| A novel framework for medical image analysis, known as Shells and Spheres, has been developed by our research lab. This framework utilizes spherical operators of variable radius, centered at each image pixel and sized to reach, but not cross, the nearest boundary. Statistical population tests are performed on the populations of pixels within adjacent spheres to compare image regions across boundaries, delineating both independent image objects and the boundaries between them. This research has focused on developing the Shells and Spheres framework and applying it to the problem of segmentation of anatomical objects. Furthermore, we have rigorously studied the framework and its applications to clinical segmentation, validating and improving our n-dimensional segmentation algorithm. To this end, we have enhanced the original Shells and Spheres segmentation algorithm by adding a priori information, developing techniques for optimizing algorithm parameters, implementing a software platform for experimentation, and performing validation experiments using real 3D ovine cardiac MRI data. The system developed provides automated 3D segmentation given a priori information in the form of a trivial 2D manual training procedure, which involves tracing a single 2D contour from which 3D algorithm parameters are then automatically derived. We apply this system to segmentation of the Right Ventricular Outflow Tract (RVOT) to aid in research toward the creation of a Tissue Engineered Pulmonary Valve (TEPV). Experimental methods are presented for the development and validation of the system, as well as a detailed description of the Shells and Spheres framework, our segmentation algorithm, and the clinical significance of this work. |
