Medical surface smoothing via adaptive diffusion of differential fields

For diagnosis, therapy planning and medical education, it is common to extract surface models from clinical data. Due to various factors, extracted surface models often contain artifacts and noise. Most existing smoothing techniques are mainly devised for non-medical models and have parameters which are difficult to be tuned manually and sensitive to smoothing, surface quality and volume shrinkage. Unlike most previous smoothing techniques that are vertex-based or normal-based, we present an adaptive smoothing algorithm based on differential fields. The accuracy and performance of our approach is experimentally validated via comparisons with conventional smoothing techniques on typical medical surface models. We show the presented algorithm can achieve a smoothing medical surface while preserving volume and fine structures of original objects.