Diabetes mellitus is a condition that is widespread around the world. A potential complication of this condition is diabetic retinopathy, which can lead to vision problems and blindness for patients. Patients who are at risk may have their eyes examined to check for the progression of this condition. This project will use an artificial intelligence approach for the classification of eye images that are taken from a patient who is at risk of developing diabetic retinopathy. The project will create a tool or algorithm that will accept an image from a patient’s eye exam as an input, and will determine if that patient has diabetic retinopathy or not. The algorithm will be created using example patient retinal images graded by ophthalmological experts collected in an Irish healthcare setting, and will be a first study of its type in Ireland. The algorithm will thus be able to ‘learn’ which images show diabetic retinopathy and which images do not. Testing will be conducted against an ophthalmologist’s grading of a final portion of new images. Previous research has indicated that such an artificial intelligence approach using deep learning is viable to detect diabetic retinopathy with high accuracy and it is expected that the tool developed may match the accuracy of an ophthalmologist’s grading. This tool could be used to great effect to augment an ophthalmologist’s workflow or it could be used for screening by non-expert healthcare workers in Ireland or in resource-poor settings around the world where ophthalmology services may be extremely limited.
