Fig. S6

Computation of CCFs on regions of interest of variable size

(a) Raw, confocal raster-­scanned xy-­image acquired by detecting the emitted signal of the fluorescent protein DsRed expressed by erythrocytes in the posterior caudal cardinal vein of a 3 d.p.f. Zebrafish embryo (we note that it is not the same of Figure 2 in the main text). λ_exc = nm, detection bandwidth=575-650 nm, f_line = 1400 Hz, δx = 0.08µm, γ = 0 °; scale bar = 10 µm intensity calibration bar on the right in arbitrary units. Starting from this 512x512 image, ROIs of variable size have been selected for the CCFs derivation, keeping the center of the regions of interest coincident with the center of the image. (b) Experimental CCFs computed on the largest useful ROI (512x274 pixels) selected in the image shown in a, for increasing column distances (J - I)=10, 30, 60, 90, 120, 150 pixels. The global fit, performed according to eq. (S.23), provides a flow speed |v|^ref = 1260 ±3 µm/s. (c),(d) CCFs derived for J - I =60 pixels (panel c) and J - I=30 pixels (panel d), on ROIs having a fixed N_y=274 and a variable x-size in the range 31-512 pixels. (e) CCFs computed for a fixed column distance J - I = 30 pixels, on ROIs having a fixed x-size of 512 pixels and a variable y-size in the range 40-­274 pixels. (f) A size reduction along the y-direction is combined to a reduction along the horizontal x-axis of the image: CCFs are shown for a fixed column distance J - I=30 pixels, a fixed ROI y‐size of 64 pixels and a variable x-size in the range 31-­512 pixels. The effect produced by a reduction of the horizontal x-­size of the region of interest is independent on the number of pixels along the vertical direction (see panel d for comparison): even for a small N_y=64, an average over ten pairs of columns provides sufficient statistics for the recovery of the CCF. (g) Estimates (filled squares) of the flow speed v obtained from the fit of the CCFs for J - I=60 pixels, shown in panel c: for a x-side between 70 and 512 pixels, the drift speed can be measured with a small discrepancy (|v|-|v|^ref)/|v|^ref = 0.02 with respect to the reference value (blue line) provided by the largest analyzed ROI. The estimates of the speed v (filled circles) provided by the global fit of the CCFs computed, for each of the ROIs, for increasing column distances in the range [10, (N_x− 10)] pixels (10-­pixels step) are also shown for comparison (we remark the similarity with the strategy adopted throughout the whole main text for the speed measurement, based on multiple column distances and on the global fit of all the corresponding CCFs). (h) Uncertainty σ_G on the peak value of the CCF as a function of the number of cross-­correlation functions averaged for each value of the ROI x-size. Data refer to N_y=64 and J - I=30 pixels (filled circles, pink), to N_y=274 and J - I=30 pixels (filled circles, blue) and to N_y=274 and J - I=60 pixels (open squares, blue). (j) Uncertainty σ_G on the peak value of the CCF as a function of the number of pixels N_y along the vertical side of the ROI, for a fixed N_x=512.