Automatic detection of microaneurysm based on the slant stacking

Abstract

The automatic detection of microaneurysms in eye fundus images can be used by medical personnel to reduce the time of analysis, permitting to cope with the high volume of exams necessary for screening the diabetic retinopathy. The goal of this work is to explore the Slant Stacking formulation of the Radon transform to automatically detect microa-neurysms on retinographies. This formulation of the Radon Transform exhibits interesting properties, namely, the invariance of the shape on the Radon domain which is explored in our proposal. The results obtained on the Di-aretDB1 with this algorithm were 89.46%, 84.16%, 84.16% for sensitivity, specificity and accuracy, respectively, while the area under the ROC was 0.83.