Enhancing the Accuracy of Glaucoma Detection from OCT Probability Maps using Convolutional Neural Networks

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Glaucoma is a leading cause of irreversible blindness.
• Early detection of glaucoma is crucial for effective treatment and vision preservation.
• Optical coherence tomography (OCT) provides detailed retinal imaging for glaucoma assessment.

Purpose of the Study:

• To evaluate the efficacy of convolutional neural network (CNN) models for glaucoma detection.
• To compare the performance of CNNs trained on natural images versus OCT data.
• To explore the potential of AI in augmenting glaucoma diagnosis.

Main Methods:

• Development and assessment of CNN models using OCT retinal nerve fiber layer (RNFL) probability maps.
• Comparison of CNNs pretrained on natural images with those trained exclusively on OCT data.
• Analysis of model performance using receiver operating characteristic area under the curve (AUC) scores.

Main Results:

• All evaluated CNN models demonstrated high accuracy in glaucoma detection (AUC 0.930-0.989).
• CNNs pretrained on natural images performed comparably to models trained solely on OCT data.
• Attention-based heat maps indicated potential for improvement by incorporating vascular information.

Conclusions:

• CNN models show significant promise for accurate glaucoma detection from OCT RNFL maps.
• AI-powered tools can potentially assist human experts in diagnosing glaucoma.
• Further refinement of CNN models, including vascular data, may enhance diagnostic capabilities and expedite eye disease detection.