Image Segmentation of Liver CT Based on Fully Convolutional Network
Xinyu Jin, Huimin Ye, LI Lan-juan, Qi Xia · 2017
Automatic segmentation of liver CT images has become significant in clinic. High-resolution CT scanners produce a large amount of data that is cumbersome and time-consuming for doctors to review. On the other hand, the traditional artificial segmentation tends to be subjective and inefficient. In addition, the accuracy of liver segmentation highly depends on the level of doctor's expertise. Thus, it is necessary to develop an automatic liver segmentation method. Nowadays, fully convolutional network (FCN) models demonstrate excellent performances in solving many computer vision problems. Due to its excellent characteristics and self-learning ability, FCN has made great achievements in the segmentation of medical image (MRI, CT, X-ray, etc). In this paper, we design a specific FCN for the segmentation in a liver CT image, which is efficient and effective in real-world applications. In addition, the data augmentation strategy is involved in the training period. Experimental results demonstrate the superiority of our approach.