EfficientNetB0-Based End-to-End Diagnostic System for Diabetic Retinopathy Grading and Macular Edema Detection

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss.
• Accurate and timely diagnosis is crucial for effective management.
• Current diagnostic methods can be time-consuming and require specialized expertise.

Purpose of the Study:

• To develop and validate a deep learning system for automated diagnosis of DR and diabetic macular edema (DME).
• To utilize fluorescein angiography (FFA) images for rapid and accurate disease detection.
• To improve the efficiency and accuracy of DR diagnosis in clinical practice.

Main Methods:

• A two-stage deep learning model using EfficientNetB0 was developed.
• The model was trained and validated on 19,031 FFA images from 2753 patients.
• Performance was evaluated using accuracy, AUC, precision, recall, F1-score, and Cohen's kappa coefficient. Grad-CAM was used for interpretability.

Main Results:

• The first stage achieved an accuracy of 0.7036 and AUC of 0.9062 for DR classification.
• The second stage achieved an accuracy of 0.7258 and AUC of 0.7530 for DME detection.
• Grad-CAM visualization enhanced the interpretability of the model's decisions.

Conclusions:

• An end-to-end deep learning system for DR and DME diagnosis using FFA images was successfully developed.
• The system automates image grading and DME detection, reducing interpretation time.
• This tool offers improved efficiency and accuracy for clinicians diagnosing DR.