A regularization approach for ultrasonic attenuation imaging

The quantitative estimation of the attenuation coefficient slope (ACS) has the potential to differentiate between healthy and pathological tissues. However, attempts to characterize ACS maps using pulse-echo data using methods such as the spectral log difference (SLD) technique have been limited by the large variability of the estimates. In the present work, ACSs were estimated using a regularized SLD technique. The performance of the proposed approach was experimentally evaluated using two physical phantoms: a homogeneous phantom, and a phantom with a cylindrical inclusion. The results obtained with the SLD and regularized SLD techniques were compared to the ACS values obtained with through-transmission techniques. In the homogeneous phantom, the use of regularization allowed reducing the standard deviation by more than 90% while keeping the estimation bias around 2%. For the inhomogenenous phantom, a trade-off between contrast-to-noise ratio (CNR) and estimation bias was observed. However, the use of regularization allowed nearly doubling the CNR from 0.54 to 0.97-1.29 when compared to the standard SLD, while achieving an estimation bias between 10% and 20%. The results suggest that the use of regularization methods can effectively reduce the variability of ACS estimation.