Hierarchical Clustering Diffusion Model for fMRI Functional Connectivity to Enhance Autism Spectrum Disorder Diagnosis

Fatima Ez-Zahraa Bazay, Djaafer Gueddou, Mourad Oussalah, Rachid Jennane, Ahmed Drissi El Maliani

Keywords: fMRI Functional Connectivity, ASD, Data Augmentation, Diffusion Models, Hierarchical Clustering

TL;DR: Hierarchical clustering-enhanced diffusion model improves autism diagnosis through synthetic fMRI data augmentation.

Abstract:

Functional magnetic resonance imaging (fMRI) data, particularly functional connectivity matrices, are crucial for studying brain disorders like Autism Spectrum Disorder (ASD). However, data scarcity often limits the performance of diagnostic models. We address this challenge by leveraging generative diffusion models for data augmentation. We introduce a novel transformer-based latent diffusion model, the Hierarchical Clustering Connective Diffusion Unit (HC-CDU), designed to synthesize realistic fMRI functional connectivity matrices. Our models effectively generate high-fidelity connectivity patterns, demonstrating an improvement of up to 3.61% in MAE reduction. In classification tasks on the ABIDE-I dataset, HC-CDU with ×1 augmentation demonstrated significant improvement, with AUC enhancing
by up to 4.29% over baseline, showcasing enhanced discriminative power.

Submission Number: 26