Synopsis

A 24 channel head and neck coil for PET/MR with an integrated moving radioactive source is presented to enable transmission measurements for accurate attenuation mapping.

Purpose

Multimodal imaging has shown to bring additional and valuable information for medical diagnostic imaging [1]. The combination of different modalities is always associated with technical and conceptual challenges, in the case of Positron Emission Tomography (PET) and MR one of these challenges is attenuation correction (AC).

We have designed a PET-optimized receive-only MR head and neck coil array. This MR coil integrates a system for AC, consisting of a point source moved hydraulically in a helical hose around the imaging coil (Fig.1). This system enables transmission measurements at 511 keV, from which accurate AC maps, including RF coil and patient, and measured at the correct energy, can be calculated.

Methods

RF coil: the coil array consists of 24-channels distributed in 4 rows with 3/7/7/7 elements (Fig. 2a). There are 17 quadratic elements (9.5cm x 9.5cm) and 7 trapezoidal elements (9cm/5cm x 10.5cm). Within rows, transformer decoupling was implemented, while neighboring rows were decoupled by overlapping the elements. Balun matching networks and preamplifier decoupling were implemented for all channels. The strongly attenuating components like preamplifiers were placed outside the PET field-of-view as recommended for PET-MR coils [3] and thin (0.75 mm diameter) copper wire was used for the coil elements. A photo of the final construction is shown (Fig.2b).

The coil housing was designed using 3D CAD software, and 3D-printed from laser-sintered polylaurinlactam. The MR coil is built on an elliptical inner former, where the elements and the electronics were placed. This inner former is inserted in a cylindrical outer shell acting both as cover for the RF coil and to support the hydraulical system, leaving free space for the shoulders. The housing thickness was minimized while maintaining stability, with a maximum wall thickness of 4 mm.

MR images (3D GRE, TR/TE=30/2.39 ms, FA=15°, in plane resolution=0.56x0.56 mm2, slice thickness=1.4 mm, MA=640x640x256) of a gel head phantom (1% Agarose, 65 µmol/kg Magnevist, 0.1% NaCl, and 0.05% NaN3) were acquired. S-parameters of the coil were measured using a network analyzer with multiport test set (Agilent Technologies, E5071C/E5092A). To obtain noise correlation information, noise data was acquired using a free induction decay sequence without RF excitation. Raw data was post-processed with in-house developed MATLAB scripts.

Hydraulical system: the information about the position of the point source (pellet) will be synchronized with the PET list-mode data during acquisition using optical triggers distributed around the cylindrical outer shell. This should allow simplified and improved reconstruction of the transmission image. Currently the hose is filled with distilled water, in the final implementation it will be exchanged by heavy water, avoiding MR signal from the hose.

Monte-Carlo simulations (Geant4 Application for Tomographic Emission, GATE [4]) were performed to determine global scatter-fraction on a virtual head phantom (Duke, Virtual Population, IT'IS Foundation) placed in a model of the RF coil and the transmission system. For the source, a 20 MBq pellet was used.

Results

Images of the phantom acquired in the complete set up, using distilled water, are shown in Figure 2. S-parameters and the noise correlation matrix are shown in Figure 3.

For the attenuation correction method, Monte-Carlo simulations showed that the loss in photon global scatter-fraction of the entire hardware set-up including a realistic head phantom is 16.4%, enabling PET reconstruction with standard algorithms.

Conclusion

Acknowledgements

References

[1] Wehrl et al., J Nucl Med 2015.

[2] Keereman et al., Magn Reson Mater Phys 2013.

[3] Sander et al., Magn Reson Med 2015.

[4] Jan et al. Phys. In Med. 2004