Mapping myocardial perfusion with an intravascular MR contrast agent, a robust estimation by a spatially constrained approach

Evaluation of quantitative parameters such as regional myocardial blood flow (rMBF), blood volume (rMBV), and mean transit time (rMTT) by MRI is gaining acceptance for clinical applications, but still lacks robust post-processing methods for map generation. Pixel by pixel analysis leads to high variance of the estimates and variance reduction by posterior spatial averaging does not produce satisfactory results. We propose a parametric estimation technique that smoothes the variance of the estimates within parametric homogeneous regions while preserving discontinuities (in the sense that smoothing is not performed across regions with different parametric contents). Our approach lies within the Bayesian framework. It is based on local autoregressive moving average (ARMA) estimation, constrained by an edge-preserving smoothness prior. The prior stems from Markov random fields (MRF) modeling and involves non-quadratic potential functions. The output of the resulting algorithm is a regularized rMBF map. The method is validated on synthetic MR kinetics and tested on first-pass T1 images of an isolated pig heart using an intravascular contrast agent. Comparison of our results with pixel by pixel estimates clearly demonstrate the ability of the proposed approach to improve parametric estimation in terms of variance reduction and discontinuity preservation