Related Experiment Video
Updated: Aug 23, 2025
05:30
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
182
Image Quality Classification for Automated Visual Evaluation of Cervical Precancer
Zhiyun Xue1, Sandeep Angara1, Peng Guo1
1National Library of Medicine, National Institutes of Health, Bethesda, MD 20894, USA.
Medical Image Learning with Limited and Noisy Data : First International Workshop, Milland 2022, Held in Conjunction with MICCAI 2022, Singapore, September 22, 2022, Proceedings. Milland (Workshop) (1St : 2022 : Singapore)
|October 31, 2022
Summary
This study introduces a deep learning framework to classify cervical images for precancer screening, improving automated visual evaluation (AVE) by identifying poor quality and mislabeled images.
Area of Science:
- Medical Imaging
- Computer Vision
- Machine Learning
Background:
- Image quality is crucial for accurate data analysis in medical diagnostics.
- Poor image quality negatively impacts both manual and automated evaluations, affecting diagnostic reliability.
- Automated Visual Evaluation (AVE) for cervical precancer screening requires robust image quality control.
Purpose of the Study:
- To develop and evaluate a deep learning ensemble framework for cervical image quality classification.
- To improve the performance of automated visual evaluation (AVE) in cervical precancer screening.
- To automatically identify mislabeled and ambiguous images, enhancing classification accuracy.
Main Methods:
- A deep learning ensemble framework integrating cervix detection, mislabel identification, and quality classification was developed.
- Images were classified into four quality categories: unusable, unsatisfactory, limited, and evaluable.
- The method was evaluated on a large, diverse dataset of 87,420 images from 14,183 patients worldwide.
Main Results:
- The proposed ensemble approach demonstrated superior performance compared to baseline methods.
- The framework effectively identified mislabeled and ambiguous images, improving overall classification accuracy.
- The system achieved high performance across images from various providers, devices, and geographic regions.
Conclusions:
- The developed deep learning ensemble framework significantly enhances cervical image quality control for AVE.
- This approach offers a scalable and reliable solution for improving cervical precancer screening accuracy.
- Automated quality assessment is vital for advancing computer-aided diagnostic tools in women's health.
