Master's Thesis Defence by Danni Dromi
Title: The use of Moving Meshes for Segmentation and Registration of 2D Medical Imaging
Abstract: The use of moving meshes is a relatively new topic in computer science research, and the potential application range of the method is far from fully explored. This thesis investigates two potential applications within the field of medical image analysis, namely the problems of segmentation and registration. These applications have been developed for use in an existing moving mesh framework. The methods are discretized specifically for use on moving meshes. The existing mesh framework is thoroughly analyzed before implementation of the two methods is carried out. Finally, a thorough testing scheme is performed for evaluating the methods, by three distinctive criteria: verification, validation and performance. Mixed results were obtained, as there were both successful tests of segmentation and registration, as well as tests which showcased errors and shortcomings in the implementations and the numerical schemes. However, even though the methods will need further work (as will be explained in the thesis), they have ultimately shown promising results. Lastly, the performance testing indicates that a moving mesh based solution of either segmentation or registration can outperform a corresponding regular mesh based solution with regards to resource consumption.
Supervisors: Kenny Erleben and Sune Darkner (DIKU)
Examinator: Niels Jørgen Christensen (DTU)