Automated 3D visualization and volume estimation of hepatic structures for treatment planning of hepatocellular carcinoma
B. Lakshmipriya,Biju Pottakkat,G. Ramkumar,K. Jayanthi
DOI: https://doi.org/10.1016/bs.adcom.2024.03.005
IF: 3.067
2024-04-30
Advances in Computers
Abstract:Hepato Cellular Carcinoma (HCC), an aggressive cancer of the liver occurs mainly in patients suffering from chronic liver disease. Anatomical structural assessment and volume calculation of the liver, tumor and the vessels within the liver are required for planning procedures in liver. Currently available pre-surgical planning softwares are mostly user guided and require manual selection of seed regions from the computed tomography (CT) images in the three-dimensional (3D) reconstruction and volumetry analyses. This chapter aims to develop a fully automated 3D visualization and volumetry estimation tool for liver without manual interface with the help of deep learning algorithms so that the physicians can use it on bed side for the treatment and surgery planning of liver tumors. Development of automated treatment planning tool involves two modules: segmentation and 3D modeling & volumetric calculation. Since abdominal CT images contain other organs and structures like heart, stomach, kidney, spleen, pancreas, ribs and spine, in addition to liver, liver has to be segmented from all slices of CT images as a first step. Considering the challenges associated with liver segmentation, it is proposed to perform automated hepatic structures segmentation using DeepLabV3+, a deep learning based semantic segmentation architecture. Segmentation of liver, tumor, hepatic artery, portal vein and hepatic veins are implemented simultaneously using DeepLabV3+ architecture with XceptionNet as base network. The segmentation performances are assessed in terms of global segmentation accuracy, Jaccard index, Dice coefficient and boundary F1 Score. Subsequently in the second module, 3D model of liver together with the associated hepatic structures is reconstructed from the segmented CT image slices of the patient using Simpleware ScanIP, a 3D image processing software. Additionally, volumes of the healthy liver, tumor and the vascular structures are also computed using this imaging tool. The visual and volumetry outcomes of the 3D framework proposed in this study is validated by reconstructing the 3D model of the liver and hepatic structures from the corresponding ground truths considered for the segmentation algorithm. The proposed 3D framework was implemented on two datasets viz. 3D-IRCAD-01 database and an institutional dataset of liver CT images from Jawaharlal Institute of Postgraduate Medical Education Research (JIPMER), Puducherry, India. The outcomes of the fully automated 3D visualization and volumetry framework proposed in this study are expected to facilitate doctors in their surgical as well as non-surgical treatment planning for patients with HCC by providing them a visually appealing model of the liver, structures inside the liver, liver tumor and their volumes.
computer science, software engineering, hardware & architecture