Past, present, and future trend of GPU computing in deep learning on medical images

A Segmentation process is labeling an image or images for obtaining more meaningfull information. On biomedical images, this activity has an important role in helping pathologist for conducting advance analysis. After Graphical Proceessing Unit (GPU) introduced not only for graphical necessary but also for general purpose computing, segmentation process which is computationally expensive can be potentially improved. The good accuracy of detection and segmentation result provides morphological information for the pathologist. Consequently, more approaches were developed to ensure the good performance of detection and segmentation such as deep learning approach. Convolutional Neural Network (CNN) is one of deep learning architecture with complex computation. This paper presents an overview of utilization of CNN as prominent deep learning architecture under GPU platform and propose an approach of using GPU as potential further parallelie techniques in CNN.