Deep Learning Methods Aid Diagnosis Of Eye Disease

The first results of a research collaboration’s investigation into the use of AI for the diagnosis and referral of retinal disease have been released. After applying its novel deep learning architecture to OCT scans from hospitalized patients, the team discovered that the AI platform’s performance in making a referral recommendation met or exceeded that of experts in various retinal diseases.

The results show deep learning can quickly and accurately interpret eye scans from routine clinical practice. It is as accurate as world-leading expert doctors in recommending how patients should be referred for treatment for over 50 sight-threatening eye diseases.

If the system can handle the wide range of patients seen in routine clinical practice, it could help doctors quickly prioritize patients who require urgent treatment, ultimately saving patients’ sight.

For some time, AI has been a promising solution for medical image interpretation, but applying it in real-world clinical scenarios has proven difficult. According to the team’s paper, AI systems must be trained on hundreds of thousands of examples from a single canonical dataset before being generalized to new populations and devices without significant performance loss or prohibitive data requirements.

Furthermore, the AI tools must apply to real-world scans and be designed for clinical deployment while matching or exceeding the performance of human experts in such scenarios.

Read more