MAC-Net: A Multi-Task Deep Learning Framework for Modeling Cognitive Function From Task-Based fMRI

2026-09-27

AI Summary

This study addresses the susceptibility of neural features to demographic and scanner-related confounds in fMRI-based cognitive prediction by proposing a covariate-aware multitask deep learning framework. Through a late fusion mechanism, the approach isolates fMRI features from covariates, preventing dominant variables from suppressing high-dimensional clinical representation learning. Furthermore, it integrates a multitask activation contrastive network, family-aware cross-validation, and integrated gradient attribution techniques to enhance individual-level predictive robustness. Evaluated on the ABC dataset, the proposed method significantly outperforms baseline approaches while precisely localizing working memory-associated brain regions. Ultimately, this work establishes a reproducible, clinically translatable framework for individualized cognitive assessment.

Abstract

Objective cognitive assessment from neural signals supports neurorehabilitation, but individual-level prediction from task-based fMRI (tfMRI) remains difficult because neural features coexist with substantial demographic and scanner-related variation. We present the Multi-task Activation and Contrast Network (MAC-Net), a covariate-aware deep learning framework for modeling individual cognitive function from regional tfMRI. By isolating tfMRI features into a dedicated neural pathway and restricting participant variables to a terminal late-fusion pathway, MAC-Net prevents dominant covariates from suppressing high-dimensional clinical representations during feature learning. Evaluating baseline data from 6,500 Adolescent Brain Cognitive Development Study participants under family-aware cross-validation, MAC-Net was benchmarked against linear models, random forests, and alternative deep architectures. The N-back plus Monetary Incentive Delay configuration achieved $R^{2}$ values of 0.174, 0.238, and 0.277 for fluid, crystallized, and total cognition, outperforming covariate-only baselines (0.178) and alternative deep models (0.217). N-back was the most informative paradigm, whereas incorporating the Stop Signal Task marginally degraded performance. Feature attributions via Integrated Gradients, DeepLIFT, and Input Gradient were highly concordant, localizing working-memory-related frontal, parietal, and cingulate regions. These findings demonstrate that covariate-aware multi-task modeling yields reproducible cognitive-function estimations, establishing a robust neural engineering framework for clinical translation.

Problem

• cognitive function prediction
• task-based fMRI
• individual-level modeling
• covariate variation

Innovation

• Multi-task deep learning
• Covariate-aware late fusion
• Task-based fMRI
• Cognitive function prediction
• Feature attribution