Google DeepMind has announced that its project with London’s Moorfields Eye Hospital will now include working with clinicians to predict eye diseases before symptoms occur.
DeepMind previously partnered with Moorfields on a collaborative research project to explore whether artificial intelligence technology could help identify eye conditions. This August, DeepMind released the first results from the project, which showed that AI could match expert human doctors’ performance on recommendations for the correct course of treatment of more than 50 eye diseases. Additionally, the research produced a major milestone toward reducing “black box” concerns in the medical community, as the AI system was able to provide doctors with explanations for its recommendations.
That significant step forward has now lead DeepMind to start on its next research challenge — predicting eye disease and preventing severe conditions even before symptoms manifest.
The research focuses on the most common blinding eye disease, Age-related Macular Degeneration (AMD). Wet AMD can cause permanent blindness, while Dry AMD often leads to mild sight loss. Currently, ophthalmologists identify Wet AMD by interpreting OCT scans — highly detailed 3D pictures of the back of the eye.
The first goal of the next research stage will be to have the AI system analyze OCT scans to identify symptoms of severe sight loss that require urgent treatment. Researchers will analyze detailed scans of approximately 7,000 patients who have previously received treatment at Moorfields Eye Hospital for Wet AMD in one eye. The AI system will then attempt to predict if the condition is likely to occur in the patient’s other, healthy eye.
The collaboration between DeepMind and Moorfields advances the role of AI technology in patient care and is expected to be watched with interest across the healthcare sector, where AI has tremendous potential. The OCT scans will be de-identified during the prediction process, which DeepMind will carry out on the Google Cloud computing infrastructure due to its reliability and processing power; and also to establish conditions for the project’s future scalability. The use of the cloud has been endorsed by the UK’s NHS Digital.
Journalist: Fangyu Cai | Editor: Michael Sarazen