Exploiting GPUs to accelerate white blood cells segmentation in microscopic blood images

White blood cell (WBC) segmentation is one of the important topics in the medical image processing field. Several researchers used K-means clustering approach to segment WBC from blood smear microscopic images. In this paper, we use the parallelism capabilities of the Graphics Processing Units (GPUs) to accelerate the segmentation of WBC from microscopic images. We implement the K-means algorithm and the preprocess steps for WBC image segmentation in CUDA programming to take the advantages of large number of cores in GPUs. We systematically implement and evaluate the performance of WBC segmentation operations on CPU, GPUs, and CPU-GPU hybrid systems. In this work, we gained about 3X faster performance than sequential implementation achieved without affecting WBC segmentation accuracy.