Advances in Prospective Motion Correction with Gradient Tones
Maximilian Haeberlin1, Alexander Aranovitch1, Bertram Wilm1, David Otto Brunner1, Benjamin Dietrich1, Barmet Christoph2, and Klaas Paul Pruessmann1

1Institute for Biomedical Engineering, University of Zurich and ETH Zurich, Zurich, Switzerland, 2Skope Magnetic Resonance Technologies, Zurich, Switzerland

Synopsis

A system for prospective motion correction using field probes and gradient tones is presented that is independent of sequence parameters and thus compatible with clinically relevant scans. An examples of a successfully corrected MPRAGE sequence is shown and the bandwidth and the amount of unintentional head motion is measured during a 32 min. scan.

Introduction

Rigid body motion is a major contributor to image quality degradation in both clinical and research applications of MRI. External marker-based prospective motion correction (PMC) methods address that problem by adapting the geometry of the MR sequence to the moving head by tracking head-mounted markers in real-time1–7. The gradient tones (“tones”) approach stands out as the only one that requires no line-of-sight access to the head and performs marker (“field probe”) tracking without additional scan time and robustness against low-frequency field fluctuations. The implementation of the tones method6, however, contained shortcomings that precluded its application to a clinical setting. The tones were placed into the image encoding part of the sequence, which required both informing the image reconstruction about the gradient system’s realization of the tones as well as applying a bandstop filter to the sequence gradients for sequence-orthogonal probe tracking. That coupled the field-probe and sequence design: The probe’s T2 relaxation time was matched to the duration of the imaging readout, its sample diameter was limited by the imaging resolution. The PMC digital signal processing (DSP) chain was implemented entirely on the MR system, which required available RF receive hardware for 19F signal reception on the spectrometer as well as computing power on the sequence controller. This lower-bounded the sequence TR to several tens of milliseconds, thus precluding scans with short TR. The current work addresses the limitations of the gradient tones implementation for improved compatibility with clinically important scans. This is achieved by employing field probes with higher sensitivity to allow very short tone durations that can be incorporated into sequence periods other than the image encoding without the need to bandstop filter them. PMC is realized with a new system that performs the probe signal acquisition and the DSP on an external computer for independent motion tracking and sequence operation. The approach is validated in vivo with a 3D-T1w-TFE (MPRAGE) sequence.

Methods

All experiments were conducted on a Philips 3T Achieva System with a volunteer equipped with a novel headset 3D-printed with ABS (Fig. 1). The headset was designed to provide stability and motion-freedom and minimal skin contact. The probes were connected via custom-built hardware to a custom-configured system that controlled the probes’ transmit/receive operation8. The system was programmed in LabVIEW to perform DSP operations for PMC including phase-unwrapping, probe position determination6, and the computation of rigid body parameters9. It transmitted the geometry updates used for PMC to the host using the manufacturer’s external data interface10. Bandwidth estimation: Three healthy volunteers were tracked using tones to estimate the bandwidth of unintended motion in order to estimate the required tracking frequency of the external PMC system. The tracking sequence contained three tones (Frequencies = 1.5 (x), 2.5 (y), 3.5 (z) kHz, Amplitudes=8.5mT/m, duration=2ms) that were repeated at 100Hz for 32min. In vivo validation: The external PMC system was used to perform PMC in a 3D T1w TFE sequence (1mm3 resolution, 240x240x180 mm FOV, TFE shot duration 1.9s, IR time 1s, flip angle 8°) augmented with gradient tones (frequencies 4.3 (x), 5.7 (y), 7.1 (z) kHz, duration=0.7ms, amplitudes = 6.6 (x), 5.5 (y), 4.4 (z) mT/m) placed on the plateau of each crusher gradient and in the inversion recovery period (Fig. 2). A healthy volunteer was instructed to remain still (Experiment 1), and to perform a head shake motion (Experiment 2) and each experiment was repeated with and without PMC. PMC was performed with field a probe design optimized for motion tracking: A very short T2 (approx. 3ms) and a thick sample diameter (1.6mm) allowed for sensitive tracking at high repetition rates. The probes’ TR during the TFE train was chosen independently of the sequence TR (8ms) and to safely fulfill the expected Nyquist rate from the preceding bandwidth estimation, and amounted to 16 ms.

