GAGM: Geometry-aware graph matching framework for weakly supervised gyral hinge correspondence.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Precise alignment of inter-subject brain landmarks like gyral hinges (GHs) is crucial for understanding brain anatomy-function relationships.
• Current methods rely on laborious point-to-point labeling, which is time-consuming for the numerous GHs in the brain.

Purpose of the Study:

• To develop a weakly supervised framework for accurate gyral hinge correspondence using only brain prior information.
• To overcome the limitations of manual labeling in establishing cross-subject brain landmark correspondences.

Main Methods:

• Proposed a Geometry-Aware Graph Matching (GAGM) framework.
• Introduced a Shape-Aware Graph Establishment (SAGE) module to model GH geometry and spatial relations.
• Developed a Region-Aware Graph Matching (RAGM) module for multi-scale matching, ensuring intra-region consistency and inter-region variability.

Main Results:

• The GAGM framework achieved accurate gyral hinge matching.
• Demonstrated superior performance compared to state-of-the-art methods on HCP and CHCP datasets.
• Successfully implemented weakly supervised correspondence based on prior brain information.

Conclusions:

• GAGM offers an efficient and accurate solution for inter-subject brain landmark alignment.
• The proposed SAGE and RAGM modules effectively address challenges in GH correspondence.
• This method advances the study of brain anatomy-function relationships and brain mechanisms.