Improved algorithm for liver tumor segmentation in CT images based on U-Net

• Published in: 2024 2nd International Conference on Signal Processing and Intelligent Computing (SPIC) pp. 1022 - 1027
• Main Authors: Qi, Yan, Zhao, Hongkun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20.09.2024
• Subjects: Computed tomography; Image segmentation; Liver; liver CT images; medical image segmentation; Robustness; Semantics; Signal processing; Signal processing algorithms; Testing; Training; Tumors; U-Net
• DOI: 10.1109/SPIC62469.2024.10691416

Aiming at the problem of inaccurate boundary segmentation of liver tumors by existing algorithms, an improved model REEC-UNet based on U-Net was proposed. Firstly, in order to alleviate the problem of gradient disappearance and semantic information loss in deep networks, residual connection is introduced into U-Net network framework. Then, the ECANet channel attention network was added in the coding stage and decoding stage to improve the performance of the convolutional neural network. The training and testing were carried out on the LiTS2017 dataset. The experimental findings demonstrated that the proposed method achieved more precise tumor boundary segmentation and enhanced the accuracy of liver tumor segmentation.