Antal Bálint, Lázár István, Hajdu András, Török Zsolt, Csutak Adrienne, Peto Tünde
Faculty of Informatics, University of Debrecen, 4010 Debrecen, POB 12, Hungary.
Annu Int Conf IEEE Eng Med Biol Soc. 2011;2011:5943-6. doi: 10.1109/IEMBS.2011.6091469.
In this paper, results of a diabetic retinopathy screening experiment are presented which is based solely on the findings of a microaneurysm detector. For this purpose, an ensemble-based algorithm developed by our research group was used; this provided promising results in our earlier experiments. At its best, the 1200 image of the Messidor database is classified by this detector with a sensitivity of 96%, a specificity of 51% and achieved an AUC of 0.87. As anticipated, larger microaneurysm counts are recognized with higher level of certainty. Therefore, this approach might be expected to have good performance in relation to the severity of the disease.
