Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique.

Area of Science:

• Medical Image Analysis
• Deep Learning
• Computational Anatomy

Background:

• Accurate abdominal multi-organ segmentation is crucial for clinical applications like diagnosis and treatment planning.
• Existing methods often face challenges in precise feature extraction and computational efficiency.

Purpose of the Study:

• To propose Swin-PSAxialNet, an efficient deep learning network for precise abdominal multi-organ segmentation in CT images.
• To enhance 3D feature extraction and capture subtle details for improved segmentation accuracy.

Main Methods:

• Developed Swin-PSAxialNet, an enhanced nnU-Net architecture incorporating Space-to-depth (SPD) modules and parameter-shared axial attention (PSAA) blocks.
• Implemented a multi-scale image fusion approach to capture detailed and spatial features.
• Utilized parameter-sharing to reduce computational cost and accelerate training.

Main Results:

• Achieved a high average Dice coefficient of 0.93342 for segmenting 11 abdominal organs.
• Demonstrated superior performance compared to previous mainstream segmentation methods.
• Exhibited excellent accuracy and computational efficiency in segmenting major abdominal organs.

Conclusions:

• Swin-PSAxialNet offers a significant advancement in abdominal multi-organ segmentation accuracy and efficiency.
• The proposed method provides a robust solution for supporting clinical diagnosis and treatment planning through precise medical image analysis.