Graph Matching Based Connectomic Biomarker with Learning for Brain Disorders.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Network Science

Background:

• Neuroimaging techniques like diffusion MRI and functional MRI allow brain connectome analysis.
• Connectomic biomarkers aid in disease diagnosis and prognosis, often using graph theory or correlation metrics.
• Current methods have limitations in simultaneously assessing individual differences, disease patterns, and network structure.

Purpose of the Study:

• To develop a novel graph matching method for quantifying connectomic similarity.
• To create a subject-specific biomarker for disease assessment at the functional systems level.
• To evaluate the method's performance and clinical relevance in traumatic brain injury (TBI).

Main Methods:

• Proposed a graph matching technique to quantify connectomic similarity.
• Trained the method on functional systems level data for disease-specific patterns.
• Validated the approach on a TBI patient dataset, comparing it with existing similarity measures.
• Assessed functional system vulnerability and correlation with clinical scores.

Main Results:

• The proposed graph matching method demonstrated superior separation between TBI patients and controls compared to standard measures.
• The method successfully identified disease-specific patterns and network vulnerabilities.
• The derived similarity score showed correlation with clinical TBI severity scores.

Conclusions:

• The graph matching based connectomic similarity measure serves as a promising subject-specific biomarker for diseases like TBI.
• This novel approach addresses limitations of existing methods by integrating network structure and disease-specific information.
• The findings highlight the potential for improved diagnostic and prognostic tools in neurological disorders.