A multiplicative model to improve microvascular flow evaluation in the context of dynamic contrast-enhanced ultrasound (DCE-US)

Estimation of perfusion parameters from dynamic contrast-enhanced ultrasound (DCE-US) data relies on locally fitting mathematical models to the time-echo-power curves derived from a sequence. The least-squares method generally used to fit a parametric perfusion model to experimental data is optimal only under the hypothesis of an additive Gaussian noise. Due to the nature of the DCE-US signal, this hypothesis is disputable. A maximum likelihood estimator based on a multiplicative noise model is proposed and tested. Results on simulated data show improvements of the precision and accuracy of commonly estimated perfusion parameters. We also analyzed the perfusion of a rather homogeneous in vivo tissue, the renal cortex of an healthy mouse. The new method leads to more homogeneous parametric maps. These improvements should contribute to a more robust estimation of perfusion parameters and an improved resolution of DCE-US parametric images.