Related Experiment Video
Updated: Jul 11, 2025
05:00
Intense Pulsed Light for the Treatment of Dry Eye Owing to Meibomian Gland Dysfunction
Published on: April 1, 2019
15.0K
A Deep Learning Approach for Meibomian Gland Appearance Evaluation
Kasandra Swiderska1, Caroline A Blackie2, Carole Maldonado-Codina1
1Eurolens Research, Division of Pharmacy and Optometry, Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom.
Ophthalmology Science
|November 3, 2023
Summary
A deep learning algorithm was developed to calculate Meibomian gland characteristics from meibography images. While it can segment glands, further development is needed for accuracy in clinical diagnosis of Meibomian gland dysfunction.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Meibomian gland dysfunction (MGD) is a common condition affecting ocular surface health.
- Accurate assessment of Meibomian gland characteristics is crucial for diagnosing and managing MGD.
- Current methods for evaluating Meibomian glands can be time-consuming and subjective.
Purpose of the Study:
- To develop and evaluate a deep learning algorithm for automated calculation of Meibomian gland characteristics.
- To assess the accuracy of the algorithm in segmenting Meibomian glands and calculating key metrics.
Main Methods:
- A dataset of 1616 meibography images from 282 individuals was used for training, validation, and testing.
- A deep learning model was trained for Meibomian gland segmentation and metric calculation.
- The automated approach was compared against manual measurements for accuracy assessment.
Main Results:
- The semantic segmentation-based approach achieved an aggregated Jaccard index of 0.4718 for glands and 0.8470 for eyelids.
- Both AI algorithms underestimated gland area, length ratio, tortuosity, and width.
- Meibomian gland intensity was overestimated by AI compared to manual assessment.
Conclusions:
- The developed deep learning approach shows potential for segmenting Meibomian glands.
- Further refinement is necessary to address challenges like gland overlap and image sharpness.
- This AI-based method could aid clinicians in the diagnosis and management of MGD.
