Classification of schizophrenia patients using a graph convolutional network: A combined functional MRI and connectomics analysis

Chen, Xiaoyi; Zhou, Jing; Ke, Pengfei; Huang, Jiayuan; Xiong, Dongsheng; Huang, Yuanyuan; Ma, Guolin; Ning, Yuping; Wu, Fengchun; Wu, Kai

doi:10.1016/j.bspc.2022.104293

Cited by 13 publications

(10 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There were three improvements in this study, compared with our previous study (Chen et al, 2023). First, previous studies have Frontiers in Neuroscience frontiersin.org indicated that multimodal MRI was more useful than that singlemodal MRI data in the discriminative analyses of SZ patients (Wu et al, 2018;Sebenius et al, 2021;Zang et al, 2021;Wang et al, 2022).…”

Section: Discussionmentioning

confidence: 79%

“…Considering that the sample size was not large and different training/testing splits lead to dramatically different rankings of models (Shchur et al, 2019;Flint et al, 2021;Sun et al, 2023), we randomly split training and testing datasets 10 times, repeated the experiments and calculated the averaged performance for each type of brain graph. It is worth emphasizing that in our recent fMRI study (Chen et al, 2023), we randomly split training and testing datasets 200 times to get the average performance and we found that the performance obtained on a single data split could be fragile and misleading, confirming the necessity of a multiple data split evaluation strategies. However, there are many models involved in this multimodal (18 types of brain graphs) study, and it would cause great computational complexity if all models repeated 200 times.…”

Section: Methodsmentioning

confidence: 91%

“…And 140 SZ patients included 61 first-episode medication-naïve SZ (FESZ) patients and 71 medicated chronic SZ (CSZ) patients. The age of the subjects was between 18–60 years and the inclusion and exclusion criteria of subjects were the same as those in our previous studies ( Wu et al, 2018 ; Kong et al, 2021 ; Huang et al, 2022 ; Chen et al, 2023 ). The SZ patients were diagnosed by veteran psychiatrists according to the structured clinical interview complying with the criteria of the Diagnostic and Statistical Manual of Mental Disorders-IV-Text Revision (DSM-IV-TR) (SCID).…”

Section: Methodsmentioning

confidence: 99%

“…GCN, an extension of CNN in the non-European domain, can simultaneously learn node characteristic information and graph structure information. The architecture of the GCN classifier ( Figure 2 ) were the same with our previous study ( Chen et al, 2023 ) that adopted the GCN framework introduced in Lee et al (2019) . The GCN classifier implemented a hierarchical pooling architecture ( Cangea et al, 2018 ; Lee et al, 2019 ) and comprised three blocks consisting of a graph convolutional layer, a graph pooling layer and a readout layer.…”

Section: Methodsmentioning

confidence: 99%

“…In our previous research, we proposed an integrated analysis of functional MRI and connectomics that considered characteristics of brain regions of interest (ROIs) and the functional connectivity between ROIs ( Chen et al, 2023 ). Specifically, we obtained an average accuracy of 92.7% based on the GCN method, which outperformed the methods that focused on features of single ROIs and the methods that only based on connectomics analysis.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Discriminative analysis of schizophrenia patients using graph convolutional networks: A combined multimodal MRI and connectomics analysis

Chen

Ke²,

Huang³

et al. 2023

Front. Neurosci.

Self Cite

View full text Add to dashboard Cite

IntroductionRecent studies in human brain connectomics with multimodal magnetic resonance imaging (MRI) data have widely reported abnormalities in brain structure, function and connectivity associated with schizophrenia (SZ). However, most previous discriminative studies of SZ patients were based on MRI features of brain regions, ignoring the complex relationships within brain networks.MethodsWe applied a graph convolutional network (GCN) to discriminating SZ patients using the features of brain region and connectivity derived from a combined multimodal MRI and connectomics analysis. Structural magnetic resonance imaging (sMRI) and resting-state functional magnetic resonance imaging (rs-fMRI) data were acquired from 140 SZ patients and 205 normal controls. Eighteen types of brain graphs were constructed for each subject using 3 types of node features, 3 types of edge features, and 2 brain atlases. We investigated the performance of 18 brain graphs and used the TopK pooling layers to highlight salient brain regions (nodes in the graph).ResultsThe GCN model, which used functional connectivity as edge features and multimodal features (sMRI + fMRI) of brain regions as node features, obtained the highest average accuracy of 95.8%, and outperformed other existing classification studies in SZ patients. In the explainability analysis, we reported that the top 10 salient brain regions, predominantly distributed in the prefrontal and occipital cortices, were mainly involved in the systems of emotion and visual processing.DiscussionOur findings demonstrated that GCN with a combined multimodal MRI and connectomics analysis can effectively improve the classification of SZ at an individual level, indicating a promising direction for the diagnosis of SZ patients. The code is available at https://github.com/CXY-scut/GCN-SZ.git.

show abstract