Autism Spectrum Disorder Classification with Interpretability in Children based on Structural MRI Features Extracted using Contrastive Variational Autoencoder

TL;DR

This work tackles early ASD screening in children by learning disentangled s-MRI features with a contrastive variational autoencoder, separating ASD-specific information from common brain patterns. A Random Forest classifier uses these features to achieve high accuracy, while a transfer learning approach from ABIDE-I (MABIDE) helps when data are scarce. The study also provides neuroanatomical interpretability via representational similarity analysis, identifying supramarginal and inferiortemporal regions as potential ASD biomarkers. Despite strong results, the authors acknowledge the small age-range dataset and propose expanding early-age data and refining the feature space in future work.

Abstract

Autism spectrum disorder (ASD) is a highly disabling mental disease that brings significant impairments of social interaction ability to the patients, making early screening and intervention of ASD critical. With the development of the machine learning and neuroimaging technology, extensive research has been conducted on machine classification of ASD based on structural Magnetic Resonance Imaging (s-MRI). However, most studies involve with datasets where participants' age are above 5 and lack interpretability. In this paper, we propose a machine learning method for ASD classification in children with age range from 0.92 to 4.83 years, based on s-MRI features extracted using contrastive variational autoencoder (CVAE). 78 s-MRIs, collected from Shenzhen Children's Hospital, are used for training CVAE, which consists of both ASD-specific feature channel and common shared feature channel. The ASD participants represented by ASD-specific features can be easily discriminated from TC participants represented by the common shared features. In case of degraded predictive accuracy when data size is extremely small, a transfer learning strategy is proposed here as a potential solution. Finally, we conduct neuroanatomical interpretation based on the correlation between s-MRI features extracted from CVAE and surface area of different cortical regions, which discloses potential biomarkers that could help target treatments of ASD in the future.

Figures (7)

• Figure 1: Several ASD and typical control (TC) examples used in our work.
• Figure 2: Overall framework of our proposed algorithm for ASD classification and interpretability study.
• Figure 3: Contrastive variational autoencoder for extracting ASD-specific features and common shared features.
• Figure 4: Classification results and the visualization analysis of the features. a) accuracy of classifying ASD with ASD participants being represented either by ASD-specific features or common shared features; b) relationship between classification accuracy and sample size, the accuracy of classifying ASD was measured based on ASD participants being represented by ASD-specific features; visualizing ASD participants and TC participants in 2-D feature space with TC participants being represented by common share features while ASD participants being represented by either c) common shared features, or d) ASD-specific features.
• Figure 5: The schematic diagram of transfer learning study of ABIDE-I.