Motivation: The cerebellum is vital for motor, sensory, and cognitive functions, yet analyzing its development across the lifespan remains challenging due to its complex structure and age-related imaging variability. No reliable methods currently support comprehensive lifespan analysis. Goal(s): To develop a pipeline for cerebellar tissue segmentation and cortical surface reconstruction for lifespan-wide analysis. Approach: We trained an age-robust segmentation model on T1-weighted cerebellar images with extensive data augmentation, followed by a surface reconstruction model trained on segmentation-derived pseudo-ground truth to deform a spherical template. Results: Our method achieves reliable cerebellar segmentation and surface reconstruction, enabling efficient lifespan-wide morphology analysis. Impact: This work addresses the gap in lifespan cerebellar analysis by providing a reliable solution for segmentation and surface reconstruction to advance our understanding of cerebellar development.
Lin et al. (Tue,) studied this question.