Artificial intelligence in glaucoma: posterior segment optical coherence tomography.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma detection and surveillance rely heavily on optical coherence tomography (OCT) imaging.
• Deep learning (DL) presents a novel approach to analyze complex OCT data for diagnosing glaucomatous optic neuropathy (GON).

Approach:

• This review summarizes recent literature on DL model applications for glaucoma detection and surveillance using posterior segment OCT.
• DL models utilize OCT-derived parameters such as retinal nerve fiber layer (RNFL), macular scans, and optic nerve head (ONH) scans, often in combination.

Key Points:

• DL models achieve high diagnostic accuracy for GON, frequently employing combined OCT parameters for a comprehensive analysis.
• While RNFL segmentation is common, DL models increasingly integrate multiple OCT parameters.
• DL models for predicting glaucoma progression are less studied but are under development.

Conclusions:

• DL models offer efficient and objective tools for glaucoma management.
• Despite AI's commercial success in other ophthalmic diseases, a DL tool for glaucoma diagnosis lacks FDA approval, potentially due to the absence of a universal OCT-based definition of glaucoma.