Reproducibility of Macular Vessel Density Calculations Via Imaging With Two Different Swept-Source Optical Coherence Tomography Angiography Systems

Abstract

PurposeTo evaluate the reproducibility of vessel density calculations using different binarization methods obtained via two commercially available swept-source optical coherence tomography angiography (SS-OCTA) systems.MethodsHealthy volunteers were imaged using two swept-source optical coherence tomography angiography (SS-OCTA) devices, PLEXElite and Triton. SS-OCTA examinations were performed using a 3 × 3-mm volume scan pattern centered on the fovea. A total of six methods were used for binarization in ImageJ, two global thresholding and four local adaptive thresholding methods. Resultant vessel density values were compared between the instruments and binarization methods. Images for 60 eyes from 30 healthy subjects were assessed by two reviewers who were blinded to the scanning system used.ResultsTwenty-two eyes were excluded due to poor image quality (17 eyes from Triton, 4 eyes from PLEXElite, and 1 eye from both instruments, P = 0.003). A final 38 eyes from 23 subjects were eligible for analysis. Each binarization method and instrument led to different median values. The coefficients of variation for vessel density measurements ranged from 0.3% to 2.3% and 0.6% to 4.7% for the PLEXElite and Triton, respectively. Local adaptive thresholding methods revealed higher reproducibility than did global thresholding methods for both devices.ConclusionsMacular scans with both SS-OCTA instruments showed good reproducibility for vessel density measurements. PLEXElite recorded fewer poor images and had higher reproducibility than did Triton. These findings will inform the selection of proper binarization methods for the clinical detection of vascular diseases affecting the central retina.Translational RelevanceThe reproducibility for macular vessel measurements with SS-OCTA instruments was good. PLEXElite recorded fewer poor images and had higher reproducibility than did Triton.

Normative database and determinants of macular vessel density measured by optical coherence tomography angiography

“…Such reference data can then be compared with measurements made in a variety of retinal diseases to detect different anomalies ideally in early disease stages. 28 Our data reveal mean VDs in the foveal subfield of 22.1% ± 5.0% and 19.9% ± 6.3% in the SCP and DCP, respectively, the difference being significant (P < .001). 5,6,[20][21][22][23][24] In this study, a commercial SS-OCTA device equipped with new software was used to quantify VD in the macula.…”

“…27 Other authors have also reported good reproducibility for the use of different OCTA devices (RTVue Avanti, Fremont, California; ICC ≥ .873) 6 or the Plex Elite (Carl Zeiss Meditec, Jena, Germany; ICC ≥ .70). 28 Our data reveal mean VDs in the foveal subfield of 22.1% ± 5.0% and 19.9% ± 6.3% in the SCP and DCP, respectively, the difference being significant (P < .001). Using the spectral domain (SD)-OCTA RTVue-XR Avanti (Optovue, Fremont, California) with external software, Shahlaee et al also noted a higher VD in 122 healthy subjects for the foveal region at the level of the SCP than DCP (32% ± 3.2% vs 27% ± 5.2%; P < .001).…”

Glaucomatous vertical vessel density asymmetry of the temporal raphe detected with optical coherence tomography angiography

“…Yarmohammedi A et al 20,39 and Takusagawa HL et al 29 used a spectral-domain OCTA device and Takusagawa HL et al 29 analysed the macular microvasculature using a 6 mm ×6 mm OCTA image. Moreover, our previous study showed that the difference in the OCTA device used and the vessel density quantification method (binarisation method) affected the VD quantification 40 .…”

Automated Quantification of Retinal Microvasculature from OCT Angiography Using Dictionary-Based Vessel Segmentation

“…Two groups compare available macular vessel density algorithms [18,20]. Shoji et al [20] assess binarization algorithms in ImageJ (National Institutes of Health, Bethesda, MD, USA) on data obtained from two OCTA systems, while Rabiolo et al [18] compare seven previously published quantification algorithms, concluding that the estimated densities are significantly di↵erent for each algorithm.…”

“…Two groups compare available macular vessel density algorithms [18,20]. Shoji et al [20] assess binarization algorithms in ImageJ (National Institutes of Health, Bethesda, MD, USA) on data obtained from two OCTA systems, while Rabiolo et al [18] compare seven previously published quantification algorithms, concluding that the estimated densities are significantly di↵erent for each algorithm. Both studies therefore recommend using the same quantification algorithm in longitudinal follow-up, and note that vessel density is dependent on both the OCTA device, the acquisition size, and the post-processing algorithm, and hence comparisons to other studies or databases should be done with care.…”