EyeCNN

EyeCNN classifies retinal images with convolutional neural networks to identify and differentiate Diabetic Retinopathy, Glaucoma, and Cataract for early diagnosis.

Key Features:

• EfficientNet B3 Architecture: EyeCNN uses EfficientNet B3, which outperformed eleven other convolutional networks with a testing accuracy of 94.30%.
• Convolutional Network Comparison: Twelve convolutional neural network architectures were trained and compared on the retinal image dataset.
• Comprehensive Retinal Dataset: Training data comprise retinal images labeled for Diabetic Retinopathy, Glaucoma, and Cataract to enable multi-disease classification.
• Automated Classification: The system performs automated classification of retinal images into disease categories using trained CNN models.

Scientific Applications:

• Early Disease Detection: Enables early identification of Diabetic Retinopathy, Glaucoma, and Cataract from retinal images via CNN-based classification.
• Clinical Decision Support: Provides automated classification outputs that can be used to support ophthalmologists' diagnostic workflows.
• Research and Development: Facilitates comparison of CNN architectures and development of diagnostic models for retinal disease classification.

Methodology:

Retinal images were preprocessed; twelve convolutional neural networks were trained and evaluated, with EfficientNet B3 achieving 94.30% testing accuracy using defined evaluation measures.