Abstract
Purpose :
To develop automated analysis of ellipsoid zone reflectivity on Optical Coherence Tomography (OCT) by measuring the amplitude of the ellipsoid zone peak and the precise segmentation of disrupted zones using machine learning algorithms to accurately record the distribution of photoreceptor integrity.
A ratio-based normalization approach between the peaks of the ellipsoid zone and external limiting membrane reflectivity has been proposed to mitigate the impact of heterogeneous retinal illumination. However, existing normalisation methodologies providing a single integer value but lack insights into the spatial variability of this parameter.
Methods :
In this study, an automated technique is presented capable of generating 2D en-face maps representing normalized ellipsoid zone reflectivity. The proposed technique involves initial segmentation of the inner/outer segment junction utilizing validated machine learning algorithms tailored for age-related macular degeneration (AMD) eyes. Subsequently, defining two sub-volumes corresponding to the ellipsoid zone and external limiting membrane regions based on this segmentation. Extracting associated peaks within each sub-volume across the OCT volume yields reflectivity maps for the ellipsoid zone and external limiting membrane. Normalized ellipsoid zone reflectivity is then computed via pixel-wise division of ellipsoid zone and external limiting membrane peaks. The methodology facilitates analysis of photoreceptor integrity in AMD eyes (n=200).
Results :
Preliminary findings on the normalized ellipsoid zone positioned within 500 microns of the lesion boundaries from 13 eyes reveal a localized reduction in normalized ellipsoid zone reflectivity proximal to GA lesions. The spatial configuration of reduced reflectivity on the 200 AMD eyes will be presented to demonstrate different disease sub-phenotypes.
Conclusions :
This novel automated technique provides insight into the spatial distribution of photoreceptor integrity in AMD eyes.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.