Related Experiment Video
Updated: Jun 9, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
475
[Colon polyp detection based on multi-scale and multi-level feature fusion and lightweight convolutional neural
Yiyang Li1, Jiayi Zhao2, Ruoyi Yu2
1School of Biomedical Engineering, Capital Medical University, Beijing 100069, P. R. China.
Summary
This study introduces a lightweight AI model for detecting colorectal polyps, crucial for early cancer prevention. The model achieves high accuracy and speed, offering a valuable tool for diagnosis.
Area of Science:
- Artificial Intelligence
- Medical Imaging
- Oncology
Context:
- Colorectal cancer (CRC) poses a significant global health burden.
- Early detection of colorectal polyps is critical for preventing CRC progression.
- Existing diagnostic methods can be labor-intensive and may miss subtle findings.
Purpose:
- To develop and evaluate a lightweight convolutional neural network (CNN) for automated detection and auxiliary diagnosis of colorectal polyps.
- To improve the efficiency and accuracy of polyp identification in endoscopic images.
- To provide a feasible tool for early CRC screening and diagnosis.
Summary:
- A 53-layer CNN with spatial pyramid pooling and feature pyramid network was designed for polyp feature extraction and fusion.
- Spatial and positional pattern attention modules were integrated to enhance boundary perception and key feature integration.
- The model achieved high performance metrics (accuracy: 0.9982, mAP: 0.9953) with a high frame rate (74 fps) and low parameter count (9.08 M).
Impact:
- The proposed lightweight CNN offers a rapid, accurate, and resource-efficient solution for colorectal polyp detection.
- This technology can significantly aid clinicians in the early diagnosis and prevention of colorectal cancer.
- The model's low operating requirements make it suitable for widespread clinical application and integration into screening programs.
