A transfer learning enabled approach for ocular disease detection and classification.

Area of Science:

• Ophthalmology and Medical Imaging
• Artificial Intelligence in Healthcare
• Computer Vision for Disease Detection

Background:

• Ocular diseases present diagnostic and treatment challenges.
• Deep learning shows promise for medical image analysis in ophthalmology.
• Accurate and efficient detection of eye conditions is crucial for patient outcomes.

Purpose of the Study:

• To introduce Ocular Net, a novel deep learning model for ocular disease detection and classification.
• To evaluate the performance of Ocular Net on a large dataset of ocular images.
• To compare Ocular Net's performance against existing methods for ocular disease diagnosis.

Main Methods:

• Utilized a dataset of 6200 ocular images, with 70% for training and 30% for testing.
• Developed Ocular Net incorporating transfer learning, average pooling, Clipped ReLU, and Leaky ReLU.
• Employed data augmentation techniques to enhance model performance and prevent overfitting.

Main Results:

• Ocular Net achieved 98.89% accuracy with a 0.12% loss value.
• The model demonstrated superior performance compared to previous methods in detecting and classifying ocular diseases.
• Performance was evaluated across different training/testing ratios and parameters.

Conclusions:

• Ocular Net is a highly accurate and efficient deep learning model for diagnosing ocular diseases.
• The model shows significant potential to enhance the accuracy and efficiency of clinical eye disease diagnosis.
• Further research can explore broader applications of Ocular Net in ophthalmology.