Assessing the accuracy of macula detection methods in retinal images

Fundus images are valuable resource in diagnosis because they often present indications about retinal, ophthalmic, and even systemic diseases such as diabetes, hypertension, and arteriosclerosis. Processing and analysis constitutes a relevant task to help specialists in detecting eye diseases. This paper focuses on algorithms to detect macula, a fundamental structure associated with the macular degeneration, that can be recognized from fundus image. We notice two typical ways in which macula detection algorithms have been evaluated: a) one uses a particular image or a single benchmark retina image database instead of several public ones; b) one selects an ad-hoc set of metrics to perform the evaluation for the lack of a standard. In this paper, we propose a set of rules to assess macula detection algorithms, then we compare four macula detection algorithms, using three public benchmark databases, using a total of 254 images. The contribution of this work is to devise a grading scheme that allows comparing different algorithms, as well as identifying cases that are likely to contain macular abnormalities. Finally, the proposed assessment methodology splits the results into success, satisfactory and failure, such that it provides an easier entry to the specialists to address and grade macular edema disease.