Comparison of Measurement Techniques for Photoreceptor Loss in Geographic Atrophy

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Ellipsoid zone (EZ) loss is a key endpoint in geographic atrophy (GA) clinical trials.
• Accurate quantification of EZ and retinal pigment epithelium (RPE) loss is crucial for developing AI models for this biomarker.

Purpose of the Study:

• To compare segmentation and edge detection methods for quantifying EZ and RPE loss using optical coherence tomography (OCT).
• To evaluate the impact of measurement methods on EZ quantification in GA.

Main Methods:

• Analysis of OCT images from 50 eyes with GA.
• Comparison of segmentation (OCT Explorer) and edge detection (3D Slicer) techniques.
• Evaluation of longitudinal progression in 15 eyes over one year.

Main Results:

• Both methods showed high agreement for RPE loss (P = 0.53) and longitudinal changes (EZ P = 0.43, RPE P = 0.23).
• Segmentation yielded significantly larger EZ loss areas (P < 0.001) and EZ/RPE ratios (P < 0.001) compared to edge detection.
• Edge detection provided more conservative EZ boundaries, excluding areas over drusen.

Conclusions:

• Measurement methodology significantly influences EZ quantification in GA.
• Systematic differences exist between segmentation and edge detection, particularly for EZ measurements.
• Standardized ground truth development for AI requires careful consideration of chosen quantification methods.