Abstract
Cerebral blood vessels change diameter with each heartbeat, in response to oxygen demand, and during pathological states. To study these dynamics, vessel diameters are often quantified from intravital microscope images. We compare three different image-based vessel diameter measurement algorithms: one that uses an intensity threshold (cross section threshold (CST)), one that performs a Radon transform then calculates its full width at half maximum (FWHM), and a newly developed algorithm that leverages the intensity gradient (find image edges (FIE)). Using synthetic data, we find that the average errors in FIE, CST, and FWHM are 3.7%, 2.4%, and 0.2%, respectively, for images with a low noise-to-signal ratio (NSR = 0.07). Of the three, FIE best maintains accuracy for images with larger NSR and brightness gradients, though a modification to FWHM allows it to maintain accuracy with larger NSR. We also introduce a novel approach to quantify the vessel pulsation amplitude based on the interquartile range (IQR) and find it to be as accurate as an existing method, phase averaging (PA). FIE is best for measuring relative changes in vessel diameter and the average diameter in the presence of noise or other lighting artifacts, and FWHM is best for measuring the average diameter.