Diabetic Retinopathy Detection and Stage Classification Using Lenet-5 Architecture

EasyChair Preprint no. 10333

Abstract

Early diagnosis and treatment of diabetic retinopathy are made possible through retinal screening.To facilitate the screening process, we are developing a deep learning system to detect and classify diabetic retinopathy.If it is untreated it may lead to other serious eye condition like macular edema and glaucoma ultimately causing vision loss.In contrast to manually created features, we employed the convolutional neural networks (CNN) model i.e Lenet-5 architecture to automatically extract features.Techniques like edge detection,thresholding and data augmentation are applied to the model during preprocessing.The efficiency of the system is assessed using statistical metrics like sensitivity (SE),specificity(SP),F-measure,and classification accuracy.The model achieved an average accuracy of 97%,recall of 22% and f1-score of 65%.

Keyphrases: Classification, deep learning, LeNet

@Booklet{EasyChair:10333,
  author = {K Priyanka and C Ananya and M Gowrilatha and K Asma and Ankita Awanty},
  title = {Diabetic Retinopathy Detection and Stage Classification Using Lenet-5 Architecture},
  howpublished = {EasyChair Preprint no. 10333},

  year = {EasyChair, 2023}}