Deep Learning-Powered Smartphone Application for Diagnosing Common Eye Diseases
Farhad Nejat1 *
- Ophthalmologist, Vision health research center, Tehran, Iran
Abstract: To develop a smartphone-based application with trained deep learning algorithms to diagnose and manage common eye diseases, including dry eye disease, conjunctival nevus, strabismus, and keratoconus, just with smartphone camera without requiring additional hardware.
Methods: The application employs convolutional neural networks (CNNs) as the core of its deep learning model. A large dataset of labeled eye images was collected and preprocessed, focusing on disease-specific features. These images were captured using a Samsung A71 smartphone by one expert person. A validation study was conducted comparing the app's performance to diagnoses made by expert ophthalmologists.
Results: The deep learning model demonstrated a diagnostic accuracy of over 95% for each targeted eye diseases, aligning closely with the performance of ophthalmologists. The application successfully identified key features of dry eye disease, conjunctival nevus, strabismus, and keratoconus in various testing scenarios. Its ease of use and reliability make it a practical tool for both clinicians and patients.
Conclusion: This deep learning-powered application serves as a crucial third party between ophthalmologists and patients, enabling continuous eye care. Its high diagnostic accuracy and accessibility ensure that patients are promptly identified and managed, reducing the neglect or delayed treatment. By providing reliable diagnostic support directly through a smartphone anytime and anywhere, the app helps ophthalmologists monitor patients and maintain consistent treatment plans, fostering better long-term outcomes. This innovation bridges the gap between specialized care and patients, empowering both clinicians and individuals to actively engage in managing eye health.
