Impact of Recirculation in Dynamic Contrast-Enhanced Ultrasound: A Simulation Study

Objectives: The impact of recirculation on the quantification of perfusion is often neglected. It can however introduce a bias or some variability in the estimation of perfusion parameters and thus hamper comparison between exams. Methods: Time-intensity curves (TICs) were simulated using a one-compartment model fed by an arterial input function (AIF). A simple model was developed to simulate recirculation in the AIR. Using AIF with and without recirculation, and sets of regional perfusion parameters, TICs corresponding to different tissue regions were simulated by convolution of the AIFs with the transfer function associated to each region. Ten noise levels and 150 simulations for each noise level were then computed. For each simulated study, six quantification methods based on either Log-Normal modeling or a compartmental modeling using a reference tissue were tested. Variations of the conventional Log-Normal model were also investigated, using parameters estimated in the reference tissue for normalization purposes, and fitting only the first phase of the TIC to avoid recirculation. Results: The impact of recirculation varies according to the quantification method. Restricting parameter estimation to the first samples of the TICs, before recirculation occurs, appears to be the worst strategy. Errors are largely minimized when using a reference tissue to establish relative parameters. The most robust approach is the compartmental modeling based on a reference tissue and applied to multiple regions with a regularization constraint. Conclusion: This simulation study demonstrates the influence of recirculation on the estimation of perfusion parameters. To reduce the impact of this unavoidable effect, the quantification method based on compartmental modeling and using a reference tissue appears to be the most reliable strategy.