NeuroLens: A Holistic Visual Analytics System for Exploring Brain Networks Across Scales

Area of Science:

• Neuroscience
• Computer Science
• Medical Informatics

Background:

• Identifying biomarkers in human brain networks is crucial for early neurological disorder detection.
• Current visual analytics methods often analyze diffusion features of individual fiber bundles and regions separately, limiting comprehensive network investigation.
• This fragmented approach may hinder accurate representation of complex brain networks from multiple perspectives.

Purpose of the Study:

• To present NeuroLens, a visual analytics system for integrated exploration of multi-level human brain network information.
• To enhance the precise modeling of fiber bundles by incorporating joint geometric features.
• To enable multi-faceted comparisons between subjects and groups at region and brain levels.

Main Methods:

• NeuroLens integrates comprehensive information across fiber bundles, local regions, and entire brain networks.
• Enhanced fiber bundle features incorporate joint geometric distributions for precise modeling.
• Attention-based graph neural networks aggregate bundle information for region-level and brain-level representations.
• A large language model is leveraged for literature-based verification of findings during exploration.

Main Results:

• NeuroLens provides a multi-level and multi-faceted interface for comparative exploration of brain network data.
• The system enables detailed analysis of complex structures involving multiple bundles at the region level.
• Brain-level representations facilitate effective comparisons across subjects and groups.

Conclusions:

• NeuroLens offers an effective visual analytics approach for comprehensive brain network investigation and biomarker discovery.
• The integration of multi-level information and advanced machine learning models enhances the understanding of neurological disorders.
• Collaboration with domain experts validated the system's effectiveness in exploring and verifying findings.