Medical Image Analysis

This research work presents a PDE-based deformable surface whose deformation behavior is governed by partial differential equations (PDEs) such as the weighted minimal surface flow. Fundamentally different from the level-set approach, our model always maintains an explicit representation of geometry and topology. The new PDE-based model can either grow from the inside or shrink from the outside of the dataset, and it can automatically split to multiple objects whenever necessary during the deformation process. More importantly, the new model can represent a series of level-of-details (LODs) through the use of global or local/adaptive subdivision. We demonstrate both the accuracy and the robustness of our method through a number of experiments on medical imaging applications such as the segmentation of both real and synthetic volumetric datasets.