My research activity concentrates in the conception of new methods for automatic analysis of retinal images and in their application to diagnosis of Diabetic Retinopathy, a severe and wide-spread eye-disease.  Our main strategy is pointed out in the abstract of my PhD thesis.

Diabetic Retinopathy is a silent disease, and people suffering from it may not be aware of their risk to become blind or their vision to become altered. Hence, in order to allow an optimal treatment of the disease, it has to be detected before vision alters. A solution to this problem may be the screening of all diabetic patients, i.e. regular eye examinations (once or twice a year) for the whole risk population. As this produces very high costs and as there exists already a lack of specialists in many areas/countries, this approach can hardly be followed without computer assistance.

Another problem in diagnosis of Diabetic Retinopathy is the quantitative comparison of images taken at successive examinations. Automatic image registration and the evaluation of the evolution of single lesions are prerequisites to that.
 

The main features in the eye are:

• The macula (dark spot in the middle of the image) is the centre of vision. All pathologies that reach this region affect vision immediately.
• The optic disc (bright spot, on the right of the macula) is the entrance of the vessels and the optic nerve. It is one of the features that is the easiest to detect. It can serve as a landmark in the image for raw registration algorithms and it its detection plays an important role for lesion detection algorithms.
• Vascular tree (reddish elongated features) supply the blood to nourish the retina. Their detection is essential for image registration algorithms, because their structure does not change from one exam to another, and their detection is also important for the elimination of false positives in microaneurysms and haemorrhages detection.

• Microaneurysms (small reddish isolated points) are tiny dilations of the capillaries. They are the first unequivocal sign of diabetic retinopathy. Their number is also a good indication of the progression of the disease. So, their detection is essential for any automatic mass screening or monitoring system.
• Hard exudates (bright features) are also developed in diabetic retinopathy, but are not caracteristic to it. However, their detection is also important, because they indicate the existence of an edema, which are one of the most feared complications of Diabetic Retinopathy.
• Hemorraghies (reddish areas, bigger than microaneurysms) indicate - if their number is very high - the existence of a local ischemia which preceeds neovascularisation, which is the second complication causing vision alteration.