Development and Validation of a Deep Learning System for Diagnosing Glaucoma Using Optical Coherence Tomography.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Glaucoma diagnosis relies heavily on expert interpretation of optical coherence tomography (OCT) scans.
• Developing automated diagnostic tools can improve efficiency and accessibility in glaucoma screening.

Purpose of the Study:

• To develop and validate a deep learning system for glaucoma diagnosis using OCT images.
• To evaluate the diagnostic performance of different OCT map inputs for the deep learning model.
• To compare the AI system's diagnostic capabilities with those of glaucoma specialists.

Main Methods:

• A deep convolutional neural network (VGG-19) was trained on a large dataset of OCT images from control and glaucoma patients.
• Retinal nerve fiber layer (RNFL) and ganglion cell-inner plexiform layer (GCIPL) thickness and deviation maps were utilized as inputs.
• The system's diagnostic performance was assessed using area under the receiver operating characteristic curve (AUROC) and compared against expert diagnoses.

Main Results:

• The deep learning system achieved high diagnostic accuracy, with the RNFL thickness map yielding the highest AUROC (0.987).
• Performance was robust across internal and external validation datasets, with minimal impact from axial length.
• The AI's detection patterns showed high agreement (up to 90%) with glaucoma specialists' diagnoses.

Conclusions:

• The developed deep learning system demonstrates significant potential for accurate and interpretable glaucoma diagnosis using OCT data.
• The AI's ability to identify glaucomatous damage patterns mirrors specialist assessments.
• This technology shows promise for clinical application as an AI-assisted diagnostic tool.