GPU-accelerated cortical surface registration for brain imaging
This project optimizes the CAT-Surface and SimNIBS registration pipeline by moving heavy spherical inflation and surface warping to the GPU. It uses custom Triton kernels to maintain topology awareness while drastically reducing processing time. It is designed for neuroscientists and researchers working with high-resolution cortical meshes who need faster, reference-accurate registration.
View on GitHub →ayakacxy/cat-surface-gpu