Description

Processing of eye fundus images for assistance towards diabetic retinopathy diagnosis. Diabetic Retinopathy (DR) is a visual complication of diabetes characterized by damage in the retina’s blood vessel (reddish lesions) and the appearance of retinal lesions (hard exudates). It’s important to diagnose the disease during its first phases, when the treatment is more effective. However, the diagnosis in the first stages is not an easy task, since the patient does not notice any visual symptoms. In order to assure an early DR detection, it would be convenient to perform regular ophtalmological tests on every diabetic patient as a part of the diagnosis protocol. The increase in diabetes cases in current society would mean a large generation of eye fundus images for experts to check. This, together with the lack of ophtalmologic specialists, causes the time to get a retinography clinic evaluation to increase. Due to all of this, the automatic detection of DR lesions in eye fund images could mean an important tool to help in the disease diagnosis. Within this line of research, an application software which contributes to significantly reduce blindness cases due to DR wants to be developed. It is supposed to make an automatic reading of eye fund images obtained by non-midriatic fundus cameras. This way, DR related lesions will be identified and the level of severity will be established by detecting those cases at risk of blindness. Values of sentitivity superior to 80% and 95% will be pursued, which are the ones recommended by the World Health Organization (WHO).

Other information