AI-Driven Prediction of Communication Score from Functional Mri with Neurocircuit Characterization in Autism

Document Type

Conference Proceeding

Source of Publication

Proceedings International Symposium on Biomedical Imaging

Publication Date

4-8-2026

Abstract

Functional magnetic resonance imaging (fMRI) provides a powerful view into how distributed brain regions interact to support human communication and social interaction. Building on this capability, this study introduces an AI-based framework that uses resting-state fMRI to directly predict continuous SRS Communication scores, offering a quantitative measure of communication-related symptom severity in autism. Preprocessing of fMRI data from 596 subjects (276 ASD, 320 TD) in the ABIDE-II dataset was performed using fMRIPrep and FastSurfer to normalize anatomical and functional data. Dynamic functional connectivity was then estimated using a Gaussian sliding window approach, and feature selection (Spearman correlation and mutual information, top 30-40%) along with PCAbased reduction to keep the most behaviorally significant components (to keep between 90-98% of the variance). Multiple regression models were evaluated, including linear, kernel-based, tree ensemble, Bayesian, and deep/foundation models. The best performance in predicting SRS Communication scores was reached by the Bayesian Ridge regressor (MAE =7.58 ± 0.30, RMSE =9.59 ± 0.48, R 2= 0.72 ± 0.05, ρ=0.84 ± 0.02). Analysis of feature importance revealed that significant predictive connectivities were primarily located within and between the default mode, language, and salience networks, reflecting their central role in social communication processing.

ISBN

[9798331577636]

ISSN

1945-7928

Publisher

IEEE

Volume

2026-April

Disciplines

Computer Sciences | Medicine and Health Sciences

Keywords

ABIDE, Autism, DL, Dynamic Functional Connectivity (dFC), fMRI, ML, PCA, SRS

Scopus ID

105041664000

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS