Explainable DL Models for Automated Ophthalmic Diagnosis

Developed a comprehensive machine learning solution for automating diagnostic processes in ophthalmology using CNN architectures. This project includes the full machine learning lifecycle: data preprocessing, model architecture design, training, evaluation, and explainability. The DA-CNN and CF-CNN models were designed with enhanced feature extraction and explainability, achieving 97.4% and 96.68% accuracy respectively on retinal OCT scans for conditions such as DME, Drusen, and CNV. The models also integrate interpretability techniques like GradCam and LIME, providing visual explanations for diagnostic decisions.