Referenceless reconstruction of spatiotemporally encoded imaging data: principles and applications to real-time MRI.

Overview

PURPOSE: Ultrafast sequences based on "Hybrid" spatiotemporal encoding (SPEN) replace echo-planar imaging's phase encoding "blips," while retaining a k-space readout acquisition. Hardware imperfections during acquisition may lead to ghosts and striped artifacts along the SPEN dimension; akin to echo-planar imaging's Nyquist ghosts, but weaker. A referenceless method to eliminate these artifacts in Hybrid SPEN is demonstrated. THEORY AND METHODS: Owing to its encoding in direct space, rather than reciprocal space, undersampling in SPEN does not generate an echo-planar-imaging-like aliasing, but instead lowers the spatial resolution. Hybrid SPEN data can be split into two undersampled signals: a reference one comprised of the odd-echos, and an even-echo set that has to be "corrected" for consistency with the former. A simple way of implementing such a correction that enables a joint high-resolution reconstruction is proposed. RESULTS: The referenceless algorithm is demonstrated with various examples, including oblique scans, large in vivo datasets from real-time dynamic contrast-enhanced perfusion experiments, and human brain imaging. CONCLUSIONS: The referenceless correction enables robust single-scan imaging under changing conditions-such as patient motion and changes in shimming over time-without the need of ancillary navigators. This opens new options for real-time MRI and interactive scanning.