GPU acceleration of a patient-specific airway image segmentation and its assessment
Yu-Wei Chang,Tony W.H. Sheu
DOI: https://doi.org/10.48550/arXiv.2012.10684
2020-12-19
Abstract:Image segmentation plays an important role in computer vision, object detection, traffic control, and video surveillance. Typically, it is a critical step in the 3D reconstruction of a specific organ in medical image processing which unveils the detailed tomography of organ, tumor, and nerve, and thus helping to improve the quality of surgical pathology. However, there may be high computational requirements in it. With the advent of GPUs, more complex and realistic models can be simulated, but the deployment of these facilities also requires a huge amount of capital. As a consequence, how to make good use of these computational resource is essential to GPU computing. This study discusses the image segmentation of 3D airway reconstruction, identifies the computing-intensive task, and parallelizes the algorithm of image segmentation in order to obtain theoretical maximum speedup in terms of the benchmark ratio of GPU to CPU. There are five steps involved, which are the image acquisition, pre-processing, segmentation, reconstruction, and object recognition. It is worth to note that it takes 85\% of time on segmentation. This study successfully accelerates the image segmentation of 3D airway reconstruction by optimizing the memory usage, grid and block setting and multiple GPUs communication, thereby gaining a total speedup of 61.8 on two GPUs (Nvidia K40).
Image and Video Processing