Abstract
Liver segmentation has always been the focus of researchers because it plays an important role in medical diagnosis. However, under the condition of low contrast between a liver and surrounding organs and tissues, CT image noise and the large difference between the liver shapes of patients, existing liver image segmentation algorithms are difficult to obtain satisfactory results. To improve this situation, we propose a liver CT sequesnce image segmentation algorithm GIU-Net, which combines an improved U-Net neural network model with graph cutting. Specifically, we initially segment a liver from a liver CT sequence using an improved U-Net and obtain the probability distribution map of the liver regions. Secondly, the sequence segmentation start slice is selected, and then the context information of the liver sequence images and the liver probability distribution map are used to construct a graph cut energy function. Finally, the segmentation is done by minimizing the graph cut energy function. Our experimental results show that GIU-Net has a good performance when segmenting liver sequence images in terms of segmentation accuracy and robustness.
Original language | English |
---|---|
Pages (from-to) | 54-63 |
Number of pages | 10 |
Journal | Expert Systems with Applications |
Volume | 126 |
DOIs | |
State | Published - Jul 15 2019 |
Keywords
- Deep learning
- GIU-Net
- Graph cut
- Liver segmentation
- U-Net