Motion Correcting Complete MR-PET Exams Via Plot Tone Navigators
Thomas Koesters1, Ryan Brown1, Tiejun Zhao2, Mattias Fenchel3, Peter Speier3, Li Feng1, Yongxian Qian1, and Fernando Emilio Boada1

1Radiology, New York University, New York, NY, United States, 2Siemens Medical Systems, Malvern, PA, United States, 3Siemens Medical Systems, Erlangen, Germany

Synopsis

We demonstrate the use of an external RF signal as a mean to provide motion tracking information for motion-state sorting of k-space line during dynamic MRI scans.

Introduction

The introduction of simultaneous MR/PET scanners has, for the first time, provided a synergistic imaging platform where the simultaneously acquired dual-modality data can be used to provide significant improvements in image quality, interpretation and quantification. Motion correction stands as one application where immediate benefit can be garnered from the use of such synergies. Motion correction of PET images relies on MR-based motion tracking information that can be used to sort the PET listmode data into different motion states or gates. The MR images corresponding to these gates are then used to modify the system matrix and produce motion-corrected PET images. Prior implementations of this approach have relied on the use of a self-navigated sequence, such as a stack of stars [1], where the motion tracking information is derived from the center of k-space. While effective, however, this approach only allows the correction of the listmode data acquired when the motion tracking sequence is used. Recently, it has been demonstrated that self-refocused [2] MR signals or signatures of coil load variations [3,4] can be used to monitor respiratory, and sometimes, cardiac motion. In this work we demonstrate that the use of a reference RF signal (the pilot tone, PT) can be used, as proposed in [5], as an effective means to identify individual motion states throughout the duration of an entire MR/PET examination.

Methods

The PT navigator approach was implemented on a Siemens MAGNETOM mMR (Siemens Healthcare, Erlangen, Germany) using a 2cm surface coil (Fig 1), which was placed outside the bore of the magnet. The coil was driven by a signal generator with -20dBm amplitude. The PT frequency was between 7 and 50Khz away from the center frequency of the scanner (123.216MHz) to avoid interference with the desired MRI data. Data were acquired with both conventional gradient echo (TR/TE=20/10ms, BW=260Hz/Pixel, 1x5mm slice) as well as a golden angle radial imaging (RI) in-house-modified prototype sequence (TR/TE=7.92/4.94ms, BW=680Hz/Pixel, 48x6mm slices, 2002 radial views) equipped with an additional rosette [6] navigator readout, which allows for comparison of the PT navigator to the conventional k-space center self-navigation [7] and coil fingerprinting approaches [2]. The PT data from different coils in the receive array were used to sort the measured echo into motion 'bins'. These motion bins were then used to reconstruct the images corresponding to the different motion states.

Results

Figure 2 shows a GRE image from a normal human volunteer acquired while using the PT signal to monitor the motion throughout the scan. A plot of the temporal variation of the PT signal illustrates that the variations in its amplitude are correlated to the motion of the chest cavity. Further analysis of this signal allows the synthesis of motion bins from which individual motion states can be reconstructed (Figure 3). Use of the rosette-navigated RI sequence allows exclusive monitoring of the motion signal via the rosette readout (Figure 4). Fourier analysis of this signal demonstrates the presence of a respiratory peak (Figure 5). This signal correlates well with the tracking signal from the center of k-space (Figure 6)and can, likewise, be used to generate motion bins (Figure 7) from which 3D motion states can be reconstructed (Figure 8).

Conclusions

We have demonstrated that an external RF signal, as proposed in [5], can indeed be used to provide effective respiratory motion tracking information in real time. The technique is straightforward to implement, requires minimal hardware, and is transparent to the pulse sequence since specialized navigator modules are not required. Integration of this tracking signal with MR image reconstruction via sorting into motion bins is compatible with most MR sequences and can provide a means to fully navigate an entire MR/PET examination.

Acknowledgements

Supported in part by PHS Grant P41 EB 017813

References

[1] Peters D. C., et al., Magn Reson Med. 2000, 43: 91-101.

[2] Vahedipour K., et al., Proc. ISMRM 2015, pp. 813.

[3] A. Andreychenko, et al., Proc. ISMRM 2010, pp. 92.

[4] I. Graesslin, et al., Proc. ISMRM 2010, pp. 3045.

[5] P. Speier et al. Proc. ESMRMB 2015, 129: 97-98.

[6] Noll D. C., et al., Magn Reson Med. 1998, 39: 709-16.

[7] Grimm, R., et al., Med. Image Anal, 2015, 19: 110-20.

Figures

Figure 1: Pilot Tone RF coil. The coil is placed outside the bore of the scanner and driven via an external signal generator operating 7-50KHz off-resonance depending on the bandwidth of the imaging sequence.

Figure 2: Single slice GRE image of a normal human volunteer alongside the pilot tone signal. The position of the signal is controlled by the resonance offset in the external oscillator (7KHz in this case).

Figure 3: Pilot Tone signal from the GRE acquisition in figure 2 (top) alongside the motion bins sorting information (bottom). In the bottom plot the vertical axis corresponds to motion state while the horizontal axis corresponds to k-space view number.

Figure 4: K-space center (thicker line) and Pilot Tone (thinner line) signal from the Rosette-navigated sequence. The k-space center signal is monitored continuously in this case by simply zeroing the gradient amplitude of the rosette readout.

Figure 5: Spectrum of the Pilot Tone signal in figure 4. The respiratory peaks are clearly visible on both sides of the spectrum.

Figure 6: Comparison of the pilot tone (red line) and k-space center (from center of radial readout) navigator signals for the radial sequence.

Figure 7: Bin sorting plot for the radial imaging scan. In this plot the vertical axis corresponds to motion state while the horizontal axis corresponds to k-space view number.

Figure 8: Synthesized motion states from the motion tracking information obtained with the Rosette-navigated RI sequence.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
4250