Premium
Image reconstruction using a gradient impulse response model for trajectory prediction
Author(s) -
Vannesjo S. Johanna,
Graedel Nadine N.,
Kasper Lars,
Gross Simon,
Busch Julia,
Haeberlin Maximilian,
Barmet Christoph,
Pruessmann Klaas P.
Publication year - 2016
Publication title -
magnetic resonance in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.696
H-Index - 225
eISSN - 1522-2594
pISSN - 0740-3194
DOI - 10.1002/mrm.25841
Subject(s) - ghosting , image quality , imaging phantom , impulse response , computer science , artificial intelligence , trajectory , iterative reconstruction , computer vision , diffusion mri , impulse (physics) , mathematics , physics , magnetic resonance imaging , image (mathematics) , optics , medicine , mathematical analysis , astronomy , radiology , quantum mechanics
Purpose Gradient imperfections remain a challenge in MRI, especially for sequences relying on long imaging readouts. This work aims to explore image reconstruction based on k‐space trajectories predicted by an impulse response model of the gradient system. Theory and Methods Gradient characterization was performed twice with 3 years interval on a commercial 3 Tesla (T) system. The measured gradient impulse response functions were used to predict actual k‐space trajectories for single‐shot echo‐planar imaging (EPI), spiral and variable‐speed EPI sequences. Image reconstruction based on the predicted trajectories was performed for phantom and in vivo data. Resulting images were compared with reconstructions based on concurrent field monitoring, separate trajectory measurements, and nominal trajectories. Results Image reconstruction using model‐based trajectories yielded high‐quality images, comparable to using separate trajectory measurements. Compared with using nominal trajectories, it strongly reduced ghosting, blurring, and geometric distortion. Equivalent image quality was obtained with the recent characterization and that performed 3 years prior. Conclusion Model‐based trajectory prediction enables high‐quality image reconstruction for technically challenging sequences such as single‐shot EPI and spiral imaging. It thus holds great promise for fast structural imaging and advanced neuroimaging techniques, including functional MRI, diffusion tensor imaging, and arterial spin labeling. The method can be based on a one‐time system characterization as demonstrated by successful use of 3‐year‐old calibration data. Magn Reson Med 76:45–58, 2016. © 2015 Wiley Periodicals, Inc.