A coupled segmentation and registration framework for medical image analysis using robust point matching and active shape model

Image segmentation and non-rigid registration are two widely investigated tasks in medical image analysis. Concurrent segmentation and registration methods have received considerable attention in recent years. While some models have been shown to give interesting results, most of them are either able to improve segmentation results alone or able to correct rigid rotation and translation only. In addition, the previous reported methods require a long waiting time to converge. In this paper, we present a fast coupled approach that performs automatic segmentation and nonrigid registration simultaneously. It can progressively identify the boundaries of several organs of interest based on the achieved transformation, and conversely, accurate knowledge of important structures enables us to achievemore precise nonrigid mapping result. We tested the method on two different applications to validate our approach: prostate image guided radiotherapy on computed tomography (CT) volumes and brain image analysis on magnetic resonance images (MRI). Quantitative analysis of experimental results shows that the segmentation result has promising global and local consistencies with the manual segmentation, and the mapping result outperforms the traditional rigid and nonrigid registration. In addition, the new method is proven to be faster and more efficient than previous reported ones.¹