A Recognition Method for Diabetic Retinopathy Based on Feature Selection

Feature selection is considered as an important data preprocessing technology because of its advantages in many aspects. In this paper, a new recognition method for diabetic retinopathy is proposed. First, we introduce the maximal information coefficient (MIC) to measure the correlation between features and the decision, and the redundancy between features. Then, by combining with this function, the mRMR algorithm is developed for searching optimal features. The classification model is used to evaluate candidate feature subsets and select the subset with the best classification performance. Finally, we employ the proposed algorithm for recognizing whether the patient has diabetic retinopathy. The experimental results show that the proposed feature selection algorithm removes irrelevant features and improves the performance of the prediction model.