Related Experiment Video
Updated: Dec 28, 2025
08:13
In Vivo Confocal Microscopy in the Diagnosis and Management of Dry Eye: A Focus on Imaging Protocols and Interpretation
Published on: November 11, 2025
258
Deep Neural Network-Based Method for Detecting Obstructive Meibomian Gland Dysfunction With in Vivo Laser Confocal
Sachiko Maruoka1, Hitoshi Tabuchi1,2, Daisuke Nagasato1,2
1Department of Ophthalmology, Tsukazaki Hospital, Himeji, Japan.
Cornea
|February 11, 2020
Summary
Deep learning models accurately detect obstructive meibomian gland dysfunction (MGD) using laser confocal microscopy. This technology shows promise for future automated MGD diagnosis in patients.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Meibomian gland dysfunction (MGD) is a common condition affecting eyelid glands.
- Accurate diagnosis of obstructive MGD is crucial for effective treatment.
Purpose of the Study:
- To assess the efficacy of deep learning (DL) models in identifying obstructive MGD.
- Utilized in vivo laser confocal microscopy images for analysis.
Main Methods:
- Trained nine distinct deep learning network structures.
- Evaluated single and ensemble DL models using metrics like AUC, sensitivity, and specificity.
- Analyzed images from 137 individuals with obstructive MGD and 84 controls.
Main Results:
- The highest performing single DL model (DenseNet-201) achieved an AUC of 0.966, sensitivity of 94.2%, and specificity of 82.1%.
- The top ensemble DL model (VGG16, DenseNet-169, DenseNet-201, InceptionV3) demonstrated superior performance with an AUC of 0.981, sensitivity of 92.1%, and specificity of 98.8%.
Conclusions:
- Deep learning models combined with laser confocal microscopy effectively differentiate between healthy and obstructive MGD.
- Achieved high diagnostic accuracy, paving the way for potential automated MGD diagnosis.
