AMD diagnosis using OCT images and Diabetic diagnosis using Fundus Images in Matlab

This study presents a diagnostic framework for detecting eye diseases, specifically Diabetic Retinopathy (DR) and Age-Related Macular Degeneration (AMD), leveraging deep learning models on medical imaging. The system allows users to select either DR diagnosis using fundus images or AMD diagnosis using OCT images. For AMD diagnosis, the process begins with OCT image input and pre-processing, where images are resized for optimal feature extraction. Using ResNet50 architecture, the model classifies images as healthy or indicative of AMD. If AMD is detected, further classification identifies stages as Dry AMD, Wet AMD, or Diabetic Macular Edema (DME), using ResNet50’s robust feature layers and training options to ensure accurate results. For DR diagnosis, fundus images are resized and processed through InceptionNet V3, classifying images as either healthy or DR-positive. Upon detecting DR, InceptionNet V3 further categorizes the disease into stages: Mild DR, Moderate, Severe, or Proliferative, allowing detailed stratification. This multi-stage approach enables precise, automated categorization of each condition’s progression, facilitating early intervention and improving disease management. The model's efficacy is evaluated through classification accuracy across both tasks, demonstrating potential as a reliable support tool in ophthalmology.

Keywords: Eye Disease Dataset, Pre-Processing, Convolutional Neural Networks, Deep learning, Feature Extraction, Classification, Accuracy.