Intelligent retinal disease detection using deep learning

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Retinal diseases pose a significant threat to vision, potentially leading to blindness.
• Early and accurate diagnosis is crucial for effective treatment and management.
• Automated diagnostic tools can assist ophthalmologists and improve healthcare efficiency.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated multi-class classification of retinal diseases using fundus images.
• To compare the performance of different deep learning architectures and feature extraction techniques.

Main Methods:

• A balanced dataset of fundus images was curated from multiple sources.
• Deep learning techniques, including Artificial Neural Networks (ANN) and transfer learning (MobileNetV2, DenseNet121), were employed.
• Feature extraction and dimensionality reduction were performed using Principal Component Analysis (PCA) and Discrete Wavelet Transform (DWT).

Main Results:

• The proposed deep learning model achieved a peak accuracy of 98.2% in classifying retinal diseases.
• The combination of ANN with MobileNetV2/DenseNet121 architectures, PCA, and DWT yielded optimal performance.
• The model successfully differentiated between healthy eyes and eyes affected by diabetic retinopathy, cataracts, and glaucoma.

Conclusions:

• Deep learning offers a highly accurate and efficient method for automated retinal disease classification.
• The developed model shows significant potential to support clinical decision-making in ophthalmology.
• This AI-driven approach can enhance the early detection and management of sight-threatening retinal conditions.