Extended Data Figure 6. Dilation measurements with circle fitting.

a, The steps of the circle fitting algorithm are illustrated for a blank and a stimulus frame corresponding to the penetrating arteriole shown in the lower panel of Fig. 1b. The raw image data (first panel) is oversampled by linear interpolation between pixels (second panel). Then a luminance threshold (a fraction of the gradient between the brightest and darkest pixel of the image) is applied (third panel). Finally a 2-D sobel filter is applied to the thresholded pixels to detect the edge of the vessel (fourth panel). The circle fit is only applied to the pixels in the fourth panel but it is overlaid on all the panels for illustration purposes. b, As the threshold is increased, fewer pixels pass the threshold and therefore the baseline diameter changes. However, the percent change in diameter across baseline and stimulus presentations (the response amplitude) and the response selectivity remain the same. Note that for vessel geometries needing an elliptical fit rather than a circular fit (see Extended Data Fig. 8c and Methods), the shorter axis of the fitted ellipse was used to estimate the vessel diameter.