A multiplicative model for improving microvascular flow estimation in dynamic contrast-enhanced ultrasound (DCE-US): theory and experimental validation

Abstract

Perfusion parameter estimation from dynamic contrast-enhanced ultrasound (DCE-US) data relies on fitting parametric models of flow to curves describing linear echo power as a function of time. The least squares criterion is generally used to fit these models to data. This criterion is optimal in the sense of maximum likelihood under the assumption of an additive white Gaussian noise. In the current work, it is demonstrated that this assumption is not held for DCEUS. A better-adapted maximum likelihood criterion based on a multiplicative model is proposed. It is tested on simulated bolus perfusion data and on 11 sequences acquired in vivo during bolus perfusion of contrast agent in the cortex of healthy murine kidney, an area where the perfusion is expected to be approximately homogeneous. Results on simulated data show a significant improvement (p < 0.05) of the precision and the accuracy for the estimations of perfusion parameters time to peak (TTP), wash-in rate (WiR), and mean transit time (MTT). On the 11 in vivo sequences, the new method leads to a significant reduction (p < 0.05) in the variation of parametric maps for 9 sequences for TTP and 10 sequences for WiR and MTT. The mean percent decreases of the coefficient of variation are 40%, 25%, and 59% for TTP, WiR, and MTT, respectively. This method should contribute to a more robust and accurate estimation of perfusion parameters and an improved resolution of parametric imaging.

Denoising of Contrast-Enhanced Ultrasound Cine Sequences Based on a Multiplicative Model

“…For lower concentrations, the variability of the measurements increased, indicating deviation from the Rayleigh model. These findings are in agreement with the results presented in [10] and [11]. The K-distribution was proposed as a model that can describe the signal and its variability over a wide range of concentrations.…”

Simulation of dynamic contrast-enhanced ultrasound sequences using example-based texture generation

“…The contrast echo-power is assumed to follow a Gamma distribution. This distribution has been shown in Barrois et al [12] to well describe the dose-ranging data.…”

Evaluation of the Accuracy of Liver Lesion DCEUS Quantification With Respiratory Gating

“…On the other hand the amplitude parameters of AUC and PI were severely affected with the value of the AUC error increasing by a maximum of 51.4% and of the PI by a maximum of 50.5%. Simulations run by Barrois et al [27] found absolute error differences for the AUC of 49.8-50.3% , the MTT of 15.5-22.0% and 4.3-5.6% for the RT. The simulations applied multiplicative noise directly onto each data point of the time intensity curve, corresponding to a mean signal from a 5x5 block of pixels, in contrast with the RMSM that takes the average within a 20px radius ROI.…”

“…Multiplicative noise was applied to the DCEUS side of the simulation in order to compare errors caused by respiratory motion to those produced by noise. The gamma distribution multiplicative noise model proposed by Barrois et al [27] was used according to (4),…”