Advances in medical imaging technique make it possible to study shape variations of neuroanatomical structures in vivo, which has been proved useful in the study of neuropathology and neurodevelopment. In this paper, we propose the use of spherical wavelet transformation to extract shape features, as it can characterize the underlying functions in a local fashion in both space and frequency, in contrast to spherical harmonics that have a noncompact basis set. The extracted shape features can be used to statistically detect and visualize group shape differences from a coarse to fine resolution, and facilitate shape-based classification. A procedure is developed to apply this method to cortical surface models, and promising results are acquired on synthetic and real data.