Automatic Liver Segmentation Using Multi-plane Integrated Fully Convolutional Neural Networks

Abstract

Automatic liver segmentation in 3D medical images is essential in many clinical applications, such as pathological diagnosis of surgical planning, postoperative assessment and hepatic diseases. However, it is still a very challenging task due to the complex background, fuzzy boundary, and various appearance of the liver. In this paper, we propose a multi-plane integrated fully convolutional neural network to segment the liver from 3D CT volumes. Our network uses multiple layers of dilated convolution filters to replace traditional ones. Residual connections and multi-scale predictions are also employed in the network to improve the segmentation performance. We extensively evaluated our method on the dataset of MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge. Our method outperformed other state-of-the-art methods with an average Dice score of 96.7% on the segmentation results of liver, which only used a single framework without any pre-processing operation on it.