Related Experiment Video
Updated: Oct 12, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
554
Automatic multi-organ segmentation from abdominal CT volumes with LLE-based graph partitioning and 3D Chan-Vese model
Ping Tang1, Yu-Qian Zhao2, Miao Liao3
1School of Automation, Central South University, Changsha, 410083, China; School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China.
Computers in Biology and Medicine
|November 20, 2021
Summary
This study introduces an automatic method for segmenting multiple abdominal organs in 3D CT scans. The approach achieves high accuracy for liver, spleen, and kidneys without extensive training, offering a competitive solution for medical image analysis.
Area of Science:
- Medical Imaging
- Computer Vision
- Computational Anatomy
Background:
- Accurate segmentation of abdominal organs in 3D CT scans is crucial for diagnosis and treatment planning.
- Existing methods often require significant manual intervention, extensive training, or registration processes.
- Challenges include variations in organ shape, location, and weak boundary definition.
Purpose of the Study:
- To develop a fully automatic method for multi-organ segmentation in 3D abdominal CT volumes.
- To improve the accuracy and efficiency of abdominal organ segmentation.
- To provide a robust method capable of handling anatomical variations and challenging image features.
Main Methods:
- Initial segmentation using Local Linear Embedding (LLE)-based graph partitioning after spine and rib removal.
- Segmentation refinement via a hybrid intensity model, 3D Chan-Vese model, and histogram equalization.
- Boundary correction using a pseudo-3D bottleneck detection algorithm.
Main Results:
- Achieved high Dice similarity coefficients (e.g., 95.9% for liver) and Jaccard indices (e.g., 92.2% for liver).
- Demonstrated competitive performance against state-of-the-art methods on the XHCSU20 dataset.
- Reported an average running time of approximately 6 minutes per CT volume with high accuracy, precision, recall, and specificity.
Conclusions:
- The proposed automatic method offers a robust and efficient solution for multi-organ segmentation in abdominal CT.
- It effectively addresses challenges like shape variations and weak organ boundaries without needing heavy training or registration.
- The method shows high accuracy and efficiency, proving competitive with existing techniques and suitable for liver-only segmentation tasks.
