Diabetic Retinopathy Detection Using Image Processing and Machine Learning Techniques

Vision related eye complications are mostly seen in the working-age diabetic population and are termed as Diabetic Retinopathy (DR). Identification of this disease at an early stage helps in preventing from growing further. The study over here describes the application of Convolutional Neural Networks (CNN) on concealing fundus pictures for predicting DR. Furthermore, the paper examined multinomial order diseases and estimated all the mistakes that were obtained because of misclassification of diseases or due to SVM and strategic relapse failure to recognize unpretentious diseases. Moreover, it has been found that pre-treatment methods in contrast with the adaptable histogram levelling procedure gave better results and assured dataset constancy by final check of multinomial order models. Furthermore, the paper developed the acknowledgement of unremarkable key points. Moreover, when learning is done on pre-trained ImageNet models like GoogLeNet, and AlexNet, the correctness of the characterization models has been seen as <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{7 5 \%, ~} \mathbf{6 9 \%}$</tex>, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{5 7. 2 \%}$</tex> for <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 - a r y, ~} \mathbf{3 - a r y}$</tex>, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4 - a r y ~ m o d e l s ~}$</tex> respectively. The proposed model is an SVM-based model and it analyses and integrates the various posterior layers of the retina called fundus images. The results show that this proposed model can revolve around various disease stages. In this way, the model makes estimates with 90 % correctness. The proposed neural model accomplished the test with an approval affectability of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 4 \%}$</tex>.