3844

Fast Irregular MRSI spiral acquisition for sparse spectra. Application to 31P MRSI in muscles.
Jabrane Karkouri1,2, Fabien Millioz3, Magalie Viallon3, Rémy Prost3, and Hélène Ratiney3

1Univ Lyon, UJM-Saint-Etienne, INSA, CNRS UMR 5520, INSERM U1206, CREATIS, F-42023, SAINT-ETIENNE, France, Lyon, France, 2Siemens Healthineers, Saint-Denis, France, 3Université de Lyon, INSA-Lyon, Université Claude Bernard Lyon 1, UJM-Saint Etienne, CNRS, Inserm, CREATIS UMR 5220, U1206, Lyon, France

Synopsis

Magnetic resonance spectroscopic imaging (MRSI) has multiple interests in clinical practice but it faces quite long acquisition time in practice which limits its use in a clinical environment. In this work, a new fast Magnetic Resonance Spectroscopic image acquisition method, inspired by Compressed Sensing using the a priori known sparse support of the metabolites chemical shift, is introduced and evaluated based on a k-t space spiral sampling. In the proposed method, the spatial and temporal interleaves are both taken in consideration during the implementation, in order to reach an even faster acquisition. This method has been evaluated using real in vivo 31P data.

INTRODUCTION

MRSI is a way to evaluate the concentration of metabolites of interest in the body both in the spectral and spatial dimension. In [5], we introduced a new fast acquisition method using a spiral k space sampling, and based on an a priori known spectrum support to select a times samples set to acquire. The FID time samples to acquire are selected thanks to the Sequential Backward Selection (SBS) algorithm [3]. In this work, we aim to extend this method. Spatial and spectral interleaves of the multiple spirals in the k-t space are taken in consideration. The method has been evaluated using real in-vivo 31P data, to prove the feasibility of the proposed approach.

METHODS

Spiral k-space sampling uses successive time varying gradients. To sufficiently cover the kx-ky space, it is often necessary to do spatial interleaves which depend on the length of a spiral, the desired spatial resolution and the MRI machine system gradient commutation and amplitude limitations.

It is also necessary to do temporal interleaves in MRSI, given the length of a spiral and the desired FID signal temporal resolution, in order to acquire all the required temporal points.

Our method is characterized by:

1) instead of sampling the full k-space at each time point for Nspat spatial interleaves, Nspat k-subspaces are sampled with spirals

2) instead of sampling the full t-space, we subsample it thanks to the SBS algorithm calculated for an a priori known spectrum support

3) different sets of time samples are selected with different initializations of SBS for each spatial interleave.

Each set of time samples will then correspond to a k-t subspace. Each time sample is now acquired with a spiral covering a kx-ky subspace where each set of time samples will have a spiral design rotated from one to another, corresponding to the spatial interleaves.

After each Repetition Time (TR), a train of spirals, of different k-subspaces, is launched in order to acquire the predefined number of points, including all the sets of time samples. (Fig. 1)

Once each k-t subspace is acquired, a LS reconstruction (Fig. 2) is performed to reconstruct the spectral support for each k-t subspace, which leads as many k-f subspaces. Then, we construct the full k-f space formed with the union of all the k-f subspaces generated. Finally, a regridding algorithm is used to interpolate the k-f space into a Cartesian grid and then reconstruct the spatial dimension.

This method has been evaluated on 31P MR spectroscopy data acquired on a Siemens Prisma 3T MRI machine. The spiral acquisition has been simulated from a cartesian acquisition, matrix size of 32x32, NA=2 [4]. Three methods have been compared: A) the conventional spiral sampling method in MRSI, B) SBS based irregular sampling for spiral trajectories [5,6] and C) the new proposal: SBS based irregular sampling coupled with spatial and temporal interleaves multiplexing. A reduction factor of 4 has been used for the time samples selection, 4 spatial interleaves and 4 temporal interleaves. Spirals length time are 1ms, in order to reach a temporal resolution of 0,25ms and a matrix size of 32x32 for a 200x200mm2 of FOV. The acquisition window is of 256ms, and the TR is of 2s.

