Skip to main content
. 2019 May 14;8:e45077. doi: 10.7554/eLife.45077

Figure 3. Automated measurement of blood velocity using Radon transform.

Figure 3.

(A) A 110 ms long space-time image of a 23.3 µm arteriole scanned obliquely, with static features removed, as described in Figure 2. Solid yellow box shows a square ROI of side 157 pixels (~15 µm x 10 ms) used to inspect single-cell velocity. Dashed green boxes show typical 75% overlap of inspection ROIs. (B) Zoomed-in version of solid yellow box in A. The ROI is circularly cropped to make the interpretation of the Radon transform easier. (C) Radon transform of the ROI in B. Local maxima in pixel intensity correspond to individual blood cell streaks in the single ROI in B. (D) Normalized standard deviation (SD) of pixel values of Radon image in C plotted as a function of orientation angle. The horizontal axis is shown both in angle space and corresponding velocity space (angle to velocity mapping given by Equation 1). Angle corresponding to peak of variance profile in D gives dominant orientation and velocity of streaks in single ROI in B. Measured cell orientation is overlaid as magenta line in B. The cell was moving at 22.3 mm s–1. To determine strength/believability of velocity measurement, a custom signal-to-noise-ratio (SNR) metric is defined as the peak standard deviation (SD) divided by the mean SD. (E) Space-time image in A with measured cell orientations in each ROI overlaid using straight lines. For visualization, only a subset of ROIs are displayed, with no overlap. Magenta lines represent ROIs which passed the custom defined SNR threshold (SNR >2.5). Notice that the magenta lines consistently report a tight range of orientation angles, as expected from normal physiology in such a small time epic. Meanwhile, cyan lines often correspond to measured orientations which seem to incorrectly report actual cell orientations, thus showing that the SNR threshold does a good job separating signal from noise. Such dense reporting of single-cell velocity enables measuring subtle fluctuations in velocity due to laminar profile and cardiac pressure wave. Velocities corresponding to measured cell orientations are displayed on a colormap, as shown in Figure 4.