Improved arrival time detection for cardiac pulse transit sonomicrometry

Routine pulse transit sonomicrometry uses simple threshold detection to determine ultrasonic signal arrival time. An improved method uses the known shape of the signal´s leading edge as a 1.5 to 3 /spl mu/sec template to find the arrival time with a cross-correlation matched energy estimator procedure. Over a sliding window in the received data, the template data and it´s Hilbert transform are cross correlated with the received data. The cross correlation generates a likelihood estimate for all possible window position, each corresponding to an arrival time. Waterbath and animal transducer signals were collected using a conventional sonomicrometer connected to a digital oscilloscope and the data analyzed off-line to implement the algorithms. Likelihood estimators behaved as expected and peak detection used on the likelihood estimators gave arrival times uniformly within 100 ns of conventional threshold detection. Various amounts of noise were added to the received signal to evaluate robustness. Even with enough noise that the original signal was barely visible in a time plot, the new likelihood procedure yielded very close arrival times.