ASDFormer: A Transformer with Mixtures of Pooling-Classifier Experts for Robust Autism Diagnosis and Biomarker Discovery

2025-08-19

AI Summary

This study addresses the dual challenges of low diagnostic accuracy and insufficient neurobiological interpretability in autism spectrum disorder (ASD) classification. We propose a Transformer-based Mixture-of-Experts (MoE) model that jointly models region-of-interest (ROI)-level functional connectivity and incorporates dynamic attention mechanisms. The model employs multiple expert branches to adaptively attend to distinct brain regions and connectivity patterns, while a hierarchical mixed pooling strategy jointly optimizes classification performance and biomarker interpretability. Evaluated on the ABIDE dataset, our method achieves state-of-the-art diagnostic accuracy (>78%) and robustly identifies aberrant functional connections within canonical networks—including the default mode and frontoparietal control networks—yielding interpretable neuroimaging biomarkers for ASD. To our knowledge, this is the first work to integrate the MoE architecture with dynamic fMRI connectivity modeling, thereby simultaneously enhancing discriminative power and neuroscientific insight.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition marked by disruptions in brain connectivity. Functional MRI (fMRI) offers a non-invasive window into large-scale neural dynamics by measuring blood-oxygen-level-dependent (BOLD) signals across the brain. These signals can be modeled as interactions among Regions of Interest (ROIs), which are grouped into functional communities based on their underlying roles in brain function. Emerging evidence suggests that connectivity patterns within and between these communities are particularly sensitive to ASD-related alterations. Effectively capturing these patterns and identifying interactions that deviate from typical development is essential for improving ASD diagnosis and enabling biomarker discovery. In this work, we introduce ASDFormer, a Transformer-based architecture that incorporates a Mixture of Pooling-Classifier Experts (MoE) to capture neural signatures associated with ASD. By integrating multiple specialized expert branches with attention mechanisms, ASDFormer adaptively emphasizes different brain regions and connectivity patterns relevant to autism. This enables both improved classification performance and more interpretable identification of disorder-related biomarkers. Applied to the ABIDE dataset, ASDFormer achieves state-of-the-art diagnostic accuracy and reveals robust insights into functional connectivity disruptions linked to ASD, highlighting its potential as a tool for biomarker discovery.

Problem

• Robust autism diagnosis using fMRI connectivity patterns
• Identifying ASD biomarkers from brain network alterations
• Capturing neural signatures with Transformer and expert modules

Innovation

• Transformer-based architecture with Mixture of Experts
• Multiple specialized expert branches with attention mechanisms
• Adaptively emphasizes brain regions and connectivity patterns