Related Experiment Video
Updated: Aug 24, 2025
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
1.3K
nnU-Net Deep Learning Method for Segmenting Parenchyma and Determining Liver Volume From Computed Tomography Images
Rowland W Pettit1, Britton B Marlatt2, Stuart J Corr3,4,5,6,7
1Department of Medicine, Baylor College of Medicine, Houston, TX.
Summary
A novel deep learning model accurately estimates liver volumes from CT scans, improving precision for liver transplantation donor matching. This automated approach offers a faster, more cost-effective solution than manual methods.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
- Transplantation Medicine
Background:
- Precise liver volume estimation is crucial for recipient-donor matching in liver transplantation.
- Current demographic-based estimates lack accuracy and specificity.
- Manual segmentation of medical images is time-consuming and costly.
Purpose of the Study:
- To develop a deep learning model for accurate liver volume estimation.
- To generate 3D organ renderings from computed tomography (CT) scans.
- To address the limitations of current manual liver volume assessment methods.
Main Methods:
- A nnU-Net deep learning model was trained to segment liver borders in 151 CT scans.
- Ground truth annotations were provided by a board-certified radiologist.
- Liver borders were identified in 3D voxels to reconstruct total organ volume.
Main Results:
- The nnU-Net model achieved 97.5% overlap accuracy in identifying liver borders.
- Calculated liver volume estimates had a mean percent error of 1.92% ± 1.54%.
- The model demonstrated high precision in automated liver segmentation.
Conclusions:
- Deep learning with nnU-Net provides accurate liver volume estimation from CT scans.
- This automated method is fast and suitable for pretransplant clinical workflows.
- The technology offers a significant advancement over manual segmentation techniques.
