Details about PET reconstructions, MRI acquisition parameters, and MRI B1 and B0 homogeneity analysis are available in the Additional file 1.
PET/MRI system description
All scans were performed on a Bruker 7 T MRI scanner (Bruker Biospin, Billerica, MA) with a Cubresa NuPET™ insert (Cubresa, Inc., Winnipeg, MB). Two 35 mm 1H volume transceiver coils were used to ensure that the PET/MRI evaluations were not dependent on a single MRI coil (Bruker Biospin, Billerica, MA and PulseTeq, Chobham, UK). PET/MR image registration and analysis were performed using VivoQuant v3.5 (Invicro, LLC, Boston, MA).
PET quantification
A 3 mL tube was filled with either fluorine-18, gallium-68, copper-64, zirconium-89, or yttrium-86 and placed inside a 15 mL conical tube filled with 2% agarose. The starting activity was determined using a CRC-15R dose calibrator (Capintec, Inc., Florham Park, NJ). This tube was placed at axial center of the PET insert, and a 12 h scan was recorded. For long-lived nuclides, multiple 12 h scans were recorded. Results were divided into different activity levels (low: less than 13 MBq (350 µCi); medium: 13–23 MBq (351–620 µCi); high: greater than 23 MBq (621 µCi)). A Quantification Calibration Factor (QCF) was determined for each nuclide and activity level in the PET insert outside the MRI magnet. This experiment and analysis were repeated with the PET insert inside the MRI magnet using the Bruker 35 mm coil. QCFs were determined for certain nuclides without continuous MRI acquisitions or with a continuous RARE MRI or FISP MRI acquisition.
To test reproducibility, phantoms with 0.37–2.59 MBq (10–70 µCi) of fluorine-18 were each placed in axial center of the PET insert inside the MRI magnet. Each tube was scanned for 30 min along with a continuous RARE MRI acquisition, and the activity from the PET image was compared to activity measured by the dose calibrator. This experiment was repeated using phantoms with 0.37–1.67 MBq (10–45 µCi) of gallium-68. A Bland–Altman analysis [20] was performed to calculate the bias and 95% limits of agreement of the activity measurements.
PET linear range
A Bland–Altman analysis [20] was performed using the QCF graphs for fluorine-18 and gallium-68 outside and inside the MRI magnet to determine how the linear range changes in the presence of the MRI magnet. The bias and 95% limits of agreement were calculated, and values within the 95% limits of agreement were considered within the linear range for that radionuclide.
PET signal-to-noise (SNR)
The fluorine-18 datasets from the QCF analysis were used to calculate SNR for each 30 min time point. SNR was calculated by subtracting the signal from the noise from the true signal and then dividing by the standard deviation of the noise. This was performed for data with the PET insert outside the MRI magnet, inside the MRI magnet with no MRI acquisition, and inside the MRI magnet with a RARE MRI acquisition and a FISP MRI acquisition.
PET Spatial resolution
A Derenzo phantom (Phantech Medical, Madison, WI) with node sizes of 1.0, 1.1, 1.2, 1.3, 1.4, and 1.5 mm was filled with ~ 3.7 MBq (100 µCi) of fluorine-18 and placed at axial center in the PET insert outside the MRI magnet. A 15 min PET scan was recorded. The phantom and insert were then moved into the MRI magnet with the Bruker 35 mm coil. A 15 min simultaneous PET/MRI scan was recorded using a continuous RARE MRI acquisition. This experiment was repeated using the Derenzo phantom filled with 3.7 MBq (100 µCi) of gallium-68.
PET partial volume effects
A partial volume correction (PVC) phantom (Phantech Medical, Madison, WI) with 5.55 MBq (150 µCi) of fluorine-18 was placed inside the PET insert. Two 15 min PET scans were recorded with the phantom and insert inside the MRI magnet using the Bruker 35 mm MRI coil and outside the MRI magnet. Using the phantom’s reference region (which assumes 100% activity recovery), recovery coefficients were calculated for spheres of different diameters. This experiment was repeated using 9.25 MBq (250 µCi) of gallium-68.
PET respiratory gating
The Derenzo phantom with ~ 3.7 MBq (100 µCi) of fluorine-18 was attached to a lever and placed in the axial center of the PET insert outside the MRI magnet. The phantom on the lever was manually moved up and down about 2–3 cm inside the PET insert continuously throughout a 15 min PET scan. A pneumatic pad was also placed inside the insert to detect the motion of the phantom (SA Instruments, Stony Brook, NY), which was relayed to the Cubresa software. PET image reconstruction could then be performed with respiratory gating by neglecting detected counts during periods of motion, or without respiratory gating by using all detected counts to reconstruct the image.
The phantom was also scanned with the PET insert inside the MRI magnet using the Bruker 35 mm MRI coil, moving up and down about 3–4 mm during a 15 min simultaneous PET/MRI scan using a continuous RARE MRI acquisition. PET image reconstruction was performed with and without respiratory gating. The experiment was repeated with ~ 3.7 MBq (100 µCi) of gallium-68.
MRI signal-to-noise, linearity
A 15 mL conical tube filled with 20 mM CuSO4 was used in the following 6 scenarios: both 35 mm MR coils were tested without the PET insert, with the PET insert turned off, and with the PET insert turned on.
A MultiSlice, MultiEcho (MSME) spin-echo acquisition was optimized to test sensitivity and linearity. The average signal-to-noise of the tube in the axial image was plotted vs. slice position. The width and height of the tube was also measured in each axial image, and the diameter of the tube was plotted in each dimension vs. axial slice position to determine linearity. These tests were repeated for all 6 coil/insert scenarios. The tube of 20 mM CuSO4 was moved in the axial direction, and these scans were repeated to ensure the entire field of view (FOV) of both MR coils were included in the tests.
MR representative images
Single-slice MSME, RARE, FLASH, MGE, True-FISP, FID-FISP, EPI, and UTE images were acquired in axial, coronal, and sagittal orientations. This set of images was repeated for all 6 coil/insert scenarios.
MRI B1 and B0 homogeneity
A MSME image set was acquired with the same parameters used for the linearity tests. However, 45° and 90° excitation angles were used. The true excitation angle was then calculated along the axis of the sample. A B0 map was also created, which was plotted vs. axial position. These tests were repeated for all 6 coil/insert scenarios.