Relative Feature Classification Method for Diabetic Retinopathy Detection from Scanned Eye Images

Objectives: Intense/unattended Diabetic Retinopathy (DR) results in vision loss for diabetic patients for which a precise computer-aided image process is required. This article introduces a Distributive Function-Relative Feature Classification Method (DFRFCM) for identifying precise DR positive rates using scanned eye inputs. Methods: The proposed method performs a linear distributive assessment for the identifying features between the positive and false rate regions. This linear assessment is based on matching features between inputs and training sets for maximum classifications. Findings: The classification of feature relativity is the key factor used for training the neural network across various regions identified. Such a process relies on one-layered neural processing for distinguishing the distribution or closure of the features between consecutive regions. Novelty: The closure pixels are identified as false rates from which DR region detection and intensity are classified. This process improves the detection accuracy with fewer errors and has an accuracy of 91.817% and an error of 7.27% when compared with conventional methods.