Automatic Classification of Arterial and Venous Flow in Super-resolution Ultrasound Images of Rat Kidneys

Abstract

Velocity is one of the clinically interesting parameters. A number of studies have shown the ability of super-resolution (SR) ultrasound imaging to visualize this parameter using velocity maps. However, manual separation of the velocity estimates for arteries from veins can be quite demanding. This study used the anatomical knowledge of rat kidneys for automatic classification of arterial and venous blood velocities in SR images and measured their variations in the medullary regions of four healthy Sprague-Dawley rat kidneys. The measurements were conducted using a modified bk5000 scanner (BK Medical, Herlev, Denmark) and a BK 9009 linear array probe with a pulse amplitude modulation scheme. Ten minutes of acquired B-mode and contrast images with frame-rate of 54 Hz were processed using a SR processing pipeline. The micro-bubble trajectories were filtered using coarse anatomy labels for classification of arterial and venous flow. The velocity estimates of separated arterioles and venules of the outer medulla showed separation of data in all rats. A Wilcoxon test showed that this difference was statistically significant (p=0.002). Considering the sample size for this study, the t-distributions predicted that the median velocity in the OM arterioles and venules were in the range of 0.84 ± 0.09 mm/s and 0.70 ± 0.07 mm/s with 95% confidence. The result showed how the blood flow in outer medulla arterioles and venules of rat kidneys can be automatically distinguished using the known anatomical information about the renal vasculature.