Cartesian MaxGIRF: Model-based EPI reconstruction incorporating gradient nonlinearity and concomitant field effects

Lower field strength scanners with large bore size or complex geometries, and scanners with stronger gradient systems experience increased gradient nonlinearity and concomitant fields, each of which causes distortions in EPI. Current correction approaches based on image-domain interpolation introduce undesirable spatial blurring. To avoid spatial blurring, we introduce a model-based EPI reconstruction framework, denoted Cartesian MaxGIRF ("Max"well field correction using "GIRF"-predicted gradients), that simultaneously compensates the effects of concomitant fields, gradient nonlinearity, and off-resonance during image reconstruction.

Performance of the proposed framework was compared against standard correction methods using phantom datasets at 0.55T: (1) 2D spin-echo EPI (SE-EPI) with reversed phase-encoding directions and (2) accelerated 2D SE-EPI with partial Fourier. Two unique EPI image artifacts induced by concomitant fields ("parabolic shift" and "slice-dependent Nyquist ghost") were demonstrated and mitigated by the proposed framework using long-ETL 3D GRE-EPI and high-resolution 3D GRE-EPI, respectively. Resolution improvements and artifact mitigations by the proposed framework were demonstrated using in-vivo human brain datasets: (1) accelerated 2D diffusion-weighted SE-EPI and (2) high-resolution 3D GRE-EPI at 0.55T.

The amount of the parabolic shift for each imaging case was theoretically analyzed. The proposed framework demonstrated the mitigation of both parabolic shifts and slice-dependent Nyquist ghosts and retained better image details than standard correction methods when mitigating geometric distortions for all scenarios.

The Cartesian MaxGIRF framework simultaneously mitigates the effects of concomitant fields, gradient nonlinearity, and static off-resonance. This approach is particularly useful to mitigate artifacts induced by second-order concomitant fields present in both symmetric and asymmetric gradient systems.