Related Experiment Video
Updated: Jan 11, 2026
05:00
Intense Pulsed Light for the Treatment of Dry Eye Owing to Meibomian Gland Dysfunction
Published on: April 1, 2019
15.5K
Quantitative evaluation of meibomian gland dysfunction via deep learning-based infrared image segmentation
Ziyang Yu1,2,3, Zhijun Wei1, Mini Han Wang3,4,5
1Beijing Institute of Technology, Zhuhai, China.
Frontiers in Artificial Intelligence
|November 14, 2025
Summary
Deep learning models accurately diagnose and grade meibomian gland dysfunction (MGD) using quantitative features from infrared images. This approach enhances clinical utility for objective MGD assessment.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Meibomian gland dysfunction (MGD) diagnosis often lacks objective quantitative measures.
- Advanced image segmentation algorithms for meibomian glands (MG) show limited clinical integration.
Purpose of the Study:
- To develop and validate deep learning models for quantitative analysis of MG images to aid MGD diagnosis and grading.
- To extract morphological and distributional features for objective MGD assessment.
Main Methods:
- Utilized DeepLabV3+, U-Net, and U-Net++ for infrared MG image segmentation.
- Extracted quantitative morphological and distributional indicators of MGD.
- Employed Spearman correlation, box plots, and logistic regression for analysis and validation.
Main Results:
- Quantitative indicators showed significant positive correlations with MGD severity (p < 0.001).
- Distinct indicator distributions across MGD grades suggest disease progression patterns.
- Logistic regression models achieved high AUC values (0.87-0.94) for MGD classification.
Conclusions:
- Deep learning-assisted quantitative analysis provides a robust and clinically relevant framework for objective MGD diagnosis and grading.
- This approach offers a promising tool for automated medical image interpretation in ophthalmology.
