Automated Liver Tumor Segmentation and Staging Method Using 3D U-Net in CT Images

Abstract

Accurate1 diagnosis and staging of Hepatocellular Carcinoma (HCC) are essential for formulating treatment strategies and assessing patient prognosis. Diagnosis and staging of HCC necessitate the review of hundreds of Computed Tomography (CT) images to identify lesions. The process is time-consuming and relies heavily on the subjective judgment of medical professionals, thereby introducing the risk of misdiagnosis. This paper proposes an automatic liver tumor segmentation and staging method to address the issues and to achieve objective and accurate diagnosis and staging of HCC. The proposed method utilizes a 3D U-Net to automatically segment the liver, liver vessels, and tumor regions from CT images, subsequently generating a surface mesh of the segmented areas. From this generated mesh, the number, size, and vascular invasion of tumors are assessed, and staging is conducted based on these parameters. The accuracy of the proposed method is evaluated through experiments. The experimental results demonstrate that the proposed method can accurately stage HCC, contributing to the objective and precise diagnosis and staging of HCC. © 2024 Copyright is held by the owner/author(s).