Related Experiment Video
Updated: Sep 6, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
A deep learning model established for evaluating lid margin signs with colour anterior segment photography
Yuexin Wang1, Xingheng Jia2, Shanshan Wei3
1Department of Ophthalmology, Beijing Key Laboratory of Restoration of Damaged Ocular Nerve, Peking University Third Hospital, Beijing, China.
Eye (London, England)
|June 23, 2022
Summary
Deep learning models accurately identify key lid margin signs from eye photographs, demonstrating high sensitivity and specificity. This AI approach shows promise for aiding dry eye disease diagnosis and management.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Imaging
Background:
- Dry eye disease diagnosis relies on identifying specific lid margin signs.
- Manual evaluation of these signs can be subjective and time-consuming.
- Objective and automated methods are needed to improve diagnostic accuracy and efficiency.
Purpose of the Study:
- To assess the feasibility of using a deep learning model for identifying lid margin abnormalities.
- To evaluate the diagnostic performance of AI in detecting various lid margin signs from anterior segment photographs.
- To explore the potential of artificial intelligence in assisting clinical decision-making for dry eye disease.
Main Methods:
- Collected 832 anterior segment photographs from 428 dry eye patients.
- Eight lid margin signs were manually labeled by ophthalmologists.
- Developed and trained eight deep learning models based on VGGNet-13 architecture.
- Evaluated model performance using sensitivity, specificity, and receiver operating characteristic (ROC) curves with area under the curve (AUC).
Main Results:
- Deep learning models achieved high AUC values, with several exceeding 0.97 for signs like posterior lid margin rounding and mucocutaneous junction retroplacement.
- High sensitivity and specificity were reported for most identified signs, with values often above 0.90.
- The models demonstrated strong performance in differentiating between normal and abnormal lid margin features.
Conclusions:
- Deep learning models can effectively identify multiple lid margin signs with high accuracy.
- The study confirms the potential of AI in automating lid margin evaluation for dry eye assessment.
- This technology can serve as a valuable tool to support ophthalmologists in diagnosing and managing dry eye disease.
