Detecting glaucoma with only OCT: Implications for the clinic, research, screening, and AI development

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Optical coherence tomography (OCT) probability (p-) maps are crucial for glaucoma detection.
• Artifacts on OCT p-maps in healthy eyes can mimic glaucoma damage.
• Distinguishing these artifacts is vital for accurate diagnosis.

Purpose of the Study:

• To develop a method for glaucoma detection using only OCT data.
• To introduce a model to differentiate glaucoma-like artifacts from actual glaucomatous damage.
• To evaluate the performance of OCT-based glaucoma detection methods, including AI models.

Main Methods:

• Development of a simple anatomical artifact model based on anatomical variations.
• Application of the model to OCT retinal nerve fiber layer (RNFL) p-maps for glaucoma detection.
• Evaluation of both clinician-judged and automated detection methods, including AI deep learning.

Main Results:

• Glaucoma-like artifacts are common in healthy eyes due to anatomical variations.
• The artifact model aids in distinguishing artifacts from glaucoma.
• The model explains limitations of summary statistics and supports the success of AI models focused on RNFL p-maps.

Conclusions:

• OCT-based glaucoma detection is valuable for clinical decisions, research, and screening.
• The proposed artifact model enhances the reliability of OCT for glaucoma diagnosis.
• Further development of AI models using RNFL p-maps shows promise for automated glaucoma detection.