Reproducibility of Retinal Blood Flow Measurements Derived From Semi-Automated Doppler OCT Analysis

Abstract

To evaluate reproducibility and intergrader agreement of total retinal blood flow (TRBF) measurements obtained by semi-automated grading of Doppler Fourier-domain optical coherence tomography (FD-OCT) scans. Doppler FD-OCT scans were obtained from 20 eyes of 18 subjects (10 glaucomatous, 10 normal). Scans were obtained using a circumpapillary scan protocol and analyzed using the Doppler OCT of Retinal Circulation software (version 2). Two masked, independent human graders manually refined the scans, adding or deleting vessels, changing vessel boundaries, and classifying vessels as veins or arteries. TRBF was calculated automatically by software summing flow in all veins. Agreement between various vessel parameters and TRBF values generated by the graders was analyzed. Mean difference and mean absolute difference (± standard deviation, range) for TRBF were -0.55 (± 5.37, -8.53 to 13.6) and 3.84 (± 3.70, 0 to 13.6) μL/min, respectively, with an intraclass correlation (ICC) of 0.933 and limits of agreement (95% confidence interval [CI]) of -11.1 to +10.0. Venous area measurements showed similar levels of agreement with mean difference and mean absolute difference (± standard deviation, range) of -2.91 (± 4.29, -10.95 to 6.43) and 3.59 (± 3.70, 0 to 10.9) mm(2) with an ICC of 0.933 and limits of agreement (95% CI) were -11.3 and +05.5. The agreement for vessel identification between graders was almost perfect with a weighted kappa of 0.86. Reproducible measurements of TRBF can be obtained from Doppler OCT data using semi-automated software with manual refinement. These findings should be of value in future studies evaluating retinal blood flow in various diseases.