Screening of diabetic retinopathy

Aim

The study aims to develop a software system that utilises deep-learning by fully convolutional networks (FCN) in order to perform automatic grading of retinal images to detect diabetic retinopathy (DR). This will be achieved through two main objectives:

1. Utilising more than 35,000 retinal images obtained at Odense University Hospital (OUH), we will firstly train an FCN to recognise characteristic DR lesions.
2. Secondly, we will validate this in a subset of images in order to facilitate automatic DR-grading.

Participants

• Department of Ophthalmology, Odense University Hospital
• Steno Diabetes Center Odense, OUH
• The Maersk Mc-Kinney Moller Institute, SDU
• Research Unit of Ophthalmology, Department of Clinical Research, SDU

Funding

The project has received support from PhD scholarships from Steno Diabetes Center Odense and Helse Sør-Øst in Norway.

The project is part of an ongoing collaboration between the Department of Ophtalmology at Odense University Hospital, Steno Diabetes Center Odense and the group of Professor Thiusius Rajeeth Savarimuthu at the Maersk Mc-Kinney Moller Institute, University of Southern Denmark. 

Additional studies include studies regarding deep-learning-based identification in patients with sight-threatening diabetic retinopathy and the use of artificial intelligence to facilitate grading skills in the virtual ocular learning platform VIOLA.