The following is a summary of “Next-Generation tear meniscus height detecting and measuring smartphone-based deep learning algorithm leads in dry eye management,” published in the May 2024 issue of Ophthalmology by Nejat et al.
Researchers conducted a prospective study developing and assessing a learning system using Python-based deep learning code to help diagnose and manage dry eye disease (DED) using smartphone images.
They reviewed the Vision Health Research Clinic data, analyzing 1,021 eye images from 734 patients, 70% females and 30% males. The images were captured by one specialist using a Samsung A71 (601 images) and an iPhone 11 (420 images) with a flashlight on and a direct gaze at the camera. For 80% of the training data, the images were segmented into 3 parts for every eye image, such as eye, lower eyelid, and iris segmentation. Segmentation and tear meniscus height measurement were done using deep learning algorithms.
The results showed that the model trained on 80% of the data and validated on 20% from both Samsung A71 and iPhone 11. The trained model achieved a dice coefficient of 98.68% and an overall accuracy of 95.39%.
Investigators concluded that the algorithm could revolutionize diagnosing and managing DED. With just smartphones, homecare devices might be able to provide effective solutions, marking a significant step forward in eye care accessibility.
Source: ophthalmologyscience.org/article/S2666-9145(24)00082-4/fulltext