Schizophrenia Classification Using fMRI Data Based on a Multiple Feature Image Capsule Network Ensemble

Abstract

Automatic diagnosis and classification of schizophrenia based on functional magnetic resonance imaging (fMRI) data have attracted increasing attention in recent years. Most previous studies abstracted highly compressed functional features from the view of brain science and fed them into shallow classifiers for this purpose. However, their classification performance in practical applications is unstable and unsatisfactory. As an acute psychotic disorder, schizophrenia shows functional complexity in fMRI data. Therefore, additional features and deep classification methods are needed to improve classification performance. In this study, we propose a multiple feature image capsule network ensemble approach for schizophrenia classification. The proposed approach proceeds in three steps: 1) extracting multiple image features from the perspective of linear sparse representation, nonlinear multiple kernel representation, and function connection of brain areas respectively; 2) feeding these image features into three specially designed independent capsule networks for classification; 3) obtaining the final results by fusing the outputs of these three deep capsule network using a ensemble approach. To further improve the classification performance, we design a optimization model of maximizing the square of correlation coefficients and propose a weighted ensemble technology based on this model, which is mathematically proved to be solved as a eigenvalue decomposition problem in certain case. Finally, the proposed approach is implemented and evaluated on the schizophrenia fMRI dataset from COBRE, UCLA and WUSTL. From the experimental results, we conclude that the proposed method outperforms some current methods and further improves the accuracy of schizophrenia classification. INDEX TERMS Schizophrenia classification, multiple features extraction, deep capsule network, classifier ensemble.

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: Systematic review and meta-analysis

“…This input was fed into an ensemble of extreme learning machines (ELM) for classification. Yang et al (2019) also used an ensemble of networks to classify an input of multiple image features (including functional connectivity, nonlinear multiple kernel learning and sparse dictionary learning) and obtained 82.8% accuracy on 3 datasets including COBRE. Kim et al (2016) used a deep learning technique to select features that could be passed on to a standard machine learning model: they used a stacked auto encoder on timeseries from the AAL atlas to encode a latent feature vector that was fed into an SVM to obtain an accuracy of 86.5%.…”

“…It is, for instance, likely that many included studies optimized parameters for their DL model but did not optimize parameters for their comparative SML model. The difference with and without optimisation can be large: In a study of ( Yang et al, 2019 , Yang et al, 2019 ) a grid search method was deployed to find the optimal parameters for SVM. They obtained a cross validated accuracy of 71.98% on the entire ABIDE I, whereas without optimisation ( Heinsfeld et al, 2018 ) report an accuracy of 65% using SVM on the ABIDE I.…”

Altered Dynamic Functional Connectivity of Cuneus in Schizophrenia Patients: A Resting-State fMRI Study

“…This is one of the few dFC studies of schizophrenia based on the COBRE public dataset [79][80][81][82][83]. In this study, we investigated differences in functional connectivity (FC) dynamics at rest between schizophrenia (SZ) patients and healthy controls (HC).…”

DeepReac+: deep active learning for quantitative modeling of organic chemical reactions

“…Furthermore, to avoid the loss of reaction information and alleviate the need for training data, we introduce the Capsule module to aggregate the feature vectors of the reaction components after message passing. Unlike most deep learning architectures, capsule networks 73 have achieved outstanding performance for small-sample learning in the fields of life sciences [74][75][76] . As the core element, the capsule is a new type of neuron that encapsulates more information than common pooling operations by computing a small vector of highly informative outputs rather than taking only a scalar output.…”