We investigated potential benefits of optimizing accelerated 3D myocardial perfusion MRI by time-adaptive data-driven k-t Cartesian sampling during the scan. To this end, the sampling mask for each following acquisition window is inferred by a neural network on the fly using the acquired history of MR signal projections. It is demonstrated that time-adaptive data-driven sampling reduces reconstruction errors by 25% along with a reduction of signal underestimation during the contrast bolus passage when compared to a predefined fixed sampling pattern.
This abstract and the presentation materials are available to members only; a login is required.