SWTV-ACE: Spatially Weighted Regularization Based Attenuation Coefficient Estimation Method for Hepatic Steatosis Detection

Abstract

We present a spatially weighted total variation regularization based method for measuring the ultrasonic attenuation coefficient estimate (ACE). We propose a new approach to adapt the local regularization by employing envelope signal-to-noise-ratio deviation, an indicator of tissue inhomogeneity. We evaluate our approach with simulations and demonstrate its utility for hepatic steatosis detection. The proposed method significantly outperforms the reference phantom method in terms of accuracy (9% reduction in ACE error) and precision (52% reduction in ACE standard deviation) for the homogeneous phantom. The method also exceeds the performance of uniform TV regularization in inhomogeneous tissue with high backscatter variation. The ACE computed using the proposed method showed a strong correlation of 0.953 (p = 0.003) with the MRI proton density fat fraction, whereas the reference phantom method and uniform TV regularization yield correlations of 0.71 (p = 0.11) and 0.44 (p = 0.38), respectively. The equivalence of SWTV-ACE with MRI proton density fat fraction, which is the current gold standard for hepatic steatosis detection, shows the potential of the proposed method to be a point-of-care tool for hepatic steatosis detection.