Evaluation of a modified autocorrelation method when applied to cardiac strain rate imaging

Abstract

When estimating tissue velocities using the con- ventional autocorrelation method (AM), the ultrasound center frequency is assumed constant and equal to the demodulation frequency, while in the modified autocorrelation method (MAM) the former is continuously estimated together with the phase shift from pulse to pulse. The AM is unbiased only if the demodulation frequency is equal to the received center frequency. This assumption is not necessarily valid. The purpose of this work has been to evaluate the performance of the MAM compared to the AM, when applied to cardiac strain rate estimation. Both methods have been implemented and evaluated in two different environments; simulated RF-data from an analytic model of the human heart and experimental data acquired from human hearts scanned with a specially programmed Vivid 7 scanner (GE Vingmed Ultrasound). When applied to simulated RF-data, the MAM demonstrated superior performance with respect to estimator variance and bias. Both methods displayed a high degree of correlation with the true strain rate of the analytic model. When estimating strain rate using experimental data, a previously presented spectral strain rate method was used as a reference. In this case the MAM did not show superior performance. On the contrary, the AM demonstrated a lower estimator variance and bias compared to the MAM.