Non-Cartesian k-space Sampling and Reconstruction in MRI Using Pulseq

Abstract

Utvikling og implementering av nye MR-sekvenser krever ofte innsikt i omfattende, lavnivå og leverandørspesifikke programmeringsmiljøer. Pulseq er et plattformuavhengig verktøy for design av MR-sekvenser som tillater rask implementering og testing av ikke-produktmetoder. Konvensjonell MR-avbildning utføres ved å lese av k-rommet langs en kartesisk eller rettlinjet bane. Motivasjonen for anvendelsen av ikke-kartesiske avlesningsmetoder er muligheten for hurtig traversering av k-rommet i tillegg til økt signal-støy-forhold og robusthet mot bevegelse. Ikke-kartesisk MR-avbildning har imidlertid økt følsomhet for forsinkede og deformerte gradientfelt. Rekonstruksjon basert på nominelle gradientfelt, og derved nominelle k-romsbaner fører ofte til degradering og artefakter i det genererte bildet.

Metoder for å korrigere effekter av ikke-ideelle gradientfelt er implementert i Pulseq for EPI- og spiralavlesninger. For EPI-data hentet fra et 7T-system ble gradientforsinkelser og faseavvik mellom alternerende linjer bestemt og korrigert for. Det faktiske gradientfeltet for spiralavlesning gjennomført på et 3T-system ble bestemt ved fantom-målinger av tynne snitt ut fra isosenter.

Korreksjon av gradientforsinkelser førte til betydelig reduksjon av ghosting artefakter i rekonstruerte EPI-bilder, selv om geometriske forvrengninger fremdeles forekom i fasekodingretningen. Rekonstruksjon av spiralbilder med målte k-rombaner medførte betydelig reduksjon av blurring artefakter, i tillegg til økt SNR og bedret bildeskarphet.

Development and implementation of new MR imaging sequences often requires insight in extensive, low-level and vendor-specific programming environments. Pulseq is a vendor-independent MR sequence prototyping tool, which allows quick implementation and testing of non-product methods. Sampling along non-Cartesian trajectories offers the advantage of efficient k-space traversal and thereby reduced acquisition times and robustness to motion. However, an impediment of the widespread adoption of such imaging sequences lies in their more complicated image reconstruction schemes and their increased sensitivity to non-ideal gradient hardware performance. Delayed and distorted gradient waveforms caused by eddy currents and hardware imperfections lead to trajectories deviating from their prescribed shape and unwanted phase accumulation as readouts grow longer. Identified gradient delays, knowledge of the true gradient field, and thereby the actual k-space trajectory is likely to improve the performance of such fast imaging techniques.

In this work an EPI sequence with active slewing during data sampling was implemented in Pulseq and executed on a 7T imaging system. Under the assumption of linear system phase response, gradient delays and constant phase offsets were estimated from corresponding reference scans acquired with phase-encoding disabled. Spiral-out imaging sequences following an Archimedean k-space trajectory were designed in Pulseq and implemented on a 3T imaging system. Trajectory measurements and B0 eddy current phase errors were estimated from phantom-based phase measurements on thin slices at different off-center positions.

Although geometric distortion in the phase-encode direction is still present in EPI images, substantial ghosting artifacts resulting from system delays are reduced. In object border areas, blurring artifacts were significantly reduced in spiral images reconstructed with measured trajectories, whereas B0 eddy current correction had minor impact on image quality.

Single time-delay correction is far more effective in EPI compared to spiral imaging. Spiral readouts require measurements of the full response of over a wide range of temporal frequencies to properly correct effects of gradient imperfections. The implemented correction methods can enable high-fidelity spiral imaging. Image degradation due to B0 eddy currents is highly dependent on the k-space trajectory and the gradient system, but is expected to be more significant in high-resolution imaging.