Results

The results of the bandwidth experiments are plotted in Figs. 3a/b. Significant motion was present up to 12Hz. The corresponding time-domain representation of the rigid body parameter fits are shown in Figs. 3c/d/e, and exhibit drifts of up to 2° and 3mm, respectively. Image reconstructions of the in vivo experiments are shown in Fig.4 (Experiment 1), and Fig.5 (Experiment 2).

Discussion

The present work demonstrates the application of PMC using gradient tones to an MPRAGE sequence by means of an external system that allows the independent operation of probe tracking and sequence tasks. The empirically determined tracking rate of 24 Hz will be a very useful guideline for future head tracking system designs. The improved probe sensitivity enables gradient tones of sub-millisecond duration, which are easily placed in available sequence windows

Acknowledgements

Giel Mens from Philips Medical System is gratefully acknowledged for help with setting up the communication link to the MR scanner.

References

1. Zaitsev, M., Dold, C., Sakas, G., Hennig, J. & Speck, O. Magnetic resonance imaging of freely moving objects: prospective real-time motion correction using an external optical motion tracking system. NeuroImage 31, 1038–1050 (2006).

2. Aksoy, M. et al. Real-time Optical Motion Correction for Diffusion Tensor Imaging. Magn. Reson. Med. Off. J. Soc. Magn. Reson. Med. Soc. Magn. Reson. Med. 66, 366–378 (2011).

3. Forman, C., Aksoy, M., Hornegger, J. & Bammer, R. Self-encoded marker for optical prospective head motion correction in MRI. Med. Image Anal. 15, 708–719 (2011).

4. Schulz, J. et al. An embedded optical tracking system for motion-corrected magnetic resonance imaging at 7T. Magn. Reson. Mater. Phys. Biol. Med. 25, 443–453 (2012).

5. Ooi, M. B., Krueger, S., Thomas, W. J., Swaminathan, S. V. & Brown, T. R. Prospective real-time correction for arbitrary head motion using active markers. Magn. Reson. Med. 62, 943–954 (2009). 6. Haeberlin, M. et al. Real-time motion correction using gradient tones and head-mounted NMR field probes. Magn. Reson. Med. n/a–n/a (2014). doi:10.1002/mrm.25432

7. Andrews-Shigaki, B. C., Armstrong, B. S. R., Zaitsev, M. & Ernst, T. Prospective motion correction for magnetic resonance spectroscopy using single camera retro-grate reflector optical tracking. J. Magn. Reson. Imaging 33, 498–504 (2011).

8. Umeyama, S. Least-Squares Estimation of Transform Parameters Between Two Point Patterns. IEEE Trans Pattern Anal Mach Intell 13, 376–380 (1991).

9. Smink, J. et al. eXTernal Control (XTC): a flexible, real-time, low-latency, bi-directional scanner interface. in 1755 (2011).

Figures

Several objectives governed the headset design. It touches the skin only at three points, behind the ears and between the nose and the forehead, and runs at an approximately 1cm distance around the forehead. Its extension was minimized to provide motion freedom inside a coil and it does not obstruct the wearer’s eyesight.

The pulse sequence diagram of the 3D-TFE (MPRAGE) sequence used in this work. Gradient tones were inserted onto the crusher gradients and at t = 900ms of the 1s-long inversion recovery phase. Field probe excitation was performed in each inversion recovery phase and every 16ms during the TFE phase.

The spectrum of the rigid body parameters contains significant contributions up to 12Hz (a,b). Even with no intention to move, the head drifted up to 3mm and 2° over a 32 minute period. The peaks in (c,d) are swallowing events one of which is shown at high resolution in e.

Image reconstructions of Experiment 1 in which the volunteer was instructed to remain still. In the left-hand column no PMC was performed, the right-hand column shows the scan with PMC. The images with PMC have sharper white/gray matter transitions and lack white matter brightening present in the uncorrected images (arrows).

Image reconstructions of Experiment 2 with instructed motion. Without PMC (left), the images in all planes are heavily corrupted with artifacts whereas gradient tones-enabled PMC significantly improves their quality (right-hand column). Motion patterns for both experiments were very similar, as indicated in the bottom row



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0337