Anatomically constrained squeeze-and-excitation graph attention network for cortical surface parcellation

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate parcellation of the cerebral cortex is crucial for understanding brain organization and neurological diseases.
• Current methods often rely on computationally intensive geometric simplification, losing intrinsic cortical surface information.

Purpose of the Study:

• To develop an end-to-end brain cortical surface parcellation method that operates directly on the original cortical surface manifold.
• To improve the accuracy and efficiency of cortical surface parcellation compared to existing techniques.

Main Methods:

• Proposed an anatomically constrained squeeze-and-excitation graph attention network (ASEGAT) for direct cortical surface parcellation.
• Incorporated self-attention and head attention mechanisms within graph attention modules.
• Introduced an anatomic constraint loss to enforce regional adjacency relationships for improved labeling consistency.

Main Results:

• ASEGAT achieved state-of-the-art performance on a public dataset of 100 manually labeled brain surfaces.
• The model obtained an accuracy of 90.65% and a Dice score of 89.00%.
• Demonstrated superior performance compared to several advanced parcellation methods.

Conclusions:

• ASEGAT provides an effective and efficient end-to-end solution for brain cortical surface parcellation.
• The method leverages intrinsic cortical geometry and anatomical priors for enhanced accuracy.
• This approach holds significant potential for advancing research in neuroscience and clinical applications.