Parallel genetic-based algorithm on multiple embedded graphic processing units for brain magnetic resonance imaging segmentation

Medical imaging has played an important role in helping physicians to make clinical diagnoses. Magnetic resonance imaging technology has been used to image the anatomy of the brain. Typically, image segmentation is utilized to observe the brain's anatomical structures and its changes, and to identify pathological regions. In this paper, we propose an efficient parallel fuzzy c-means clustering algorithm for segmenting images on multiple embedded graphic processing unit systems, NVIDIA TK1. The experimental results demonstrate that the maximum speedups of the proposed algorithm on 15 TK1s greater than 12 times and 7 times than that of fuzzy c-means algorithm with single ARM and Intel Xeon CPUs, respectively. These experimental results show that the proposed algorithm can significantly address the complexity and challenges of the brain magnetic resonance imaging segmentation problem.