2G-5 Methods for Improvement on Liver Scatterer Size Estimates

Abstract

For ultrasound scatterer size imaging, a large estimator variance caused by statistical fluctuations in echoes from random media is an important problem. This paper reviews causes of scatterer size estimator variance when using a reference phantom to measure backscatter and discusses methods to reduce the variance.A Siemens Antares machine equipped with a VFX9-4 linear array transducer is used to acquire RF data both from a reference phantom and excised human livers. The ultrasound bandwidth is in the range from 2.5 MHz to 5.5 MHz. Scatterer sizes are estimated from the frequency dependence of the backscatter coefficient using the reference phantom method. Welch's averaging method with 20 -30 independent 10 -15 wavelength gated reference data segments provides a reasonable power spectrum. The sample power spectra were also computed using 20 -30 averages of the spectral estimates for 10 -15 wavelength gated windows which provide optimum estimation of the scatterer size. Angular compounding of partially uncorrelated scatterer size estimates acquired from different angular insonifications is one method to reduce the variance. A 2 -3 degree step size is a good choice for angular compounding with this transducer setting. Phantom studies show that the correlation between echo power spectrum estimates drops to less than 0.5 for this step size. Thus, beam steered RF data acquisition from -10 to +10 degree along with compounding plus averaging across 2 -3 uncorrelated Alines laterally was chosen for acquiring sample data.Another error source of scatterer size estimates is due to frequency dependent attenuation. In large homogeneous regions the size estimates should not change significantly with depth. Our experiments show that this criterion can be used to improve the estimates of the attenuation value.Differentiation of benign from malignant liver masses, particularly liver hemangiomas from metastases is clinically important. We are acquiring and analyze liver tumor data from excised human liver. Our experiments show that the effective scatterer diameter in human liver is 140 -190 µm over the measurement bandwidth. Preliminary in vivo results suggest that ultrasound scatterer size imaging may be an effective processing method for imaging liver hemangiomas.