Fast functional MRI using inverse imaging with dynamic off-resonance artifacts correction

The phase of NMR signal can drift significantly over time in fMRI experiments due to systematic instability, head motion, or thoracic/pelvic cavity motion. Artifacts related to phase drift can be corrected by retrospective signal processing. However, time series images can be shifted or distorted seriously such that they cannot be recovered by these methods. Navigator echoes have been proposed to correct time-invariant artifacts related to phase drifting before MRI reconstruction. Respiration-induced phase drift can also be estimated and corrected dynamically by measuring navigator echoes in each acquisition in fMRI. In this study, we use the “dynamic off-resonance in k-space” (DORK) method to correct the phase drift in magnetic resonance inverse imaging (InI), which is a method using minimally gradient encoded data and parallel detection to achieve massively accelerated fMRI. As the phase of each accelerated InI scan becomes farther away from the initial value, the discrepancy between the reference scan and the instantaneous accelerated InI acquisition becomes more severe. We hypothesize that DORK can significantly improve the InI reconstructions by reducing such data inconsistency. Empirical results show that DORK can reduce the InI fluctuation in the respiratory frequencies, improve the stability of the fMRI time series, and increase the peak value of hemodynamic response estimates.