Peripapillary Atrophy Detection in Fundus Images Based on Sectors with Scan Lines Approach

Abstract

Peripapillary atrophy (PPA) is one indication of glaucoma and myopia that can be examined in color fundus images. Glaucoma is an eye disease that causes the second largest loss of vision in the world, while myopia is generally suffered by adults and even children. PPA needs to be detected so that both diseases can also be detected earlier. In this work, we propose a method of features extraction and classification for automatic detection of PPA. In the features extraction stage, the texture features are statistical values generated from the extraction area. The extraction area is divided into two areas: no-PPA and suspect-PPA. The no-PPA area captures the area with no PPA, while the suspect-PPA captures the area where PPA may be present. In order to detect the presence of PPA in the extraction area, we propose a knowledge base technique for classification. Our method obtains the accuracy of 0.92, 0.90, and 0.89 on 155 images from three different datasets.