Simultaneous Segmentation and Registration of MR Images
Tejos C, Guarini M, Irarrazaval P
Pontificia Universidad Catolica de Chile
In medical imaging studies that analyse the evolution of structures through time, objects of interest need to be not only segmented, but also registered to correct pose changes resulting from the re-positioning of patients in the scanner. A few researchers have successfully integrated both processes, however, the existing joint segmentation and registration techniques provide geometrically distorted results due to their pixel-based characteristics, or correspond to 2D schemes without 3D extensions. The Simplex Mesh Diffusion Snake technique, a recently developed prior knowledge-based deformable model segmentation algorithm, is here employed to perform simultaneously segmentation and registration of synthetic and real MR images, with sub-pixel accuracy.