Patch-based segmentation using refined multifeature for magnetic resonance prostate images

In this paper, a new method is proposed for prostate segmentation in magnetic resonance (MR) images, which is an important step for image-guided brachytherapy system. Different from the conventional patch-based segmentation method, a multicharacter feature set is proposed within the range of statistical feature on voxel intensity, texture feature description, shape features, and spatial relation feature, to resolve the limitations of incomprehensive feature set. Furthermore, an Adaptive Boosting algorithm is used to refine these features in order to reduce the redundancy of features. Using the refined features, the patch-based segmentation framework has a better effect. The method is tested by 40 prostate MR images and compared with several classical image segmentation methods; the test and comparison results show that the proposed method consistently achieves higher segmentation accuracy.