Quantification of Internal Carotid Artery Flow with Digital Subtraction Angiography: Validation of an Optical Flow Approach with Doppler Ultrasound

A and B, The regression fit at Q_I = 1.5 mL/s for Q_D ≤ 4 mL/s (left) and Q_D ≤ 7 mL/s (right). Q_X is represented on the y-axis, and Q_D, on the x-axis. The data (blue star points), fit (red line), and 95% CI limits (magenta) are all represented. C and D, The regression fit at Q_i = 2.0 mL/s for Q_D ≤ 3 mL/s (left) and Q_D ≤ 7 mL/s (right). Q_X is represented on the y-axis and Q_D on the x-axis. The data (blue star points), fit (red line), and 95% CI limits (magenta) are all represented. E and F, The regression fit at Q_i = 3.0 mL/s for Q_D ≤ 3 mL/s (left) and Q_D ≤ 7 mL/s (right). Q_X is represented on the y-axis, and Q_D, on the x-axis. The data (blue star points), fit (red line), and 95% CI limits (magenta) are all represented.

A, the boxplot for the fitted slope (y-axis) at 1.5, 2.0, and 3 mL/s (x-axis). Each boxplot represents the 5 slope values fitted for each Q_D subsample, with Q_D thresholds ≤3.0, ≤4.0, ≤5.0, ≤6.0, and ≤7.0 mL/s. B, results of the nonparametric ANOVA multiple comparison of the regression slope upper limit (x-axis) for each injection rate group (y-axis). These summarize the corresponding maximum slope values of 0.84, 1.21, and 0.72 with an SE of 0.12. The 95% CI error margins are also represented.

A, Slope versus Q_D for injection rates Q_i = 1.5 (blue), 2.0 (red), and 3.0 mL/s (green). The 95% CI limits have been superimposed to Q_i = 2.0 mL/s points. The graph shows that the slope bias calculated is not significantly different from 1 up to Q_D = 6 mL/s (P > .05). The Q_i = 1.5 mL/s group (blue points) follows the same evolution trend as Q_i = 2.0 mL/s, though with accentuated bias. A nonsignificant slope bias is likely for the Q_D range below 3 mL/s (P = .06). B, Injection rate quotients for Q_i = 1.5 mL/s (group 1), 2.0 (group 2), and 3.0 mL/s (group 3). The blue bars correspond to quotients <1.2, and the red bars, to quotients >1.2. The fact that the Q_i = 1.5 and 2.0 mL/s groups are reconstructed in similar ways with minimum slope bias, unlike the Q_i = 3.0 mL/s group, is partly due to differences in injection rate quotients (P = .01).