The effect of low scatterer number density on ultrasound attenuation estimation

The ultrasound attenuation coefficient has been established as a marker of thermal ablation therapy. The reference phantom method (RPM) has been proposed as a practical technique for estimating the attenuation coefficient in patients. However, tissue is frequently modeled with a high scatterer number density (SND). In this work, attenuation slope estimation was performed in simulations, modeling a linear attenuation coefficient and both high and low SND. For the numerical phantom studies, tissue was simulated having an SND ranging from 0.2 to 128 scatterers/mm³. Using a fixed kernel size of 9.6 mm axially along with 10 independent A-lines and a center frequency of 5.5 MHz, it was found that the bias in the estimate of the attenuation coefficient was no more than 0.10 dB/cm MHz, while the standard deviation increased to over 1.0 dB/cm MHz for different SND values. Attenuation slope was also estimated using the reference phantom method in five ex vivo bovine livers and another five bovine livers following intense heating in a water bath. In normal ex vivo bovine liver and heated bovine liver it was found that the received ultrasound echo signal from a Siemens 9L4 linear array transducer, operating at a center frequency of 6 MHz, exhibited a low envelope SNR, 1.10 in normal liver and 1.28 in heated liver. These results deviate from the SNR value of 1.91 expected for Rayleigh distributed envelope values. The same transducer was used for the attenuation slope estimates with a variable kernel size. As the kernel size was varied from 6 to 16 mm axially the mean attenuation slope estimate varied between 0.4 and 0.5 dB/cm MHz in unheated liver and 1.1 and 0.9 dB/cm MHz in heated liver. Meanwhile, the standard deviation in the attenuation estimates varied from 1.8 to 0.5 dB/cm MHz in unheated liver and 1.6 to 0.5 dB/cm MHz in heated liver. From the simulation and ex vivo results, we conclude that the variance in attenuation slope estimation in ex vivo bovine liver- is primarily due to the low SND. Future work in attenuation estimation should consider both high and low SND in tissue when RPM algorithms are utilized.