1: \begin{abstract}

2: 

3: Automatic liver segmentation from CT volumes is a crucial prerequisite yet challenging task for computer-aided hepatic disease diagnosis and treatment.

4: In this paper, we present a novel 3D deeply supervised network (3D DSN) to address this challenging task.

5: The proposed 3D DSN takes advantage of a fully convolutional architecture which performs efficient end-to-end learning and inference.

6: More importantly, we introduce a deep supervision mechanism during the learning process to combat potential optimization difficulties,

7: and thus the model can acquire a much faster convergence rate and more powerful discrimination capability.

8: On top of the high-quality score map produced by the 3D DSN, a conditional random field model is further employed to obtain refined segmentation results.

9: We evaluated our framework on the public MICCAI-SLiver07 dataset.

10: Extensive experiments demonstrated that our method achieves competitive segmentation results to state-of-the-art approaches with a much faster processing speed.

11: 

12: \end{abstract}

13: