Related Experiment Video
Updated: May 17, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
333
COUPLED SWIN TRANSFORMERS AND MULTI-APERTURES NETWORK(CSTA-NET) IMPROVES MEDICAL IMAGE SEGMENTATION
Siyavash Shabani1, Muhammad Sohaib1, Sahar A Mohamed1
1Department of Electrical and Biomedical Engineering, University of Nevada, Reno.
Summary
This study introduces CSTA-Net, a novel Vision Transformer model for 3D medical image segmentation. It achieves high accuracy, effectively delineating fine details in medical scans.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Vision Transformers (ViTs) show superior performance over traditional methods in visual tasks.
- 3D medical image segmentation remains a challenging area requiring advanced deep learning models.
Purpose of the Study:
- To introduce the Coupled Swin Transformers and Multi-Apertures Networks (CSTA-Net) for enhanced 3D medical image segmentation.
- To improve the delineation of fine details in medical images.
Main Methods:
- The CSTA-Net architecture integrates Swin Transformer outputs with an Aperture Network.
- Each aperture network combines global and local feature maps using convolution and fusion blocks.
Main Results:
- The model was evaluated on the Synapse multi-organ and ACDC datasets.
- Achieved an average Dice score of 90.19±0.05 on Synapse and 93.77±0.04 on ACDC.
- Demonstrated effective delineation of fine details.
Conclusions:
- CSTA-Net advances 3D medical image segmentation using Vision Transformers.
- The proposed architecture offers a promising approach for accurate and detailed medical image analysis.
