Measuring red blood cell velocity with a keyhole tracking algorithm

Abstract

A tracking algorithm is proposed to measure the velocity of red blood cells traveling through microvessels of tumors growing in skin flaps implanted on mice. The tracking is based on a keyhole model that describes the probable movement of a segmented cell between contiguous frames in a video sequence. When a history of movements exists, past, present and a predicted landing position define two regions of probability with a keyhole shape. This keyhole is used to determine if cells in contiguous frames should be linked to form tracks. Preprocessing segments cells from background and post-processing joins tracks and discards links that could have been formed due to noise or uncertainty. The algorithm presents several advantages over traditional methods such as kymographs or particle image velocimetry: manual intervention is restricted to the thresholding, several vessels can be analyzed simultaneously, algorithm is robust to noise and a wealth of statistical measures can be obtained. Two tumors with different geometries were analyzed; average velocities were 211±136 [ m/s] (mean±std) with a range 15.9-797 [ m/s], and 89±62 [ m/s] with a range 5.5-300 [ m/s] respectively, which are consistent with previous results in the literature.