There was no significant difference between the sensitivity of the PET insert inside and outside the MR bore. The relatively high sensitivity of 10.15% achieved by the system can be attributed to its geometry, i.e. the long AFoV and narrow ring diameter, as well as the high detection efficiency of the dual-layer LYSO detectors. This sensitivity compares favourably with other MR-compatible PET inserts, such as the Bruker Si 103 (Bruker Corporation, Billerica, USA) with 11% absolute peak sensitivity, and dedicated small animal PET systems, such as the nanoScan PET/CT (Mediso Medical Imaging Systems, Hungary) with 8.41% absolute peak sensitivity [10], which have similarly long axial fields of view and employ similar detector materials. This value is comparable as well with the absolute peak sensitivity of the MR Solutions PET insert (model 80 − 2), 7.9% (measured inside the MR bore without RF coil), that has a longer axial FoV, 103 mm but a wider inner diameter bore, 116 mm [23] (Table 4).
The possible effect of the static magnetic field on spatial resolution was tested by performing measurements outside the MR bore and inside the MR bore with the MR idle. Spatial resolution was not significantly affected by the presence of the magnetic field. This is probably because of the relatively low static magnetic field of our 3T system. At higher field strengths, the positron range is slightly reduced in directions orthogonal to B0 (transverse and radial) leading to improved in-plane spatial resolution [31, 32]. However, axial elongation of positron range is also observed which degrades axial resolution and introduces artefacts [33]. Improvements in spatial resolution are also dependent on the energy of the positron emitter [34]. In the future, we will repeat this study in the 7T system to see whether we observe these effects at higher field strength. The measured radial/tangential and axial components of spatial resolution were < 1.2mm near the centre of the FoV which compares favourably with the MADPET4 [14] and MR Solutions PET insert 80 − 2 [23] (Table 4) but also with other current-generation small animal PET/MRI systems [35], [11] and [16]. Additionally, radial resolution was relatively uniform over 5-15mm radial offset from the centre of the FoV, indicating that the DOI method implemented on this system is effective at minimising resolution degradation due to parallax error.
Studies performed with the IQ phantom demonstrated that there is no significant difference in quantitative accuracy between scans performed outside or inside the MR bore or during simultaneous imaging. There was no significant difference between the uniformity of the phantom inside the bore and outside the bore, with the observed differences being within experimental error. The consistency in uniformity and spill over ratios suggest that attenuation and scatter correction accuracy is not affected by the magnetic field or pulsing of RF gradients. Similarly, the consistency of recovery coefficients across scanner configurations suggests that there is no impact of the static magnetic field or RF pulsing on partial volume effects, consistent with our spatial resolution results. Thus, the same scatter and attenuation corrections can be applied in any configuration for low energy radioisotopes such as 18F. A bubble was present inside the 1mm rod of the IQ phantom, which prevented accurate calculation of the recovery coefficient for this rod. However, since the IQ phantom was not re-filled between scans, results can be compared across the three configurations studied. The difference in RC of the fourth rod is within experimental error.
Concerning the scatter and random rates, the increase observed inside the bore can be explained either by higher backscatter from the magnet or by impacts of the gradient fields and RF pulsing on the timing resolution of the detectors. Degraded timing resolution would increase the number of random events registered as coincidences and, conversely, decrease the number of true coincidences registered within the coincidence time window. The combination of increased scatter and randoms rates and decreased trues rate explains the decrease in NECR seen inside the MR bore, and during simultaneous PET/MR (at high count rates). The NECR of the MR Solutions PET insert 80 − 2 presented a 30% decrease when measured inside the MR bore with the RF coil in comparison with the NECR measured without the RF coil [23], (Table 4). The NECR of the GE SIGNA clinical PET/MR [18] when the MR was running was also found to be lower than the NECR when the MR was idle. Grant et al. [18] and Deller et al. [20] identified interference of the RF pulsing on the PET electronics as the cause of longer dead time, and therefore, loss of counts: this would explain the decreased NECR when the MR is active. This could also explain our results but does not explain the difference observed between the NECR outside the bore and inside the bore with no RF pulsing. Increased backscatter may explain the difference in these configurations. In addition, the SiPMs may behave slightly differently since there is less shielding from light outside the bore, and SiPMs are very sensitive to light. In general the peak NECR of 175 kcps at 17MBq compares favourably with other SiPM-based systems [23], [35], [11], [16].
At typical activity concentrations, the differences in count loss due to dead time in the three configurations were negligible and dead time correction using a non-paralyzable model and a single dead time parameter of 122ns was sufficiently accurate. Previous studies of simultaneous PET/MR systems have shown that RF leakage into PET electronics results in an increase in dead time [18, 20]. Imperfections in the shielding of PET electronics may account for the differences we observed at higher count rates. The strength and duration of the RF pulses depend on the MR imaging sequence chosen. Therefore, a further investigation with different MR sequences and their effect on dead time may be warranted. However, this effect does not explain the small difference in dead time we observed between scans done outside and inside the MR bore with the MR idle. A possible explanation might be increased backscatter from the magnet causing an increase in the scatter count rate. It should be noted that these small dead time differences can be incorporated into the dead time correction for more accurate quantification at very high activities.
The PET insert exhibited good temperature stability in the range of 15–30˚C. Even during fast MRI sequences (EPI or TOF 3D), which typically raise the temperature of the gradients, and with the gradient chiller temperature at 18.5 ˚C, the SiPM temperatures never reached > 20 ˚C. This is an important result as SiPMs are known to be affected by environmental temperature [36].
We did not perform animal studies on the benchtop due to the lack of portable physiological monitoring systems during these experiments. PET images acquired with the MR idle and during simultaneous imaging were highly comparable for both radiotracers. Taken all together, the NEMA and non-NEMA experiments demonstrate that the performance of the PET insert is not affected by being operated inside the MR bore or as a standalone unit on the benchtop. This flexibility offers some practical advantages. First, it allows more efficient use of the PET insert in a busy preclinical imaging laboratory since the PET insert can be used on the benchtop concurrently with MR-only studies. Second, more complex PET studies requiring, for example, microdialysis or voltammetry, may need to be performed outside the MR bore because of either space or MR-compatibility issues with ancillary instrumentation. Finally, in other situations it may be desirable to perform PET scans inside the MR bore, enabling use of complementary information from the two modalities to improve the accuracy and reliability of parameter estimates [4].
Table 4 compares the key NEMA standards measurements between similar PET/MRI systems, namely, the PET insert presented in the current study, model I-402, the MR Solutions PET insert, model I-802 [23] and the MADPET4 [14].