Skip to navigation Skip to main content Skip to footer

Approved research

Automatic Diabetic Retinopathy Detection based on Deep Learning Using Retinal Images and Clinical Data

Principal Investigator: Dr Rodrigo Varejao Andreao
Approved Research ID: 46758
Approval date: May 14th 2019

Lay summary

Diabetes mellitus (DM) is an important and growing health problem for all countries, regardless of their degree of development. In 2015, the International Diabetes Federation (IDF) estimated that 8.8% (95% confidence interval [CI]: 7.2 to 11.4) of the world population aged 20-79 years (415 million people) lived with diabetes. If current trends persist, the number of people with diabetes is projected to exceed 642 million by 2040. One consequence of DM is diabetic retinopathy (DR). Such disease affects vision and is the leading cause of vision loss. However, the early stage of the DR is asymptomatic and when detected it could be late. Early diagnosis and control of the disease are very important to reduce the damage caused by diabetic retinopathy. Many studies have shown diabetic retinopathy detection systems based on retinal image classification alone. However, in a more realistic setting, the use of clinical data can be very relevant both to improve the accuracy of diagnostic aid systems, to help the screening of such diseases at the Primary Health Care and to understand the evolution of the disease at early stage. High-performance diabetic retinopathy detection systems could help the screening of such disease even though a specialized opinion is not present which is the case at the Primary Health Care. As a consequence, the prognosis of the disease will improve as well as the effect of the treatment. The objective of this project is to investigate whether automatic diabetic retinopathy detection can benefit from the usage of both retinal images through its features extracted by deep learning algorithms and clinical data such as weight, height, waist circumference, blood glucose, glycated hemoglobin, and others. The project should be completed in 36 months.