Spiral Acquisition for High-Speed Anatomical Imaging at 7T
Lars Kasper1,2, Christoph Barmet1,3, Maria Engel1, Maximilian Haeberlin1, Bertram J Wilm1, Benjamin E Dietrich1, Thomas Schmid1, David O Brunner1, Klaas E Stephan2,4,5, and Klaas P Pruessmann1

1Institute for Biomedical Engineering, University of Zurich and ETH Zurich, Zuerich, Switzerland, 2Translational Neuromodeling Unit, IBT, University of Zurich and ETH Zurich, Zuerich, Switzerland, 3Skope Magnetic Resonance Technologies, Zurich, Switzerland, 4Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom, 5Max Planck Institute for Metabolism Research, Cologne, Germany

Synopsis

We present whole-brain, high-resolution (0.5mm) spiral imaging with proton-density and T2* contrast at 7T in less than a minute. Owing to a comprehensive characterization of the imaging process, artifact-free image reconstruction from long-readout spiral shots (20 ms) becomes feasible via an iterative SENSE algorithm. In particular, trajectory imperfections as well as dynamic off-resonance changes are captured via concurrent field monitoring, while static off-resonance as well as coil sensitivities are mapped in a multi-echo reference scan and augment image reconstruction. The resulting images exhibit the same geometric fidelity as spin-warp images at a fraction of the total acquisition duration.

Introduction

Ultra-high field MRI holds considerable promise for advanced diagnostic imaging due to its unique contrast properties and high baseline sensitivity. This superior sensitivity enables shorter scans and can be traded for higher spatial or temporal resolution. In practice, however, many high-field protocols make grossly inefficient use of the available sensitivity. Cause are high-field specific conditions (SAR limitations, long T1/short T2) that enforce lengthy repetition times, in combination with short readouts. In the ensuing small acquisition duty cycle the superior high-field sensitivity is squandered. To leverage high-field imaging, full use must be made of the available acquisition windows, as can be achieved with efficient and long k-space readouts. In this work, 7T anatomical whole-brain imaging in under 1 minute is shown, presenting a 5-10 fold scan time reduction as compared to conventional readouts. To this end, spiral acquisitions are used, combining efficient k-space readouts with flexible acquisition times. Impeccable image quality is achieved by controlling for physiology- and system-induced encoding field imperfection, combining iterative, off-resonance-map-based image reconstruction with concurrent magnetic field monitoring (Fig. 1, [1]).

Methods

SETUP: The brains of 3 healthy volunteers were scanned on a 7T system (Philips Achieva, Best), using a 1-channel transmit, 32-channel receive coil (Nova Medical, Wilmington). 16 NMR field probes were placed in optimal positions (cf. below) around the receive (and inside the transmit) coil using a clip mount system and connected to a dedicated MR acquisition system [2].

SEQUENCE AND TRAJECTORIES: For the anatomical scans, a set of multi-slice sequences was designed utilizing spiral readouts, which achieved time optimality by fully exploiting the system’s slew-rate and gradient strength limits [3]. The selected sequences covered different contrasts (PD/T2*; spiral in/out) and timing constraints (36-90s total duration, see Fig. 2 for details), but shared slice geometry (thickness 1.5-3mm; -10 deg tilt oblique transverse). A multi-slice 2D setup allowed for whole-brain coverage by slice-interleaved acquisition of up to 50 slices within one TR (3s). For determining the sensitivity and off-resonance maps, a spin-warp multi-echo sequence (delta TE 1ms) was acquired with identical slice geometry, but 1mm in-plane resolution. In all scans the field probes were excited before the start of the readout trajectory and sampled with 1MHz bandwidth during the readout. MR signals from the head-coil were acquired via the scanner spectrometer.

SIGNAL PROCESSING AND IMAGE RECONSTRUCTION (Fig. 1): For every time point during the imaging readout, 16 k-space coefficients (k0-k15) – corresponding to a real-valued spherical harmonic basis function set up to 3rd order – were computed from the probe signals. To minimize error in the measurements of the encoding dynamic fields, the probe positions were chosen such as to reduce the error in the fitted k-space coefficients [4,5] , respecting the coil geometry. On the coil raw data, an iterative, CG-based image reconstruction was performed (SENSE with multi-frequency interpolation, [6,7]). For the encoding model, the measured field dynamics up to first order (k0-k3) were used, in combination with the sensitivity- and static off-resonance maps [8,9], which were smoothed via a CG algorithm.

