Related Experiment Video
Updated: Sep 25, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
532
Semantic segmentation in medical images through transfused convolution and transformer networks
Tashvik Dhamija1, Anunay Gupta2, Shreyansh Gupta3
1Department of Electronics and Communication Engineering, Delhi Technological University, New Delhi, India.
Applied Intelligence (Dordrecht, Netherlands)
|May 2, 2022
Summary
Two new deep learning models, USegTransformer-P and USegTransformer-S, enhance medical image segmentation by combining local and global features. These models achieve superior precision in segmenting brain tumors, lung nodules, skin lesions, and nuclei.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Deep learning, particularly fully convolutional neural networks, has advanced automated medical image segmentation.
- Existing models often focus on localized features, neglecting global context crucial for accurate segmentation.
Purpose of the Study:
- To introduce two novel deep learning models, USegTransformer-P and USegTransformer-S, for precise medical image segmentation.
- To address the limitations of existing models by integrating both local and global feature extraction.
Main Methods:
- Developed USegTransformer-P and USegTransformer-S, which amalgamate transformer-based and convolution-based encoders.
- These models are designed to capture both localized and global contextual information within medical images.
Main Results:
- The proposed USegTransformer models demonstrate superior performance compared to state-of-the-art methods.
- Achieved high precision in diverse segmentation tasks, including brain tumors, lung nodules, skin lesions, and nuclei segmentation.
Conclusions:
- USegTransformer-P and USegTransformer-S offer a significant advancement in medical image segmentation accuracy.
- These models have the potential to greatly assist medical practitioners and radiologists in clinical practice.
