Multi-Band Brain Network Analysis for Functional Neuroimaging Biomarker Identification.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomarker Discovery

Background:

• Functional connectivity networks (FCNs) are key biomarkers for neuro-diseases but are limited by mixed brain signal frequencies.
• Single FCNs often lack the power to detect subtle, disease-specific functional patterns.

Purpose of the Study:

• To develop a novel framework for semi-supervised, personalized diagnosis of neuro-diseases using multi-band functional connectivity.
• To enhance diagnostic power by integrating information from various brain signal frequencies.

Main Methods:

• Decomposition of Blood Oxygenation Level Dependent (BOLD) signals into frequency bands using discrete wavelet transform.
• A parameter-free multi-band fusion model to align and fuse FCNs from different frequency bands into sparse FCNs.
• Utilizing l1-SVM for joint brain region selection and disease classification.

Main Results:

• The proposed framework demonstrated superior classification performance compared to state-of-the-art methods for Fronto-Temporal Dementia (FTD), Obsessive-Compulsive Disorder (OCD), and Alzheimer's Disease (AD).
• The method effectively identified disease-relevant brain regions, enhancing the interpretability of the diagnostic biomarkers.
• Generated sparse FCNs effectively captured individual subject's functional patterns while maintaining group-level similarities.

Conclusions:

• The multi-band functional connectivity analysis framework offers a robust and effective approach for computer-assisted diagnosis of neuro-diseases.
• This method overcomes the limitations of single-frequency analysis, providing more sensitive and specific diagnostic biomarkers.
• The framework's ability to perform joint feature learning and personalized diagnosis holds significant promise for clinical applications.