Development of a Mobile Application for at-Home Analysis of Retinal Fundus Images

TL;DR

This work addresses the challenge of translating retinal fundus imaging into a practical at-home monitoring tool by proposing a mobile application that captures fundus photographs with a clip-on lens and computes longitudinal metrics linked to age-related ocular diseases, rather than issuing single-shot diagnostics. It combines Messidor- and MAPLES-DR–based modeling (with a ResNet-50 for retinopathy and edema and an EfficientNet-B0 + transformer variant) along with a tortuosity analysis and DeepSeeNet-derived glaucoma indicators to enable trend tracking and early health insights. The platform emphasizes user-friendly visualization, calendar-based trend analysis, notes, and secure data sharing, aiming for personal use, clinical collaboration, and research facilitation while acknowledging that AI outputs are not medical diagnoses. Key findings show modest retinopathy accuracy on Messidor and higher edema accuracy with MAPLES-DR, highlighting data imbalance and suggesting that feature segmentation benefits training data efficiency but may not drastically reduce runtime. Overall, the work offers a practical framework for at-home ocular health monitoring, with potential for expansion to additional metrics and broader research use cases.

Abstract

Machine learning is gaining significant attention as a diagnostic tool in medical imaging, particularly in the analysis of retinal fundus images. However, this approach is not yet clinically applicable, as it still depends on human validation from a professional. Therefore, we present the design for a mobile application that monitors metrics related to retinal fundus images correlating to age-related conditions. The purpose of this platform is to observe for a change in these metrics over time, offering early insights into potential ocular diseases without explicitly delivering diagnostics. Metrics analysed include vessel tortuosity, as well as signs of glaucoma, retinopathy and macular edema. To evaluate retinopathy grade and risk of macular edema, a model was trained on the Messidor dataset and compared to a similar model trained on the MAPLES-DR dataset. Information from the DeepSeeNet glaucoma detection model, as well as tortuosity calculations, is additionally incorporated to ultimately present a retinal fundus image monitoring platform. As a result, the mobile application permits monitoring of trends or changes in ocular metrics correlated to age-related conditions with regularly uploaded photographs.

Figures (4)

• Figure 1: A confusion matrix for the retinopathy grade predicting model, trained on 1100 images and tested on 100 images from the Messidor dataset. The different grades are represented by the numbers 0-3, with 0 indicating no retinopathy, and 3 indicating severe retinopathy.
• Figure 2: A confusion matrix for the risk of macular edema predicting model, trained on 1100 images and tested on 100 images from the Messidor dataset. The numbers 0-2 represent the different risks of edema, with 0 being the lowest and 2 the highest.
• Figure 3: An excerpt of the application program highlighting the prompt given to DeepSeek regarding three metrics output by various models. The prompt requests a bullet point summary of the interpretation of the metrics, followed by a brief summary of the pertinent medical conditions, as well as suggested courses of action. Note that although not visible in this figure, the platform provides warnings to the user that the generative AI feature does not represent a medical professional and should not be treated as such.
• Figure 4: Sample screenshots of the user interface of our application describing age-related eye diseases with diagrams [21], in addition to a sample information page detailing the interpretation of curved eye vessels on eye-health.