Explainable Deep Learning of Resting-State Functional Connectomes Reveals Network Biomarkers of Adolescent Intelligence

2026-09-27

AI Summary

This study addresses the limited interpretability of deep learning in brain connectome-based intelligence prediction by proposing an interpretable framework based on a sparse projection residual network. Using the ABCD dataset to predict adolescent intelligence, the method integrates gradient-based attribution, SHAP, and occlusion techniques for multidimensional feature analysis. This reveals a dual-layer functional architecture wherein local computational hubs and long-range relay pathways synergistically drive intelligence. Results demonstrate a 6%–9% improvement in predictive performance, with correlation coefficients reaching 0.44–0.58. Furthermore, the framework generates brain functional maps with clinical reference value, establishing a novel paradigm that combines high accuracy with strong interpretability for elucidating the neural mechanisms underlying intelligence.

Abstract

Mapping resting-state brain organization to individual differences in cognitive ability remains a major challenge in population neuroinformatics. Although deep learning enables flexible modeling of brain connectivity, limited interpretability restricts its scientific and clinical utility. To address this objective, we developed an explainable deep learning framework based on sparse projected residual networks to predict fluid, crystallized, and total intelligence from resting-state functional magnetic resonance imaging in 5,285 participants from the Adolescent Brain Cognitive Development study. We incorporated three complementary explainability methods (Integrated Gradients, Gradient Shapley Additive Explanations, and Occlusion) to interpret model behavior. The framework outperformed existing approaches, achieving Pearson correlations of 0.44, 0.58, and 0.56 for fluid, crystallized, and total intelligence, respectively, corresponding to predictive improvements of 6 to 9 percent. All three explainability methods produced near-identical feature rankings (pairwise rank correlations greater than 0.99). Consensus maps revealed a dual-layered functional architecture where primary predictive hubs localized within canonical systems, while the strongest global predictive pathways frequently bypassed these hubs through distributed, long-range relay connections. These findings suggest that intelligence emerges from the interaction between localized computational hubs and distributed communication pathways. Ultimately, these normative network architectures provide clinical reference maps to detect individual deviations, supporting earlier diagnosis, cognitive subtype stratification, and treatment monitoring in atypical neurodevelopment.

Problem

• Resting-State Functional Connectomes
• Adolescent Intelligence
• Explainable Deep Learning
• Population Neuroinformatics
• Network Biomarkers

Innovation

• Explainable Deep Learning
• Sparse Projected Residual Networks
• Resting-State Functional Connectomes
• Network Biomarkers
• Adolescent Intelligence