Multi-Organ Segmentation Based on Convolutional Neural Network and Random Walk for CT Image

作者：Luo, Zengbo; Jiang, Huiyan^*; Li, Siqi; Yang, Benqiang

来源：Journal of Medical Imaging and Health Informatics, 2018, 8(5): 1057-1063.

DOI：10.1166/jmihi.2018.2395

• Multi-Organ Segmentation
• Convolutional Neural Network
• Gauss Mixture Model
• Random Walk

Multi-organ segmentation is very important in medical image processing field. This paper proposes a new segmentation method based on convolutional neural network and random walk for multi-organ ( liver, kidney, and spleen) on CT images. Firstly, a convolutional neural network is utilized to extract deep characteristics for organ global segmentation. Then, the Gauss mixture model is employed to select the seeds from the initial segmentation results. Finally, based on these seeds, random walk is used to optimize local segmentation results. The training and testing set adopt original CT images and artificial labeling organ images by doctors from the hospital respectively. Experimental results with about 100 CT images demonstrate that the combination of convolutional neural network and random walk could be effective for the segmentation of most organs. Comparison results show the performance advantages of our method as compared with related work and better evaluation results under various criteria. Besides, because of fitting for most organs segmentation, it provides a powerful help for 3D visualization.

• 出版日期2018-6
• 单位中国人民解放军总医院; 东北大学