An angle independent pattern recognition algorithm for ultrasound image blood flow estimation

The authors present a new algorithm for determining the blood flow in arteries based upon consecutive speckle pattern images from an ultrasound array device. The algorithm is composed of two parts: (i) an object recognition phase in which the speckle patterns of platelets, red blood cells, and white blood cells are identified as contiguous objects: and (ii) a tracking phase in which the maximum likelihood direction is chosen as the best candidate for the velocity vector. The algorithm is able to detect 2-D motion from B-mode images taken at 2 MHz. When 3-D array techniques become available, the algorithm can be easily adapted to detecting 3-D motion. The advantages of this algorithm over existing ones are discussed, with emphasis on accuracy, robustness to background noise, and low computational needs. Both simulated and real data tests and results are presented. Using real data, the algorithm was able to measure the blood flow velocity to 1-pixel/frame accuracy.<>