P4B-5 The Role of Local Center Frequency Estimation in Doppler-Based Strain Imaging

Ultrasound frequency-dependent attenuation causes a downshift in the spectrum as ultrasound waves propagate through attenuating tissue. Therefore, assuming a constant center frequency in Doppler (autocorrelation) based strain imaging will result in displacement estimation error and ultimately affect strain values, especially when using broadband signals. Here we derive a theoretical expression, based on a Gaussian-enveloped pulse, for the strain estimation error due to errors in the assumed center frequency. The expression relates the strain estimation error to the attenuation coefficient, pulse center frequency, relative bandwidth, imaging depth, and true strain. For a 6 MHz transducer (50% BW) imaging tissue with attenuation coefficient 0.5 dB/MHz/cm, we show that the center frequency downshift introduces a range of -24% to -74% relative strain bias error from 2 cm to 6 cm depths, respectively. We support our theoretical derivations with Doppler-based strain estimation results using Field II simulations, phantom, and ex-vivo bovine liver experiments. In simulation, the localized-center-frequency-estimation (LCFE) based Doppler approach reduced the bias and variance of the strain estimates, as theoretically predicted. A Philips 6.2 MHz linear transducer was used to acquire high frame rate RF data on an elasticity phantom (0.5 dB/MHz/cm) under controlled compression. In three ROIs (5 mm axial times 13 mm lateral) centered at 2, 3, and 4 cm depths, the estimated strain values are -3.7%plusmn0.9%, -3.6%plusmn0.8%, and -2.6%plusmn0.9%, respectively, with LCFE, and -3.3%plusmn1.3%, -2.4%plusmn1.6%, and -1.6%plusmn2.0% without LCFE. Respectively, the relative biases between the two methods are -11%, -33% and -39%, close to the theoretical biases of -8%, -24% and -51%. Results on an ex-vivo liver phantom with a post RFA hard lesion show a 20% reduction in strain standard deviation at the lesion center (4.5 cm depth) when using LCFE. Experimental results show that for broa- dband ultrasound signals, LCFE is important in Doppler-based elasticity imaging to reduce the bias and variance of the strain estimates, thus providing more diagnostically accurate information.