Foundation Models for Retina Images in Diabetic Retinopathy Classification and Cardiovascular Risk Assessment

Diabetes affects multiple parts of the body and can damage the eyes and circulatory system. Diabetic retinopathy is one complication that may lead to sight loss, and people with diabetes also face a higher chance of heart and blood vessel disease. Retinal photographs capture both the eye's small vessels and aspects of the wider cardiovascular system, making it possible to assess several health areas from one image taken without an invasive procedure.

The thesis investigated whether contemporary artificial intelligence methods can extract information of clinical value from such photographs, and how consistently they work across differing groups and imaging conditions. It points to retinal imaging as a promising source of health information while stressing that AI models must be rigorously assessed before being introduced into clinical care.