Abstract
Purpose :
To remove decorrelation signal due to bulk motion (BM) while preserving true flow signal in optical coherence tomography angiography (OCTA).
Methods :
Four volumetric OCTA scans of the macular region were acquired from a healthy human subject using a wide-field 200 kHz swept-source OCT system. The scan pattern consisted of 800x399 transverse pixels covering an 8x6mm2 area. Two B-scans at each raster position were used to compute flow signal using the split-spectrum amplitude-decorrelation angiography algorithm. Each B-frame was divided into 5 segments for regression analysis. Within each segment, the first 15 percentile of A-lines (representing bulk tissue) with lowest flow signal were used for regression analysis of decorrelation (D) vs. logarithm of reflectance. The regression analysis provided a reflectance adjusted threshold for distinguishing flow from BM voxels, as well as the estimated BM velocity. The BM velocity was subtracted from the vascular voxels using a nonlinear model that related D and velocity in laboratory blood flow phantoms. The effectiveness of the algorithm was compared with an earlier method in which the median decorrelation value in each B-frame was subtracted from all voxels in the B-frame.
Results :
The step of filtering out BM voxels improved the contrast between capillaries and background and the step of subtracting BM velocity from vascular voxels further removed line artifacts (Fig.1). Compared to the median subtraction algorithm (Table 1), the regression-based BM subtraction algorithm removed a larger percentage of D noise from the foveal avascular zone (p<0.01), achieved a greater improvement in vessel density measurement repeatability, and better signal to noise ratio for flow detection (p<0.01). Two methods preserved vascular continuity (p>0.05).
Conclusions :
The regression-based BM subtraction algorithm appeared to more completely remove BM noise from OCTA compared to the median subtraction algorithm. This could improve image interpretation by reducing line artifacts and make quantification of vessel density more accurate.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.