Abstract
Purpose :
Optical coherence tomographic angiography (OCTA) images have been used to measure retinal vessel density and quantify the effect of eye diseases on retinal perfusion. Another OCTA parameter, the flow index, has also been investigated. The flow index is the average value of the flow signal in an en face OCTA image. It may be a more sensitive indicator to detect changes in velocity or volumetric flow than vessel density. But, due to its strong dependence on the strength OCT reflectance signal, it is less reliable for disease diagnosis. In this work, we present a compensation method to remove this dependence.
Methods :
3×3 mm2 macular OCTA scans were acquired by a spectral-domain OCT system (AngioVue). The flow index of the superficial vascular complex (SVC) was defined as the averaged OCTA decorrelation outside the foveal avascular zone (FAZ) - defined as a circle of 0.3-mm radius centered in the fovea - excluding large vessels. Sixty images were acquired from 10 healthy eyes by attenuating the signal strength with neutral density filters (NDF). The decorrelation values of background voxels were set to zero by a thresholding algorithm that accounted for voxel reflectance, which yields a vessel density that is independent of signal strength. Then, capillary flow signal was normalized to the averaged large vessel decorrelation signal of the scan, by compensating the linear dependences of the average large vessel and capillary decorrelation on average OCT reflectance (Fig. 1).
Results :
The flow signal compensation was tested on a separate set of nine healthy eyes scanned repeatedly between two and six times under optimal imaging conditions. The dependence of flow index on OCT signal strength (Fig. 2) was reduced after applying the compensation scheme (Pearson’s R = 0.86, p<0.01 before compensation vs. R = 0.49, p < 0.01 afterwards). The coefficient of variation on the flow index in the test set was reduced by the compensation scheme from 14.8% to 10.8 %. The repeatability of the flow index, measured as the pooled coefficient of variation per eye was improved from 5.8% before compensation to 4.9% afterwards.
Conclusions :
The algorithm successfully reduced the flow index dependence on OCT signal strength and yielded a more repeatable flow index measurement. This metric has potential to describe capillary flow changes in diabetic retinopathy, retinal vein occlusion and glaucoma.
This is a 2021 ARVO Annual Meeting abstract.