RESULTS

The anatomic image and the in-vivo 31P acquired spectrum are represented in Fig. 3. Fig. 4 shows the in-vivo 31P image data, acquired with a conventional phase-phase encoding method, the used phantom for our simulation based on the in vivo image, and the reconstructed image with our algorithm. Fig. 5 shows the magnitude of the reconstructed 31P spectrum for method B and C for a voxel in the left ellipse. In a noise free acquisition the Signal to Reconstructed Error Ratio (SRER) were respectively 25.8 dB for method A, 24.9 dB for method B and 21.5 dB for method C for a pixel in the left ellipse. For a noisy acquisition (sigma=10%PCr), the results are: 23.6dB (A) 19.8dB (B) and 17.5 dB (C). For A), the minimum acquisition time is Tacqu = 32s. The gain in terms of acquisition is, compared to A), of 2 for B) and 2.3 for C).

CONCLUSION

A new MRSI fast acquisition strategy is proposed by exploiting the irregular sampling of the spectroscopic dimension to lower the acquisition time using both spatial and spectral interleaves. We showed that we were able with this new method to reduce the scan time which represents an additional contribution in the context of fast MRSI acquisition and will be studied in the future in dynamic 31P/13C MRSI acquisition [7,8].

Acknowledgements

This work has been supported by Siemens Healthineers and the LABEX PRIMES (ANR-11-LABX-0063), program ”Investissements d’Avenir” (ANR-11-IDEX-0007).

References

1. Posse, S., et al., MR spectroscopic imaging: principles and recent advances. Journal of magnetic resonance imaging : JMRI, 2013. 37(6): p.1301-25.

2. Adalsteinsson, E., et al., Volumetric spectroscopic imaging with spiral-based k-space trajectories. Magnetic Resonance in Medicine, 1998.39(6): p. 889-898.

3. Reeves, S.J. and L.P. Heck. Selection of observations in signal reconstruction. IEEE Trans. Signal Proc. 1995; 43: 788-791

4. J. Fessler, Image Reconstruction Toolbox, https://web.eecs.umich.edu/~fessler/code

5. Karkouri, J. et al., Irregular spiral acquisition for compressive sensing in MRSI, ISMRM 2017

6, Karkouri, J et al., Time samples selection in spiral acquisition for sparse magnetic resonance spectroscopic imaging, IEEE ICIP 2017

7. Valkovic, Ladislav, CHMELÍK, Marek, MEYERSPEER, Martin, et al. Dynamic 31P–MRSI using spiral spectroscopic imaging can map mitochondrial capacity in muscles of the human calf during plantar flexion exercise at 7 T. NMR in Biomedicine, 2016, vol. 29, no 12, p. 1825-1834.

8. Wiesinger, F., Weidl, E., Menzel, M. I., Janich, M. A., Khegai, O., Glaser, S. J., ... & Schulte, R. F. (2012). IDEAL spiral CSI for dynamic metabolic MR imaging of hyperpolarized [1‐13C] pyruvate. Magnetic resonance in medicine, 68(1), 8-16.


Figures

Fig. 1: Sparse reconstruction using LS estimation and noise minimization using SBS. Illustration of the algorithm used to reconstruct the under-sampled FID signal acquired. The SBS algorithm allows an iregular sampling of the FID signal y. The vector ym is the result of an irregular sampling of the vector y.

Fig. 2: a) Time samples required to acquire and selected by first SBS in blue, 2nd in red, third in yellow, fourth in purple. Time samples required to be acquired after the first excitation in b), second excitation in c), last excitation (7th here) in d). We see that after an excitation we can acquire any spirals by the different 4 given sets of time samples selected by SBS. Illustration of spiral time sampling for 1st, 2nd and last excitation. All the points of the spirals are assumed to be acquired instantaneously for this illustration. Only 100 temporal points over 1024 are shown here.

Fig. 3: In vivo anatomic and 31P spectrum (from the highlighted voxel) acquired on the quadriceps of a man at 3T. The three metabolites Phosphocreatine (PCr), Adenosine Triphosphate(ATP) and Phosphate Inorganic (PI) are indicated and show to constitute a sparse spectrum

Fig.4: For t=0, a) Acquired in vivo 31P image at 3T b) Phantom (based on in vivo image) ; c) Reconstructed image with a gridding algorithm

Fig. 5: Reconstructed spectrum (sigma noise 10% PCr) from a voxel in the left ellipse a) for the method A), b) for the method B), and for the new proposed approach C). The reconstruction error (shifted by -100) is given on the bottom.

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)
3844