An Accurate Approach for Microaneurysm Detection in Digital Fundus Images

Diabetic Retinopathy (DR), a common and severe complication of diabetes, damages the retina and even causes blindness. Considering that micro aneurysm (MA) is regarded as the first clinical symptoms of DR, diagnosis of MAs in retinal images is a critical step to protect the vision of diabetes suffers in present time. This paper presents a new approach to detect MAs. The contribution of this approach is mainly twofold. First, the dynamic multi-parameter template (DMPT) matching scheme is proposed in this paper is more realistic compared to conventional schemes. It adopts double restraining factors scheme to measure the matching degree by sum of errors combining correlation coefficient. The scheme can balance the matching degree of the template and MAs between the correlation and relative error. Second, we have proposed an adaptive weighted scoring (AWS) algorithm for the MAs detection in feature extraction. It can lower false positive (FP) in case of keeping the same true positive (TP) effectively. Through extensive analysis about results, our approaches proved to be effective and efficient.