Comparison of quantitative ultrasound parameters for fat content liver detection and monitoring

Quantitative ultrasound parameter estimates from the Lizzi-Feleppa (LF slope, intercept and midband) and quantitative ultrasound (mean BSC, mean AC, ESD and EAC) approaches were compared for in vivo and ex vivo mouse liver fat content monitoring. The quantitative ultrasound parameters having the higher correlations with fat content were EAC (R
²
=0.68), mean BSC (R
²
=0.60), mean AC (R
²
=0.60) and LF midband (R
²
=0.70) for ex vivo conditions and mean BSC (R
²
=0.45), LF midband (R
²
=0.44) and mean AC (R
²
=0.37) for in vivo conditions. By defining a threshold fat content of 11.6% between normal vs. fatty mouse livers, the parameters yielding the better sensitivities and specificities to differentiate normal and fatty liver ex vivo were the mean AC (sensitivity=1, specificity=1), then mean BSC and LF midband (sensitivity=0.87, specificity=1). In vivo, the parameters yielding the better sensitivities and specificities to differentiate normal vs. fatty liver were mean BSC and LF midband (sensitivity=0.93, specificity=0.87) and then the mean AC (sensitivity=0.67, specificity=0.93).