Messing up to clean up: Semi-randomized frequency selective space-filling curves to suppress physiological signal fluctuations in MRI
Abstract
Purpose
Rapid 3D steady-state sequences are widely used but are also known to be sensitive to semi-periodic physiological signal fluctuations due to, for example, cardiac pulsation, breathing, and eye/eyelids movement. This semi-periodicity results in repeating artifacts in the image whose intensity depends on the scan parameters. The purpose of this study is to design a reordering of the 2D phase encodes (within the 3D acquisition) that reduces these artifacts.
Methods
A randomized order of the phase encodes can suppress repeating artifact but may also introduce its own apparent noise, for example, in cases of slow subject movement or gradual changes in eddy currents. In a new design a semi-randomized space-filling curve is generated by scrambling the local order of the phase encodes to achieve a controlled frequency selective effect, that is, eliminating artifacts above a chosen (fluctuation) frequency threshold while leaving lower frequencies untouched, thus overcoming the limitations of a randomized order. The method was characterized in simulations and substantiated by human brain imaging at 7 T using two steady-state gradient echo variants that suffer from pulsation, either near blood vessels or near the ventricles.
Results
The simulations with a point source show that the maximum artifact intensity can be reduced by factors of 10–50 depending on the scan parameters. In human scanning, the new approach drastically reduced physiologically induced artifacts and was superior in this regard to both full randomization and a generalized Hilbert curve, another semi-randomized approach.
Conclusion
The phase-encodes reordering presented here effectively removes artifacts arising from local fluctuations.