Synopsis

Pushing the spatiotemporal resolution using spiral acquisition has been an active area of research in dynamic speech imaging. However application of long readout spirals at high field have been limited by off-resonance blurring around the mouth and airway. Low field MRI offers an advantage to speech imaging, due to reduced local susceptibility gradients. Here, we demonstrate reconstructions of single shot spiral (up to 9.4 ms readouts) without accruing any blurring artifacts resulting in high temporal resolution (11 ms) dynamic speech imaging.

Introduction

Dynamic MRI provides a viable non-invasive method for studying the upper airway and visualization of morphological and functional aspects of speech. High spatial (1-4 mm²) [1] and temporal (< 40 ms) [2] resolution demands are typically achieved by highly undersampled acquisitions with compressed sensing image reconstructions. Spiral acquisitions are a popular choice, for high spatiotemporal resolution. Often spiral acquisitions suffer from blurring artifacts caused by local susceptibility gradients at air-tissue boundaries in speech imaging and, as a result, acquisitions are limited to 2.5ms in length at 1.5T [3] or require off-resonance correction [4].

Here, we demonstrate the prospect of speech imaging using a high performance prototype 0.55T MRI system. Low field offers the advantage of reduced local susceptibility gradients, enabling longer readouts without blurring artifacts. We show the ability to image with a single shot spiral of duration 9.4 ms at 0.55 T without the need for off-resonance correction.

Methods

A custom 0.55T MRI system (modified MAGNETOM Aera, Siemens Healthcare, Erlangen, Germany) was used for imaging. Gradient performance was maintained (45 mT/m maximum amplitude and 200T/m/s slew rate) to allow real-time imaging.

Human subjects imaging was approved by the local Institutional Review Board. Data was acquired in healthy volunteers with a 16 channel head/neck coil while they repeated a simple speech task of counting numbers one through five. A pseudo golden angle spiral GRE sequence was used (Imaging parameters: thickness = 10 mm, TE = 0.82 ms, FOV = 280 mm, flip angle = 20°, spatial resolution = 2.2X2.2 mm² ). Datasets with spiral readout duration 2.4, 7, 9.4 and 14 ms (TR = 4.32, 9, 11.3 and 16 ms) were collected and reconstructed with one spiral arm per frame (resulting in temporal resolution = 1 TR). Shimming volume was localized to the mouth and no additional center frequency adjustment was performed. A dataset with same imaging parameters and spiral readout duration of 9.3 ms (TR = 11.2 ms) was collected on a different volunteer at 1.5 T, for comparison.

We corrected the spiral trajectories using GIRF [5] prior to reconstruction. All the reconstructions were performed off-line using a sparse-SENSE method constraining temporal finite difference similar to the approach described in [6]. Sparsity penalty on temporal finite difference penalizes rapidly varying pixel-time profiles which has been used previously for dynamic imaging [7,8]. Single shot spiral reconstructions were performed for datasets with spiral readout lengths 7, 9.4 and 14 ms (R = 8,6,4) collected at 0.55T and also for the 1.5T dataset. A fully sampled NuFFT reconstruction was used to generate reference data from the short readout spiral (2.4 ms) resulting in a low temporal resolution of 146 ms/frame.

Results

A comparison of readout lengths in a volunteer illustrates the unique ability to extend readout lengths up to 9.4 ms (Figure 1) without observable image blurring around airways, velum, tongue or lips. On extending the readout length to 14 ms the blurring around airways becomes apparent. As expected , at 1.5T the 9.4 ms readout results in extreme blurring and the spatial details around the velum are lost.

The reconstruction of a single shot enables temporal resolution of 11ms. The 9.4 ms readout was preferred over 7 ms due to the increase in SNR with sampling time. The animation of the reconstructed dynamic series is shown in Fig 2. Representative time frames from the speech task are shown in Figure 3 revealing the dynamics during a speech task. The visual quality of dynamic frames do not suffer from any degradation and the spatiotemporal profiles shown in Figure 4, through the main articulators (velum, tongue and lips) exhibit clear delineation of the movement of the soft palate and demarcation of the tongue from the soft palate and lips.

Discussion and Conclusion

Acknowledgements

References

1. Lingala, Sajan Goud, Brad P. Sutton, Marc E. Miquel, and Krishna S. Nayak. "Recommendations for real‐time speech MRI." Journal of Magnetic Resonance Imaging 43, no. 1 (2016): 28-44.

2. Beer AJ, Hellerhoff P, Zimmermann A, Mady K, Sader R, Rummeny EJ, Hannig C. Dynamic near-real-time magnetic resonance imaging for analyzing the velopharyngeal closure in comparison with videofluoroscopy. J Magn Reson Imaging 2004;20:791–797.

3. Lim, Yongwan, Sajan Goud Lingala, Shrikanth S. Narayanan, and Krishna S. Nayak. "Dynamic off‐resonance correction for spiral real‐time MRI of speech." Magnetic resonance in medicine (2018).

4. Niebergall, Aaron, Shuo Zhang, Esther Kunay, Götz Keydana, Michael Job, Martin Uecker, and Jens Frahm. "Real‐time MRI of speaking at a resolution of 33 ms: Undersampled radial FLASH with nonlinear inverse reconstruction." Magnetic Resonance in Medicine 69, no. 2 (2013): 477-485.

5. Campbell‐Washburn, A. E., Xue, H., Lederman, R. J., Faranesh, A. Z., & Hansen, M. S. (2016). Real‐time distortion correction of spiral and echo planar images using the gradient system impulse response function. Magnetic resonance in medicine, 75(6), 2278-2285.

6. Lingala, Sajan Goud, Yinghua Zhu, Yoon‐Chul Kim, Asterios Toutios, Shrikanth Narayanan, and Krishna S. Nayak. "A fast and flexible MRI system for the study of dynamic vocal tract shaping." Magnetic resonance in medicine 77, no. 1 (2017): 112-125.

7. Feng, Li, Robert Grimm, Kai Tobias Block, Hersh Chandarana, Sungheon Kim, Jian Xu, Leon Axel, Daniel K. Sodickson, and Ricardo Otazo. "Golden‐angle radial sparse parallel MRI: combination of compressed sensing, parallel imaging, and golden‐angle radial sampling for fast and flexible dynamic volumetric MRI." Magnetic resonance in medicine 72, no. 3 (2014): 707-717.

8. Adluru, Ganesh, Chris McGann, Peter Speier, Eugene G. Kholmovski, Akram Shaaban, and Edward VR DiBella. "Acquisition and reconstruction of undersampled radial data for myocardial perfusion magnetic resonance imaging." Journal of Magnetic Resonance Imaging: An Official Journal of the International Society for Magnetic Resonance in Medicine29, no. 2 (2009): 466-473.