Compensation for reflectance variation in vessel density quantification by optical coherence tomography angiography

PURPOSE. To compensate for reflectance variation when quantifying vessel density by optical coherence tomography angiography (OCTA). METHODS. Healthy participants received 636-mm macular and 4.534.5-mm optic nerve head (ONH) angiography scans on a 70-kHz spectral-domain optical coherence tomography system. The split-spectrum amplitude-decorrelation angiography (SSADA) algorithm was used to compute the OCTA signal. Mean reflectance projection and maximum decorrelation projection were used to create en face OCT and OCTA images. Background OCTA noise in static tissue was evaluated in the foveal avascular zone (FAZ). Vessel density was calculated from en face retinal OCTA that was binarized according to a decorrelation threshold. RESULTS. The average retinal decorrelation noise in the FAZ was linearly related to the average logarithmic-scale OCT reflectance signal. Based on this relationship, a reflectance-adjusted decorrelation threshold equation was developed to filter out 97.5% of background OCTA noise. A fixed threshold was also used for comparison. The superficial vascular complex vessel density in the macula and ONH were significantly correlated with reflectance signal strength index (SSI) using the fixed threshold. This correlation was removed by using the reflectance-adjusted threshold. Reflectance compensation reduced population variation in 25 healthy eyes from 8.5% to 4.8% (coefficient of variation) in the macula and from 6.7% to 5.4% in the peripapillary region. Within-visit repeatability also improved from 4.4% to 1.8% in the macula and from 3% to 1.7% in the peripapillary region. CONCLUSIONS. Compensating for reflectance variation resulted in more reliable vessel density quantification in OCTA.