Comparative Analysis of Brain Tumor Segmentation with Fuzzy C-Means Using Multicore CPU and CUDA on GPU

Magnetic resonance imaging is widely applied in medical practice. It has become a difficult task to divide the brain's image into distinct groups due to the symbiosis of intensity and noise. In recent years, due to the enhanced soft tissue contrast of non-invasive imaging and magnetic resonance imaging (MRI) images, MRI-based brain tumor segmentation studies are gaining more attention. With nearly two decades of development, innovative approaches to use computer-aided techniques to the field of brain tumors are becoming more mature and approaching common clinical applications. In order to enhance the segmentation performance of MRI brain images, fuzzy C-means (FCM) method based on similarity measurement is implemented in this paper. However, high computational requirements when working with big datasets are the principal problem with these algorithms. GPU today plays a major role in implementing time-consuming algorithms to decrease the complexity of time. With the use of FCM algorithm in GPU reduces the time required for processing the large amount of data, and it is 19 times better than serial execution.