Segmentation in noisy medical images using PCA model based particle filtering

Existing common medical image segmentation algorithms such as snake or graph cut usually could not generate satisfying results for noisy medical images such as X-ray angiographical and ultrasound images where the image quality is very poor including substantial background noise, low contrast, clutter, etc. In this paper, we present a robust segmentation method for noisy medical image analysis using Principle Component Analysis (PCA) based particle filtering. It exploits the prior clinical knowledge of desired object's shape through a PCA model. The preliminary results have shown the effectiveness and efficiency of the proposed approach on both synthetic and real clinical data.