Clutter-Generating Phantom Material. Part II: Utilization in the Comparison of Conventional and Regularized...

A particular challenge in clinical ultrasound imaging is acoustic clutter, which arises from the heterogeneity of the speed of sound and reverberations between layered tissue types. Although clutter is common when imaging through complex tissue such as the abdominal wall, few studies have investigated its effects on quantitative ultrasound (QUS) parameter estimation. The ultrasonic attenuation coefficient (AC) has shown promise as a biomarker for multiple applications. Recently, multiple regularized methods of AC estimation have been developed; however, their performance must be evaluated in clinically relevant scenarios such as in the presence of clutter. In a companion paper to this work, a material that produces clutter similar to that seen in clinical imaging was developed and characterized. Here, we utilize this clutter-generating phantom material to compare the bias and variance of AC estimates resulting from a conventional estimation method known as the spectral difference method (SDM) and a regularized method known as Analytical Global Regularized Backscatter Quantitative Ultrasound (ALGEBRA), which can either be implemented in 1D or 2D. A B-mode, target-based contrast-to-noise ratio was used to quantify the amount of clutter in data collected from a phantom with known AC. Estimation reliability was determined using the normalized root mean square error (NRMSE) and the percent bias. On average, 1D-ALGEBRA had a 22.86% smaller bias and a 32.19% smaller NRMSE than the SDM, while 2D-ALGEBRA had a 17.59% smaller bias and a 25.66% smaller NRMSE than the SDM. An analysis of variance model indicated that ALGEBRA is more robust to the presence of clutter than the SDM. Further statistical tests showed that the reduction in variance resulting from ALGEBRA was the main contributor to the reduction in NRMSE. This work demonstrates the utility of this clutter-generating phantom material in objective testing of QUS parameter estimation, as well as performance improvements obtained in phantoms with regularized methods from QUS parameter estimation in the presence of acoustic clutter.