Fusion Learning from Dynamic Functional Connectivity: Combining the Amplitude and Phase of fMRI Signals to Identify Brain Disorders

2026-03-14 · Source: cs.AI updates on arXiv.org · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Mathematics & Computational Sciences · Depth: Expert, short

Summary

A new multi-scale fusion learning framework, MSFL, has been introduced to enhance the detection of brain disorders using resting-state functional magnetic resonance imaging (fMRI) data. This approach integrates both amplitude and phase information from fMRI signals, moving beyond the traditional sliding window correlation (SWC) method which primarily captures amplitude correlations. MSFL specifically combines SWC features with phase synchronization (PS) features, which measure phase coherence within dynamic functional connectivity (dFC). The framework's effectiveness was validated by classifying autism spectrum disorder and major depressive disorder using the publicly available ABIDE I and REST-meta-MDD datasets, respectively. Results demonstrate that MSFL significantly surpasses existing comparative models in performance. Furthermore, a SHAP-based model explanation analysis confirmed that both SWC and PS derived dFC features are crucial contributors to its diagnostic capabilities.

Key takeaway

For AI Scientists and Research Scientists developing diagnostic tools for neurological conditions, integrating both amplitude and phase information from fMRI signals significantly improves classification accuracy. You should consider adopting multi-scale fusion learning frameworks like MSFL to leverage these complementary dynamic functional connectivity features. This approach, validated on autism spectrum disorder and major depressive disorder, offers a more robust method for identifying brain disorders than amplitude-only analyses.

Key insights

Combining fMRI signal amplitude and phase via fusion learning improves brain disorder detection.

Principles

• Dynamic functional connectivity benefits from multi-modal feature integration.
• Both amplitude and phase information contribute to dFC analysis.
• Model interpretability confirms feature importance in diagnostic tasks.

Method

The MSFL framework integrates sliding window correlation (SWC) for amplitude and phase synchronization (PS) for phase coherence into a multi-scale fusion learning model to classify brain disorders.

In practice

• Apply SWC and PS features for enhanced dFC analysis.
• Use SHAP for explaining multi-feature diagnostic models.
• Evaluate fusion models on public fMRI datasets like ABIDE I.

Topics

• Dynamic Functional Connectivity
• fMRI Signal Analysis
• Fusion Learning
• Autism Spectrum Disorder
• Major Depressive Disorder
• SHAP Framework

Best for: AI Scientist, Research Scientist

Related on AIssential

• End-to-End Machine Learning for Depressive State Classification via EEG and fNIRS — Biological signal analysis via EEG and fNIRS offers an objective framework for depressive state classification.
• Learning Multi-Scale Hypergraph for High-Order Brain Connectivity Analysis — MuHL dynamically learns multi-scale, high-order brain connectivity for improved neurodegenerative disease classification.
• Isolating Nonlinear Independent Sources in fMRI with $\beta$-TCVAE Models — $\beta$-TCVAE effectively disentangles nonlinear fMRI data into biologically meaningful and spatially coherent brain networks.
• Stringology-Based Motif Discovery from EEG Signals: an ADHD Case Study — Stringology-based motif discovery offers a novel, complementary approach to analyze EEG dynamics in neurodevelopmental disorders like ADHD.
• Continual Learning for fMRI-Based Brain Disorder Diagnosis via Functional Connectivity Matrices Generative Replay — A continual learning framework uses generative replay and knowledge distillation to diagnose brain disorders from sequential fMRI data.
• Learning Under Extreme Data Scarcity: Subject-Level Evaluation of Lightweight CNNs for fMRI-Based Prodromal Parkinsons Detection — Subject-level data splitting and model capacity are crucial for reliable deep learning evaluation in extreme data scarcity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.