A Novel Deformable Model for Medical US Image Segmentation

Medical ultrasound (US) images are characterized by inherent noise perturbations which introduce uncertainty in their interpretation even by experienced radiologists. In this paper we propose a novel deformable model for medical US image segmentation, which incorporates a combination of domain and boundary integrals for the alleviation of the noise effects. It can be applied as a diagnostic aid for the delineation of hypo-echoic regions, which in many cases are associated with pathological findings. The proposed model was experimentally evaluated using various types of medical US images. The results show that it leads to more accurate segmentation results whereas it converges faster than other relevant models.