two different datasets and compared with state-of-the-art methods in the literature. We also analyzed the functional networks learned by DAE, estimated the functional connectivities by decoding hidden states in HMM, and investigated the estimated functional connectivities by means of a graph-theoretic approach.",
keywords = "Deep learning, Dynamic functional connectivity, Hidden Markov model, Mild cognitive impairment, Resting-state functional magnetic resonance imaging",
author = "Heung-Il Suk and Wee, {Chong Yaw} and Lee, {Seong Whan} and Dinggang Shen",
year = "2016",
month = "4",
day = "1",
doi = "10.1016/j.neuroimage.2016.01.005",
language = "English",
volume = "129",
pages = "292--307",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - State-space model with deep learning for functional dynamics estimation in resting-state fMRI

AU - Suk, Heung-Il

AU - Wee, Chong Yaw

AU - Lee, Seong Whan

AU - Shen, Dinggang

PY - 2016/4/1

Y1 - 2016/4/1

N2 - Studies on resting-state functional Magnetic Resonance Imaging (rs-fMRI) have shown that different brain regions still actively interact with each other while a subject is at rest, and such functional interaction is not stationary but changes over time. In terms of a large-scale brain network, in this paper, we focus on time-varying patterns of functional networks, i.e., functional dynamics, inherent in rs-fMRI, which is one of the emerging issues along with the network modelling. Specifically, we propose a novel methodological architecture that combines deep learning and state-space modelling, and apply it to rs-fMRI based Mild Cognitive Impairment (MCI) diagnosis. We first devise a Deep Auto-Encoder (DAE) to discover hierarchical non-linear functional relations among regions, by which we transform the regional features into an embedding space, whose bases are complex functional networks. Given the embedded functional features, we then use a Hidden Markov Model (HMM) to estimate dynamic characteristics of functional networks inherent in rs-fMRI via internal states, which are unobservable but can be inferred from observations statistically. By building a generative model with an HMM, we estimate the likelihood of the input features of rs-fMRI as belonging to the corresponding status, i.e., MCI or normal healthy control, based on which we identify the clinical label of a testing subject. In order to validate the effectiveness of the proposed method, we performed experiments on two different datasets and compared with state-of-the-art methods in the literature. We also analyzed the functional networks learned by DAE, estimated the functional connectivities by decoding hidden states in HMM, and investigated the estimated functional connectivities by means of a graph-theoretic approach.

AB - Studies on resting-state functional Magnetic Resonance Imaging (rs-fMRI) have shown that different brain regions still actively interact with each other while a subject is at rest, and such functional interaction is not stationary but changes over time. In terms of a large-scale brain network, in this paper, we focus on time-varying patterns of functional networks, i.e., functional dynamics, inherent in rs-fMRI, which is one of the emerging issues along with the network modelling. Specifically, we propose a novel methodological architecture that combines deep learning and state-space modelling, and apply it to rs-fMRI based Mild Cognitive Impairment (MCI) diagnosis. We first devise a Deep Auto-Encoder (DAE) to discover hierarchical non-linear functional relations among regions, by which we transform the regional features into an embedding space, whose bases are complex functional networks. Given the embedded functional features, we then use a Hidden Markov Model (HMM) to estimate dynamic characteristics of functional networks inherent in rs-fMRI via internal states, which are unobservable but can be inferred from observations statistically. By building a generative model with an HMM, we estimate the likelihood of the input features of rs-fMRI as belonging to the corresponding status, i.e., MCI or normal healthy control, based on which we identify the clinical label of a testing subject. In order to validate the effectiveness of the proposed method, we performed experiments on two different datasets and compared with state-of-the-art methods in the literature. We also analyzed the functional networks learned by DAE, estimated the functional connectivities by decoding hidden states in HMM, and investigated the estimated functional connectivities by means of a graph-theoretic approach.

KW - Deep learning

KW - Dynamic functional connectivity

KW - Hidden Markov model

KW - Mild cognitive impairment

KW - Resting-state functional magnetic resonance imaging

UR - http://www.scopus.com/inward/record.url?scp=84957052106&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84957052106&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2016.01.005

DO - 10.1016/j.neuroimage.2016.01.005

M3 - Article

C2 - 26774612

AN - SCOPUS:84957052106

VL - 129

SP - 292

EP - 307

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

ER -