An Artificial Intelligence Driven Approach for Classification of Ophthalmic Images using Convolutional Neural...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Early detection of eye diseases is crucial for timely treatment and mitigating vision loss.
• Deep learning (DL) models, particularly Convolutional Neural Networks (CNNs), are increasingly utilized for analyzing clinical images in ophthalmology.
• Existing DL algorithms like DenseNet, ResNet, and VGG-16 show promise for disease detection.

Purpose of the Study:

• To develop and assess a novel ensembled deep learning CNN model for classifying retinal color fundus images (RCFIs).
• To evaluate the model's performance in identifying specific ocular diseases: cataract, glaucoma, and diabetic retinopathy.
• To determine the diagnostic potential of the model as a screening tool for these conditions.

Main Methods:

• The study involved creating an ensembled deep learning CNN model by augmenting the VGG-16 architecture with additional convolutional layers.
• The model was trained and evaluated on a dataset of shuffled RCFIs exhibiting features of various ocular diseases.
• Performance metrics focused on classification accuracy and diagnostic potential for binary disease detection.

Main Results:

• The proposed model, an enhanced VGG-16 with added convolutional layers, demonstrated significantly improved performance.
• The model achieved a high accuracy of 98% (p<0.05) in classifying RCFIs.
• The enhanced model showed good diagnostic potential for detecting cataract, glaucoma, and diabetic retinopathy.

Conclusions:

• The developed deep learning model is accurate and suitable for integration into clinical decision support systems in ophthalmology.
• The model's high accuracy and diagnostic potential support its use as an early screening tool for common eye diseases.
• This research highlights the value of advanced DL techniques in improving ophthalmic diagnostics.