Automatic Liver and Spleen Segmentation with CT Images Using Multi-channel U-net Deep Learning Approach

The detection and evaluation of the shape of liver from abdominal computed tomography (CT) images are fundamental tasks in computer-assisted liver surgery planning such as radiation therapy. The contour of spleen is also a significant factor highly related to liver diseases. However, automatic and accurate liver segmentation still remains many challenges to be solve, such as ambiguous boundaries, heterogeneous appearances and highly varied shapes of the liver and spleen. To address these difficulties, we developed an automatic segmentation model based on multi-channel U-net network. Some preprocessing steps were done to elevate the performance first. Also, an approximate liver and spleen map was generated by calculating the gradient of CT images. The area which have high possibility to be liver and spleen would be select as the training set to make sure the balance of data. Then, a deep learning U-net structure was applied for the processed training data. Finally, some post-processing methods, which include k-means clustering and morphology algorithms, would be applied in our protocol. The results indicated that a high structure similarity index (SSIM) and dice score coefficient of liver and spleen segmentation model can be achieved, which were 0.9731 and 0.9508 respectively, demonstrating the potential clinical applicability of the proposed approach.