Hyperpolarized (HP) 13C MRI is a powerful tool for imaging tissue metabolism. As HP MRI moves into the clinic, the use of surface coils will become more and more important. Surface coils permit increased sensitivity for small metabolites and coil arrays enable rapid parallel MRI techniques, which are especially important given the relatively short duration available to image HP substrates¹. A major challenge for parallel HP ¹³C MRI is determining the coil sensitivities. Unlike conventional 1H imaging, there is too little natural abundance ¹³C signal for sensitivity calibration. Because of the small matrix size of most HP acquisitions, autocalibrated techniques dramatically limit the achievable acceleration². We propose to address the problem of coil sensitivity measurement by computing sensitivity profiles based on electromagnetic simulation informed by coil positions measured using an array of small fiducial markers filled with a dense signal source (⁵⁵Mn) placed at known locations with respect to the coil conductors. ⁵⁵Mn has a resonance frequency close enough to ¹³C to be detected using the same hardware³. However the resonant frequency is far enough from ¹³C that it can be detected with no background, permitting the locations to be measured automatically and without expending any HP ¹³C signal.

Fiducial markers were constructed from hollow 6.35 mmhigh density polyethylene spheres (Precision Plastic Ball Co, Fig 1) that were filled with approximately 40 µl 3M ⁵⁵Mn solution. A 10 cm x 4.8 cm transmit/receive radiofrequency surface coil (Fig 2a) was constructed and tuned to 32 MHz. Three ⁵⁵Mn markers were placed alongside the coil conductors (arraows, Fig 2a). A CT image of the coil together with the markers was obtained (Fig 2b) in order to determine the relationship between the markers and the conductor path.

MR experiments were performed using a 3T scanner (MR750, General Electric Medical Systems). The RF coil was placed on top of a cylindrical phantom, which was placed into the MR scanner at an oblique angle (Fig 3a, approximately 45 degrees with respect to the z-axis. ⁵⁵Mn fiducial markers were localized via three 1D gradient-echo projections with a readout gradient along either the x-, y-, or z-axis (BW 125khz, readout gradient 4 G/cm, 313 spectral points). Spatial position of the fiducial markers were determined by extracting the peak positions on each of the 3 projections. Reference coil sensitivity was measured by obtaining a 2D CSI image of the ethylene glycol phantom (matrix 16x32, FOV 8cm x 16cm, voxel 5 mm x 5 mm, slice 10 mm).

Data processing was performed in Matlab (Mathworks, Inc). The position of the fiducial markers measured using the CT scan was mapped to the MR-derived spatial position by least-squares estimation⁴. The spatial transformation derived from the fiducial markers was then applied to the coil array conductor pattern. The coil sensitivity expected from the conductor pattern was then calculated using the principle of reciprocity and Biot-Savart quasi-static approximation (valid for the long carbon wavelength at 3T).

CT topogram of the coil with fiducial markers is shown in Fig. 2b. After extracting the spatial positions from the CT scan, the digitized conductor path is displayed together with the triangle formed by the three markers (Fig. 2c). Spectra obtained along each spatial projection are shown in Fig. 3b. After applying the 3D least-squares spatial transformation, the transformed CT-derived fiducial markers align with the measured fiducial marker positions nearly perfectly (RMS error 0.54mm). In addition, the oblique orientation of the transformed conductor path matches the orientation of the RF coil in the MR scanner (Fig 3c). The coil sensitivity pattern predicted by the transformed simulated conductor elements matches the measured sensitivity pattern qualitatively. Note the difference in shape between two profiles reflected by the round ethylene glycol phantom. The plot through the midline of both images shows good agreement between measured and calculated values.⁵⁵Mn-based fiducial markers are a practical way to localize RF coils used for hyperpolarized 13C imaging. Using point sources that resonate at a frequency far from ¹³C, the markers can be localized by peaks in a projection image, which is easy to automate. The markers do not appear in the ¹³C image and so they can be left permanently on the coil. Future studies are necessary to determine the optimal number of projections required to localize RF coils reliably. In addition, this method can be extended to coil arrays to determine coil sensitivities for parallel imagining reconstructions.