Abstract
Purpose :
Various studies have implicated that disruption of vascular autoregulation plays a role in the pathogenesis of glaucomatous optic neuropathy. Coefficient of variation (CoV) of OCTA intensity has been proposed to assess perfusion and hemodynamics; however, established methods rely on OCTA images from commercial OCT systems that heavily manipulate the data using proprietary processing algorithms. To this end, a new processing approach was developed to eliminate system or processing biases to perform more reliable CoV-based perfusion analysis.
Methods :
Our approach was validated using data acquired from three patient groups: normal tension glaucoma (n=10), open angle glaucoma (n=10), and control (n=10), which is the same dataset previously used for a cross-sectional cohort study to evaluate retinal microcirculation heterogeneity. The Zeiss PlexElite 9000 OCT was used to acquire a total of ten serial volumes for each eye, covering a 3 × 3 mm area centered around the fovea. The OCT volumes were fed through a custom 3D registration algorithm to compute a series of outputs which were subsequently used to register corresponding OCTA volumes. En face OCTA images were generated using a DNN-based retinal layer segmentation algorithm followed by vessel segmentation to remove background OCTA signals. Sequential volumes were compiled to produce a time-series en face OCTA stack and pixel-wise CoV was computed to visualize perfusion heterogeneity in macular circulation between patient groups.
Results :
The variability in intensity distribution of the en face OCTA image extracted from the commercial OCT machine is higher compared to images processed using our custom algorithm. While this is observed in the CoV map, the intensity variation is exaggerated and does not truly reflect perfusion heterogeneity upon comparison with the custom image stack. This can be attributed to commercial OCT processing algorithms that aim to enhance vascular contrast, rendering results less reliable for temporal analysis.
Conclusions :
The proposed processing technique in this study is capable of visualizing retinal microcirculation more reliably compared to previous methods, improving diagnostic accuracy and differentiating between disease stages using commercialized OCTA. The techniques developed in this study could also serve as a foundation for future research to advance our knowledge of retinal vascular pathogenesis.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.