Real-time blood flow estimation using a recursive least-squares lattice filter

Ultrasonic flow estimation involves Fourier-transforming data from successive pulses. The standard periodogram spectral estimate does not reflect the true velocity distribution in the blood and assumes quasi-stationarity in the data. Last year (see J.A. Jensen et al., IEEE Ultrasonics Symposium Proceedings, p. 1221-4, 1996), the authors demonstrated that a recursive lattice filter can yield results much closer to the correct velocity distribution. They have now implemented it in real time on a system with sixteen ADSP-21060 processors, interfaced to a commercial scanner. The system can perform real-time processing for both the periodogram and lattice-filter approaches and displays both results on a PC for comparison. Results are shown for phantom data and for demodulated data from the aorta and hepatic vein of a healthy subject. This demonstrates under clinical conditions that the lattice filter gives a more realistic velocity distribution and can track rapid changes in the flow.