Improving microaneurysm detection in color fundus images by using an optimal combination of preprocessing methods and candidate extractors

In this paper, we present an approach to improve microaneurysm detection in color fundus images. This task is usually realized by candidate extraction, which is followed by a classification step. The proposed method aims to increase the number of true positives in the first phase of the microaneurysm detection process. Thus, we establish a framework for selecting an optimal combination of preprocessing methods and candidate extractors. Our investigation shows that the state-of-the-art candidate extractors provide significantly improved results, when they are optimally combined with preprocessing approaches. We show that this performance can be further increased with an ensemble formed by a globally optimal combination of the preprocessing methods and candidate extractors.