An improved algorithm based on convolution dynamic multi-parameter template for microaneurysms detection

Abstract

Diabetic Retinopathy (DR) is a serious diabetic complication which may lead to new-onset blindness or visual injury. As the smallest lesions and the earliest sign that can be observed, the screening and localization of MAs is especially important for the diabetes diagnose of early lesions. In this paper, a combination of algorithms is proposed to detect MAs accurately. In the proposed algorithm, a primary candidate set will be detected by using the convolution dynamic multiparameter template (CDMPT) matching scheme and then uses a Random Forest to obtain true MA classification from the candidate set. In this work, the proposed algorithm is tested on a public dataset. The experimental results validate the effectiveness of the new algorithm.