With k-t-sub-Nyquist sampling, dynamic magnetic resonance imaging (MRI) measurements become more efficient, since time-consuming gradient encoding steps are omitted. Efficiency is yet only guaranteed, if the imaging information can be reconstructed without compromising the spatio-temporal resolution by image artifacts, noise or spatial blurring. In this work, the topic of k-t-sub-Nyquist sampling with time-resolved parallel imaging reconstruction is investigated from two ends: from the theoretical and the practical end.
A signal and noise transfer analysis for time-resolved parallel imaging methods is derived as one key element towards a unified theoretical framework of these methods. Two prominent methodological approaches of parallel imaging in the case of static MRI are connected by analogous expressions and a generalization of the known geometry (g)-factor for k-t-kernel based time-resolved parallel imaging methods is provided. This allows to evaluate static and dynamic parallel imaging methods, as well as solely temporally based kernel approaches, within a common framework.
Furthermore, time-resolved parallel imaging is explored in the practical context of k-t-sub-Nyquist sampled echo planar imaging (EPI), where signal is sensitive to disturbances such as the off-resonance phenomena. The development of a k-t-sub-Nyquist sampled EPI sequence (k-t-EPI) and image reconstruction scheme is described. The issues and implications balancing signal encoding and image reconstruction in these scenarios are discussed. The developed dynamic MR acquisition schemes are applied in dynamic susceptibility contrast (DSC) weighted cerebral perfusion imaging of tumor patients. The proposed k-t-EPI acquisition facilitates a higher spatial resolution compared to clinical routine measurements. This is achieved without sacrificing temporal resolution and while mitigating the EPI inherent in-plane susceptibility and geometric distortion artifacts.