Related Experiment Video
Updated: May 13, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
WCEDSAM: A Lightweight Multi-Scale Colonoscopy Polyp-Segmentation Network Combining Frequency-Domain Decomposition
Lei Wang1, Tongyu Wang1, Sitong Liu2
1School of Medical Technology and Information Engineering, Zhejiang Chinese Medical University, Hangzhou 310053, China.
Biology
|May 12, 2026
Summary
A new lightweight model, WCEDSAM, enhances colorectal cancer polyp segmentation by efficiently processing complex image features. This advanced AI model achieves superior accuracy in identifying polyps, improving screening effectiveness.
Area of Science:
- Medical Image Analysis
- Artificial Intelligence in Healthcare
- Cancer Diagnostics
Background:
- Colorectal cancer screening faces challenges due to polyp variability and computational demands.
- Existing segmentation models struggle with indistinct boundaries and overlapping features.
- Need for efficient and accurate AI models for polyp detection.
Purpose of the Study:
- To develop a lightweight medical image segmentation model, WCEDSAM, for improved colorectal polyp detection.
- To enhance feature extraction and reduce computational cost in polyp segmentation.
- To achieve state-of-the-art performance in polyp segmentation across diverse datasets.
Main Methods:
- WCEDSAM is a compact model based on MedSAM, incorporating a Wavelet Transform for pixel-level feature separation.
- A DSConv-ECA module is integrated before the ViT encoder for efficient local feature capture and reduced parameters.
- The model was evaluated on five public datasets (Kvasir-SEG, CVC-ClinicDB, CVC-ColonDB, CVC-300, ETIS).
Main Results:
- WCEDSAM achieved high performance with 15.38 million parameters.
- Mean Dice (mDice) scores of 0.9383 on Kvasir-SEG and 0.9376 on CVC-ClinicDB.
- Cross-domain mDice scores of 0.9189 (CVC-ColonDB), 0.8961 (CVC-300), and 0.7765 (ETIS), outperforming UNet++ and TransUNet.
Conclusions:
- WCEDSAM demonstrates superior performance in colorectal polyp segmentation compared to existing methods.
- The model's lightweight design and novel feature extraction modules offer computational efficiency.
- WCEDSAM shows significant potential for improving the accuracy and efficiency of colorectal cancer screening.