Detection of Diabetic Eye Disease in a Danish Context using Deep Learning

Research output: ThesisPh.D. thesis


Diabetes is a global epidemic with more than 460 million people affected across the world. This number is expected to rise dramatically in the coming decades, as people on average live
longer into old age and our lifestyles become increasingly sedentary. Diabetes leads to high blood pressure which is the cause of many complications in organs such as the heart, kidneys
and eyes. Diabetic retinopathy (DR) is one of the most frequent complication in diabetes. It is a chronic and progressive disease in the eye where microvascular changes brought on by high blood pressure and blood glucose can lead to reduced vision and blindness, and it is a leading cause of preventable blindness in the working aged population.

The risk of DR related vision loss and blindness can be decreased by early detection of the disease. Diabetes patients in many countries attend regular screenings at hospitals or healthcare clinics, where ophthalmologist or trained health professionals will examine their eyes using fundoscobic images and look for microvascular changes and abnormalities indicating disease.
As the incidence of diabetes continues to rise, the demand for expert healthcare professionals also increases, and there is a fear that healthcare systems will not be able to keep up with demand.

Within the last five years the interest in and research into the use of computer assisted diagnostic methods have increased, in large part due to the performance of deep learning methods and specifically convolutional neural networks for automatic image recognition. Deep learning has been extensively studied for use in automated screening and diagnosis of DR and is currently on the level of human experts for detecting certain severity levels of the disease.

Denmark has an established DR screening program that is tasked with regular screenings of the about 300,000 people in Denmark with diabetes. The Danish screening program is build on recommendations and guideline established by The Danish Ophthalmological Society. These guidelines propose the use of specific methods for identifying and stratifying DR disease severity and managing patients to reduce the risk of DR related vision loss.
Translated title of the contributionDetektion af Diabetisk Øjensygdom i en Dansk Kontekst ved brug af Deep Learning
Original languageEnglish
Awarding Institution
  • University of Southern Denmark
  • Savarimuthu, Thiusius R., Principal supervisor
  • Grauslund, Jakob, Co-supervisor
Publication statusPublished - 1. Feb 2022

Note re. dissertation

Print copy of the thesis is restricted to reference use in the Library.


Dive into the research topics of 'Detection of Diabetic Eye Disease in a Danish Context using Deep Learning'. Together they form a unique fingerprint.

Cite this