Related Experiment Video
Updated: Jul 4, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
405
Automated pancreatic segmentation and fat fraction evaluation based on a self-supervised transfer learning network
Gaofeng Zhang1, Qian Zhan2, Qingyu Gao1
1School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China; Department of Radiology, Changhai Hospital of Shanghai, Naval Medical University, Shanghai, 200433, China.
Computers in Biology and Medicine
|January 29, 2024
Summary
A new 3D segmentation model, nnTransfer, accurately segments the pancreas in CT scans. This method also enables automated measurement of pancreatic fat, aiding in diagnosing diseases like diabetes.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Pancreas segmentation in CT images is crucial for diagnosing pancreatic disorders like tumors and diabetes.
- Accurate segmentation remains a challenge due to image complexity.
Purpose of the Study:
- To develop and evaluate a novel 3D segmentation model for pancreas segmentation using CT images.
- To enable automated measurement of pancreatic fat volume fraction (FVF) for disease assessment.
Main Methods:
- A novel 3D segmentation model, nnTransfer (nonisomorphic transfer learning) net, was proposed, utilizing generative models for self-supervised learning on unlabeled data.
- A dataset of 229 high-resolution CT images was created and annotated.
- Automated whole-volume measurement of pancreatic fat (FVF) was achieved using histogram analysis with local thresholding.
Main Results:
- The nnTransfer model achieved high performance in pancreas segmentation with a Dice Similarity Coefficient (DSC) of 0.937 ± 0.019 and a Hausdorff Distance (HD) of 2.655 ± 1.479.
- The mean pancreas volume was 91.95 ± 23.90 cm³ and the mean FVF was 12.67% ± 9.84%.
- The model demonstrated flawless and autonomous operation.
Conclusions:
- The nnTransfer model provides accurate pancreas segmentation and automated FVF measurement from CT images.
- This technique facilitates the evaluation of pancreatic diseases, especially in diabetic patients.
- The self-supervised learning approach enhances the model's ability to learn from image attributes.
