Effective Depth Expansion For Reliable Fatty Liver Assessment Using A Double Nakagami Distribution Model

Quantitative ultrasound (QUS) methods have been widely used for soft tissue characterization. Spatial resolution (i.e., ultrasound frequency) is an important factor for QUS methods. In our previous study, we proposed double Nakagami distribution (DND) model for the analysis of fatty liver and high frequency ultrasound (HFU) which allows finer-resolution QUS. Healthy liver structure filter (HLSF) classified each ROI based on the DND model parameter distribution which acquired from healthy liver samples. This approach was able to successfully diagnose fatty livers (>20 % steatosis percentage) in a dataset of 12 livers ranging from 0 to 90 % steatosis. This study proposed a compensation method to expand effective depth range of HLSF based on DND model using HFU measurement. Radio-frequency data was experimentally acquired from 12 excised rat livers (three healthy (0 % of hepatocytes contain lipid droplets) and nine fatty (10 to 70 %)). Healthy liver structure filter (HLSF) classified each ROI based on the DND model parameter distribution which acquired from healthy liver samples. The functions of the depth-dependent Nakagami parameters were obtained by fitting the modified Gaussian distribution to the Nakagami parameters obtained from the three normal liver samples. HLSF(x) was constructed using healthy liver datasets from focal depth - 0.5 mm to focal deplth + 3.5 mm in 1 mm interval. The filter applied to estimated DND parameters at the same depth. For comparison, the conventional method used a fixed value of the Nakagami parameter for DND model parameter estimation and HLSF constructed at focal depth. Depth dependent of the Nakagami parameter and HLSF decreased the depth dependency of DND model parameter. AUROC classifying over than 15 % steatosis progress improved the performance at a distance from focal depth of +3.5 mm (0.64 to 0.86). The proposed method expanded reliable QUS (area under the receiver operating characteristic > 0.85) depth range by 250 % against half of depth of field and demonstrate QUS can be used reliably with clinical HFU.