T he recent study by Kakihara et al. provides a compelling analysis of foveal avascular zone (FAZ) enlargement and geometric perfusion deficits (GPDd) as indicators of regional retinal nonperfusion. While their findings suggest a significant clinical role for these biomarkers, the reliance on intensive post-processing techniques—specifically image averaging and manual delineation—warrants a closer examination of the methodology’s reproducibility and the risk of data distortion.
A primary concern lies in the “image manipulation” involved in creating the averaged en-face OCTA images. The authors registered and averaged five repeat scans to enhance the signal-to-noise ratio. While this is a standard practice for improving image quality, the study lacks granular detail on the specific registration algorithms used within the Fiji software, which could lead to “ghosting” or the artificial creation of vessel-like structures if motion artifacts are not perfectly aligned.
Furthermore, the replication of this study by other centers faces hurdles due to the manual nature of several key steps:
-
•
Structural vs. Functional FAZ: The calculation of “FAZ enlargement” relies on the manual delineation of both structural OCT and functional OCTA boundaries. Even with masked graders and reported high intergrader reliability, the absence of an automated, objective segmentation algorithm limits the scalability of this metric in a high-volume clinical setting.
-
•
UWF-FA Quantification: The “ROI-free” manual tracing of nonperfusion on ultra-widefield images is inherently subjective. Although stereographic projection was used to correct for peripheral magnification, the specific software settings and the technician’s threshold for “darker underlying choroid” are difficult to standardize across different devices or clinical sites.
-
•
GPDd Calculation: The methodology defines GPDd using a threshold from the nearest perfused capillary. To replicate this accurately, researchers need clarity on how the “nearest capillary” is defined within the software—whether through skeletonized vessel maps or raw binarized images—as this choice significantly impacts the final percentage.
While the study’s findings are a valuable step toward noninvasive ischemic assessment, the transition from research to clinical standard will require a transition from manual, high-manipulation workflows to automated, robustly validated pipelines. Authors are encouraged to provide more detailed documentation of their Fiji macros and binarization thresholds to facilitate external validation.
CRediT authorship contribution statement
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
