Diabetic retinopathy detection at early stage is critical and ignorance could lead to loss of vision also. Diabetic retinopathy detection at early stage is the prime objective of this study. To detect the diabtetic retinopathy at early stage, deep learning mechanism is collaborated with the Gaussian filtering and multi support vector machine. Multi Support vector machine is used at classification phase. Simulation is conducted within MATLAB 2017 using image processing toolbox. Operation is performed in coloured fundus images. To tackle the problems of large image set, deep learning mechanism is merged along with genetic approach. Layers are defined including input, processing and output layers along with filtering layer. Once output is generated using this network, it is fed into the CNN for segmentation. Genetic algorithm with MSVM is used at classification section to improve the accuracy of overall result. Classification results shows improvement by 2 % in terms of specificity, accuracy and error rate.
