Medical Doctor & ph.d.-student Sebastian Dinesen Department of Ophthalmology
Projektet i tal
OPEN undersøgelse/kliniske data
Forventet # af deltagere
Inkluderet antal deltagere
Inkluderede deltagere med prøver
Deep-learning in diabetic eye screening: detection of sight-threatening proliferative diabetic retinopathy and prediction of long-term diabetic complications.
Aims and Outcomes
Deep-learning (DL) is a form of artificial intelligence based on folded neural networks and the gold-standard method for automatic image analysis. We aim to (A1) build a DL-model able to detect sight-threatening active proliferative DR (PDR), and (A2) to validate this on a multi-ethnic cohort from Singapore. Further, we will (B1) build a DL-model from fundus photos that can predict the development of PDR, diabetic nephropathy or -neuropathy and cardiovascular disease (CVD), as
Diabetic retinopathy (DR) is the most common complication to diabetes. A Danish national screening programme has been implemented, but manual grading of retinal images is a strenuous task, and automatic detection of patients in need of treatment would improve diagnostic accuracy and free up health-care resources.
Description of the cohort
Patients with diabetes who attended the Danish screening program for diabetic retinopathy.