Health-monitoring with AI-enabled smartphone-based imaging of the eye

The gift of longevity carries with it a host of complexities in ensuring that lifespans are not only prolonged but also healthy and functional.

Effective health strategies must be deeply embedded in a framework that provides continuous access to healthcare services. This involves persistent monitoring of both health and illness.

The human retina stands out as an invaluable window into ones health, shown to reflect not only ocular diseases but also aging, neurodegeneration, and heart function.

High-end optical imaging devices such as optical coherence tomography (OCT) provide clinicians with remarkably clear pictures of the living retina. However, they are not widely accessible and portable for personal use.

With the drastic improvements in smartphone optical and sensory capabilities, taking diagnostic-quality snapshots of the eye fundus is becoming a viable option.

This poses a paramount opportunity to turn the smartphone imaging with the support of artificial intelligence (AI) into a powerful health monitoring tool.The proposed research will advance the AI methodology and provide a proof-of-concept in retinal disease monitoring.

It is based on the hypothesis that the disease-specific biomarkers visible on gold-standard 3D OCT imaging are recoverable with AI from 2D portable fundus imaging provided by smartphones.

In a unique interdisciplinary setting, it aims to (i) build multimodal foundation AI models based on a large amount of available retrospective imaging data, (ii) advance the AI methodology to distill across the imaging modalities the knowledge accrued by the models, and (iii) in a clinical study demonstrate that AI models operating on smartphone-based fundus images are comparable to their high-end optical imaging counterparts across diagnostic tasks.

The work will lead to a potentially disruptive, accessible and scalable technology for monitoring of retinal changes and disease activity having a lasting impact on patients and public health.