Abstract
Purpose :
Ellipsoid zone (EZ) reflectivity has potential to serve as a biomarker of photoreceptor integrity, but there is variability in the methods used to extract reflectivity values. Here we evaluate normalization methods of EZ reflectivity across four different OCT devices.
Methods :
Horizontal 6mm line scans (20 averaged B-scans) were acquired in five control eyes using the Cirrus, Bioptigen, Optovue Avanti, and Spectralis OCT devices. Eight longitudinal reflectivity profiles (LRPs; each 7 pixels wide by 250 pixels in depth) were extracted using ImageJ from each OCT between 0.5–2.0mm (0.5mm intervals) from the fovea. Boundaries of the ganglion cell and inner plexiform layers (GCIPL) and the peak and full width half max (FWHM) of the EZ were identified. Intensity values between the GCIPL boundaries were averaged. Intensity values between the EZ FWHM were averaged, and peak EZ intensity was extracted. Values were averaged across LRPs for each scan. The average intensity of a 300x50 pixel vitreous region was also extracted from each scan. Four intensity normalization methods were evaluated with ICC and Friedman’s test: 1) peak EZ, 2) EZ FWHM, 3) GCIPL normalization (EZ FWHM/GCIPL), and 4) vitreous normalization (subtract vitreous from 3 before taking ratio).
Results :
The ICC for mean peak EZ (-0.11) and mean EZ FWHM (-0.09) intensity across devices were poor. Friedman test revealed significant differences in peak EZ intensity across devices (Q(3)=86.2, p<0.0001). Dunn’s multiple comparisons test indicated peak EZ differed across all devices except between Optovue vs Spectralis. Mean EZ FWHM across devices also significantly differed (Q(3)=66.0, p<0.0001), and Dunn’s test revealed values from the Bioptigen were significantly different from the other devices (p<0.0001). The ICC for GCIPL normalization across devices was poor (-0.21), and Friedman test revealed significant differences across devices (Q(3)=13.6, p=0.0001), including Spectralis vs Bioptigen (p=0.009) and Spectralis vs Optovue (p=0.018). The ICC for vitreous normalization across devices was poor (0.042), with significant differences revealed by Friedman test (Q(3)=11.2, p=0.002), specifically between Bioptigen vs Optovue (p=0.009).
Conclusions :
Our data demonstrate that the normalization method alters the relationship of EZ reflectivity across devices. Such variability poses challenges to utilization of EZ reflectivity as a biomarker in clinical studies.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.