Evaluating topological and graph-theoretical approaches to extract complex multimodal brain connectivity patterns in...

Area of Science:

• Neuroscience
• Data Science
• Medical Imaging

Background:

• Brain networks derived from MRI are crucial for understanding brain organization.
• Graph-theoretical metrics analyze network properties like efficiency and integration.
• Topological data analysis, including Betti curves, offers novel methods for capturing complex network patterns.

Purpose of the Study:

• To compare Betti curves and graph-theoretical metrics for feature extraction in neurodegenerative disease.
• To evaluate these methods in distinguishing people with multiple sclerosis (PwMS) from healthy volunteers (HV).
• To assess the impact of multimodal data and multilayer architectures on classification performance.

Main Methods:

• Features were extracted from structural connectivity, gray matter morphology, and resting-state functional networks.
• Both single-layer and multilayer graph architectures were utilized.
• Betti curves and graph-theoretical metrics were compared for their discriminative power.

Main Results:

• Betti curve-derived features generally showed superior performance compared to graph-theoretical metrics.
• Multimodal data integration and multilayer graph architectures enhanced the representation of brain alterations.
• Improved classification accuracy was observed when using combined features and advanced network structures.

Conclusions:

• Topological features, specifically Betti curves, show significant potential for neurodegenerative disease analysis.
• Multimodal data integration and multilayer network analysis are effective strategies for improving diagnostic capabilities.
• These findings support the use of advanced topological and network approaches for understanding and diagnosing conditions like MS.