Related Experiment Video
Updated: Jan 2, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
709
Automatic multi-organ segmentation in dual-energy CT (DECT) with dedicated 3D fully convolutional DECT networks.
Shuqing Chen1, Xia Zhong1, Shiyang Hu1,2
1Pattern Recognition Lab, Universität Erlangen-Nürnberg, Erlangen, 91058, Germany.
Medical Physics
|December 10, 2019
Summary
This study introduces four new deep learning models for segmenting organs in dual-energy computed tomography (DECT) scans. These advanced algorithms effectively utilize DECT data, achieving high accuracy in segmenting multiple thoracic and abdominal organs.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Dual-energy computed tomography (DECT) offers enhanced contrast and tissue differentiation compared to single-energy CT (SECT).
- Automatic multi-organ segmentation is crucial for advancing DECT clinical applications.
- Existing segmentation methods primarily focus on SECT, leaving a gap in DECT-specific research.
Purpose of the Study:
- To develop novel algorithms for automatic multi-organ segmentation of DECT data.
- To leverage the additional spectral information provided by DECT for improved segmentation accuracy.
- To address the limited research in DECT-specific segmentation methods.
Main Methods:
- Proposed four 3D fully convolutional neural network (CNN) architectures.
- Integrated and fused dual-energy spectral information within each network.
- Employed a fivefold cross-validation strategy for evaluation.
Main Results:
- Achieved high average Dice coefficients: 98% (right lung), 98% (left lung), 96% (liver), 92% (spleen), 95% (right kidney), 93% (left kidney).
- Demonstrated superior performance compared to deep learning methods designed for SECT.
- Validated on 45 thorax/abdomen DECT datasets from a clinical dual-source CT system.
Conclusions:
- The developed DECT segmentation methods are feasible and show promising results.
- The algorithms exhibit high adaptability for practical clinical applications.
- The study highlights the potential of exploiting DECT data for advanced medical image analysis.
