A Deep Learning-Based Framework for Retinal Disease Classification.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate and early detection of retinal diseases is crucial for effective treatment.
• Current diagnostic methods can be time-consuming and require specialized expertise.

Purpose of the Study:

• To develop and evaluate an artificial intelligence model for the automatic detection and classification of retinal diseases.
• To utilize deep learning, specifically the VGG-19 architecture with transfer learning, for analyzing optical coherence tomography (OCT) images.

Main Methods:

• A customized 19-layer deep convolutional neural network (VGG-19 architecture) was employed.
• Transfer learning was utilized to enhance the model's learning capabilities.
• The model was trained on a dataset of 84,568 OCT retinal images across four classes: choroidal neovascularization, drusen, diabetic macular edema, and normal.

Main Results:

• The proposed VGG-19 model achieved a classification accuracy of 99.17%.
• The model demonstrated high performance with 0.995 specificity and 0.99 sensitivity.
• Statistical evaluations, including ROC curve analysis and confusion matrix, confirmed the model's effectiveness.

Conclusions:

• The VGG-19 architecture combined with transfer learning is a highly effective technique for automated retinal disease detection.
• The developed AI model shows significant potential to improve the diagnostic process for retinal conditions.