Describing the Structural Phenotype of the Glaucomatous Optic Nerve Head Using Artificial Intelligence

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Glaucoma diagnosis relies on identifying structural changes in the optic nerve head (ONH).
• Accurate assessment of ONH morphology is crucial for early glaucoma detection and management.
• Current diagnostic tools may have limitations in capturing subtle structural alterations.

Purpose of the Study:

• To develop a novel deep-learning approach for describing the structural phenotype of the glaucomatous optic nerve head (ONH).
• To create a robust artificial intelligence tool for glaucoma diagnosis using optical coherence tomography (OCT) imaging.
• To identify novel structural biomarkers associated with glaucoma.

Main Methods:

• A deep-learning network was trained to segment neural and connective tissue layers of the ONH from OCT images.
• A customized autoencoder network with a parallel classification branch processed segmented images.
• Principal component analysis (PCA) was used to analyze latent space parameters and their impact on ONH morphology.

Main Results:

• The deep-learning model achieved a diagnostic accuracy of 92.0 ± 2.3% for glaucoma detection.
• Sensitivity reached 90.0 ± 2.4% at 95% specificity, with a Dice coefficient of 0.86 ± 0.04 for image reconstruction.
• Altering principal component magnitudes revealed morphological changes associated with glaucoma progression.

Conclusions:

• The developed deep-learning network effectively describes the ONH structural phenotype in glaucoma.
• The algorithm identified novel biomarkers related to clinical glaucoma observations.
• This approach offers a robust tool for glaucoma diagnosis and understanding disease mechanisms.