Abstract
We propose a novel approach for semantic CT image segmentation based only on a fully convolutional network （FCN）, which accomplishes an end-to-end, voxel-wise multiple-class prediction to map each voxel in a CT image directly to an anatomical label. The proposed method simplifies the segmentation of the anatomical structures（including multiple organs）in a CT image（generally in 3D）to majority voting for the semantic segmentation of multiple 2D slices drawn from three orthogonal viewpoints with redundancy. An FCN consisting of “convolution” and “de-convolution” parts is trained and re-used for the 2D semantic image segmentation of different slices of CT scans. We applied the proposed method to segment a wide range of anatomical structures that consisted of 19 types of targets in the human torso. A database consisting of 240 3D CT scans and a humanly annotated ground truth was used for training（230 cases）and testing（the remaining 10 cases）. The results showed that the target regions for the entire set of CT test scans were segmented with acceptable accuracies（89% voxels were labeled correctly）against the human annotations. This performance was comparable to other recently reported state-of-the-art results. Compared to previous segmentation methods that have to be guided by human expertise, this data-driven approach showed better efficiency, generality, and flexibility.