Results

Field evolutions during spiral encoding of several tens of ms could be successfully monitored (Fig. 3). Combined with static B0 off-resonance maps, virtually artifact-free T2* spiral-out images (Fig.4) could be reconstructed from multiple slices and volunteers, exhibiting both high contrast-to-noise ratio (CNR) and resolution (0.5mm). Furthermore, the geometric congruence of the images with the undistorted multi-echo spin-warp reference scan was confirmed for all proposed spirals (Fig. 5). Total acquisition durations of 90s were sufficient for convincing 0.5mm resolution images. Slight reduction in resolution (0.7mm) enabled sub-minute scan time. Alternatively, a SENSE reconstruction (R=2) with half of the interleaves of a 0.5 mm spiral-out trajectories succeeded, enabling a flexible trade-off between CNR and scan time.

Discussion & Conclusion

Oftentimes poor use is made of the sensitivity available in ultra-high field imaging, resulting in lengthy scan protocols, particularly for high-resolution scans. Spiral readouts with optimal acquisition windows provide a powerful remedy, if the image reconstruction is performed on the actual encoding fields (monitored k-trajectory, static off-resonance and sensitivity maps). Anatomical whole brain coverage is shown in a sub-minute scan with image quality identical to conventional spin-warp acquisitions. Proton-density (TE 3ms) as well as a T2*-weighted contrast is shown, the latter coming in two flavors: outside-in and inside-out spirals.

Spiral imaging is suitable to other ultra-high field applications. As single-shot readout for fMRI, ASL, fQSM, and in multi-shot acquisitions for angiography, susceptibility imaging and different anatomical contrasts. Optimized spiral trajectories such as variable density spirals have the potential of further increasing imaging speed and image SNR, leveraging the superb sensitivity of ultra-high field MRI.

Acknowledgements

No acknowledgement found.

References

[1] B.J. Wilm et al., Magnetic Resonance in Medicine, 2015, in press.

[2] B.E. Dietrich et al., Magnetic Resonance in Medicine, 2015, in press.

[3] M. Lustig, S.-J. Kim, J.M. Pauly, IEEE Transactions on Medical Imaging, 2008, 27, 866.

[4] C. Barmet et al., in Proc. Intl. Soc. Mag. Reson. Med. 18, 2010, 216.

[5] C. Barmet, N. De Zanche, K.P. Pruessmann, Magnetic Resonance in Medicine, 2008, 60, 187?197.

[6] K.P. Pruessmann et al., Magnetic Resonance in Medicine, 2001, 46, 638?651.

[7] L.C. Man, J.M. Pauly, A. Macovski, Magnetic Resonance in Medicine, 1997, 37, 785.

[8] B.P. Sutton, D.C. Noll, J.A. Fessler, IEEE Transactions on Medical Imaging, 2003, 22, 178.

[9] C. Barmet, J. Tsao, K.P. Pruessmann, in Proceedings of the ISMRM, 2005, 682.

Figures

Schematic showing the comprehensive signal model used for image reconstruction. Concurrently measured field dynamics up to first order in space are used in combination with static off-resonance- and sensitivity maps. Algebraic reconstruction on the basis of a conjugate gradient (CG) method is performed on a cluster with 32 CPUs.

Parameters of all performed scans. The multi-echo scan was used to determine off-resonance and sensitivity maps, as well as being the reference for geometrical congruence with the spiral scans (cf. Fig. 5). Inside-out (‘Spiral Out’) and outside-in (‘Spiral In’) spirals were acquired with 0.5mm resolution in 90 seconds. For rapid acquisition below 1 minute, 12-interleaf Spiral Out with 0.7mm resolution was utilized, as well as parallel imaging of 15 interleaves with 0.5mm resolution (SENSE=2).

Encoding field dynamics measured during an 0.5 mm inside-out spiral (ID 2). Three of the total 30 interleaves are shown in parametric view (bottom right). The temporal dynamics of one interleaf is shown (left and top right), each subplot grouping all phase coefficients pertaining to one spatial order. ‘Max rad’ indicates the maximum phase excursion induced in the imaging volume (20cm sphere).

T2*-weighted images from a 0.5mm Spiral Out acquisition (ID 3). 30 interleaves, acquired in 90 seconds, provide rich anatomical information. Displayed are 5 slices each for 2 subjects, with sequence timing allowing for whole brain coverage (50 slices) within TR (3s). While thicker slices (3mm, top row) provide superior contrast-to-noise, fine-grained details, e.g. blood vessels, are discernible in thinner slices (1.5 mm, bottom row).

Overview of anatomical utility for all proposed spirals in the same slice (1.5 mm, subject 3). Numbers indicate the sequence ID (cf. Fig. 2). Note the icons: only the top left shows a spin-warp image (ID 1, 1st echo). All 0.5mm spirals (ID 2,3,4) provide high geometric congruency to the spin-warp, with higher resolution/contrast, at half the imaging time. Slightly reduced resolution (0.7mm, ID 5) is possible in 36s. Parallel imaging (SENSE=2) accelerates the 0.5mm spirals to 45s (ID3=>ID6).



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0741