Related Experiment Video
Updated: Mar 13, 2026
08:41
Novel In Vivo Micro-Computed Tomography Imaging Techniques for Assessing the Progression of Non-Alcoholic Fatty Liver Disease
Published on: March 24, 2023
1.9K
Scale-adaptive supervoxel-based random forests for liver tumor segmentation in dynamic contrast-enhanced CT scans
Pierre-Henri Conze1, Vincent Noblet2, François Rousseau3
1ICube UMR 7357, University of Strasbourg, CNRS, FMTS, 300 bd Sébastien Brant, 67412, Illkirch, France. conze@unistra.fr.
International Journal of Computer Assisted Radiology and Surgery
|October 25, 2016
Summary
We developed a machine learning framework for classifying hepatocellular carcinoma (HCC) tumor necrosis using CT scans. This method efficiently segments liver tissues, improving clinical management.
Area of Science:
- Medical Imaging
- Machine Learning
- Oncology
Background:
- Hepatocellular carcinoma (HCC) management requires accurate tumor assessment.
- Dynamic contrast-enhanced CT (DCE-CT) is crucial for evaluating HCC.
- Automated tissue classification in HCC aids clinical decision-making.
Purpose of the Study:
- To propose a machine learning classification framework for estimating tumor necrosis rate in HCC from DCE-CT scans.
- To enable efficient segmentation of healthy, active, and necrotic liver tissues with minimal user interaction.
Main Methods:
- Utilized random forest (RF) on supervoxels with multi-phase features to differentiate liver tissues based on contrast agent dynamics.
- Extended the RF approach hierarchically and multi-scale for diverse spatial extents and appearance variations.
- Implemented an adaptive data sampling scheme combining RF and a hierarchical multi-scale tree from recursive supervoxel decomposition.
Main Results:
- Clinical data assessment validated the efficacy of multi-phase information within a multi-scale supervoxel representation for HCC tumor segmentation.
- The proposed framework demonstrated improved accuracy in classifying liver tissues within HCC tumors.
Conclusions:
- The developed framework advances accurate multi-label tissue classification for HCC management.
- This approach has broader applications beyond HCC, contributing to more precise tissue classification in medical imaging.
