Direct parametric reconstruction from undersampled (k, t)-space data in dynamic contrast enhancement MRI

Dynamic contrast enhancement (DCE) magnetic resonance imaging (MRI) requires high temporal resolution to capture enhancement processes that occur rapidly after the injection of contrast agent. Undersampling can accelerate the acquisition but will result in image degradation. Typically the DCE images are reconstructed individually and kinetic parameters are estimated by fitting a pharmacokinetic model to the time-enhancement response; this method is denoted as “indirect”. Techniques such as ktFOCUSS can be employed in the reconstruction step to avoid aliasing artifacts due to undersampling. This paper suggests a Bayesian inference framework to estimate kinetic parameters (related to the extended Toft model) directly from undersampled (k, t)-space DCE MRI. The proposed scheme is evaluated on a simulated abdominal DCE phantom, for fully sampled, 4 and 8-fold undersampled (k, t)-space data. The estimated kinetic parameters with the proposed algorithm improves correspondence (measured with mean absolute percentage difference) with the ground truth kinetic parameters up to 7% for fully sampled data and up to 19% for 4-fold and 8-fold undersampling, compared to the indirect approach.