Extensive Spherical Region Enlargement with Isotropic Deformation for Sulcal Labeling

The cortical sulci of the posterior medial cortex play a crucial role in cognitive and behavioral functions. However, automatic labeling of these sulci is challenging because of class imbalance (i.e., only a small portion of the cortex are labeled) and neuroanatomical variability. To tackle these challenges, we propose a novel spherical deformation-based method that can automatically enlarge the sulcal regions of interest (ROIs) while contracting other regions. This allows the proposed method to capture detailed sulcal patterns more effectively. Specifically, we design a learnable isotropic deformation field in the spherical domain while ensuring no self-intersections and maintaining an invertible mapping. Moreover, such an addition minimally impacts the model complexity of underlying spherical convolutional neural network. In the qualitative and quantitative evaluations, we show that the proposed method outperforms baseline approaches particularly in ROIs with high anatomical variability. The most notable improvement of 15.51%p is observed in the smallest sulcal ROI.