Parallel Segmentation of Tumor from MRI Using Image Processing

Abstract

Modern medical image analysis relies heavily on medical picture segmentation, which has important ramifications for computer-aided diagnosis and therapy planning. This research presents a novel framework that combines morphological methods, K-means clustering, and U-Net architecture for the parallel segmentation of brain tumours using magnetic resonance imaging (MRI). In response to issues brought up in the literature, our work includes a thorough analysis of current segmentation techniques and makes explicit links between the literature review and the suggested methodology. We go into detail about the iterative design process, explaining the results of each design iteration and how the literature influenced our approach. Important design parameters are clearly stated, illustrating the framework's evolutionary path. Our method, which makes use of Single Instruction Multiple Data (SIMD) parallelization, improves speed while maintaining accuracy. Our framework's thorough analysis of design parameters, literature influences, and design iterations establishes it as a notable development in the field of medical picture segmentation.