Related Experiment Video
Updated: Jun 19, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.8K
Detection of Cervical Lesion Cell/Clumps Based on Adaptive Feature Extraction
Gang Li1, Xingguang Li1, Yuting Wang2,3,4,5
1School of Artificial Intelligence, Chongqing University of Technology, Chongqing 401135, China.
Bioengineering (Basel, Switzerland)
|July 27, 2024
Summary
This study introduces AFE-Net, an adaptive feature extraction network for improved automated detection of cervical lesions in cytological images. The novel approach enhances detection accuracy and reduces model complexity for computer-aided diagnosis.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Computational Pathology
Background:
- Automated detection of cervical lesion cells/clumps in cytological images is crucial for computer-aided diagnosis.
- Variations in lesion cell/clump shape and size challenge existing detection methods, reducing performance.
Purpose of the Study:
- To develop an improved automated detection system for cervical lesions.
- To enhance the accuracy and efficiency of cervical lesion cell/clump detection in cytological images.
Main Methods:
- Proposes an adaptive feature extraction network (AFE-Net) incorporating an adaptive module and a global bias mechanism.
- Introduces a novel bounding box loss function, tendency-IoU (TIoU).
- Evaluates the network on the CDetector dataset.
Main Results:
- AFE-Net achieved a mean Average Precision (mAP) of 64.8% on the CDetector dataset.
- The model demonstrated a 2.2% improvement in mAP compared to YOLOv7.
- AFE-Net reduced the number of parameters by 11.8% (30.7 million) compared to YOLOv7 (34.8 million).
Conclusions:
- AFE-Net effectively improves the detection performance of cervical lesion cell/clumps by combining adaptive and global features.
- The proposed TIoU loss function contributes to enhanced bounding box prediction.
- AFE-Net offers a more efficient and accurate solution for computer-aided cervical cancer screening.
