Validation of the Octamouse for Simultaneous ¹⁸F-Fallypride Small-Animal PET Recordings from 8 Mice

DISCUSSION

The goal of this study was to establish and validate the use of the Octamouse, an acrylic insert for simultaneous small-animal PET scanning of 8 isoflurane-anesthetized mice, to optimize the yield of data acquisition from a single radiosynthesis. We first conducted phantom studies to assess scatter effects arising from the Octamouse on the quantitation of sources placed within the chambers and systematically investigated the contributions of several postprocessing steps on the quantitation of ¹⁸F-fallypride binding to dopamine D_2/3 receptors in the striatum of living mice. There have hitherto been few reports describing a multichamber approach for small-animal PET recordings. One study has described an insert for simultaneous nose-to-nose scanning of 2 rats (4). Inserts have also been used for obtaining precise repositioning of animals in serial PET studies or multimodal imaging studies of rats (11) and also mice (12). Simultaneous small-animal PET recordings have been obtained in as many as 4 mice (5). However, this is the first report, to our knowledge, on the use of a multichamber holder for the scanning of 8 mice, and at least 16 mice from a single radiosynthesis. Although the use of an acrylic insert presents a clear advantage in economy, especially when large groups and long dynamic acquisitions are required, several technical issues relevant to quantitation are raised.

The phantom studies showed linear response of the small-animal PET scanner for a wide range of source radioactivities placed within the Octamouse chambers. Counting efficiency was close to 82%, as expected from the geometry of cylindric sources with diameters twice the full width at half maximum of the small-animal PET scanner. However, the phantom images revealed the presence of considerable radioactivity throughout the air spaces of the chambers, which we attributed to scatter from the relatively massive Octamouse. Indeed, application of a standard scatter-correction algorithm substantially improved the image quality (Fig. 1B). The line profiles placed across the sources indicated that scatter correction decreased the trough activities toward the true background activity, which propagated to a 6% increase in ¹⁸F-fallypride BP_ND. This effect was highly significant, because it occurred in almost every animal. Thus, scatter is an avoidable source of bias in the quantitation of signals arising from within the Octamouse.

We have previously characterized the binding of ¹⁸F-fallypride in the striatum of mice scanned in groups of 2–4 animals without an acrylic holder (1,2). In those studies, we used a zoom factor of 1.3 to restrict the reconstructed volume, which led to a decreased voxel size and, consequently, fewer partial-volume effects, in turn propagating a somewhat higher BP_ND. However, because the 1.3 zoom factor does not encompass the entire Octamouse, the attenuation and scatter correction would not properly account for mass outside the field of view. Lowering the zoom factor to 1.0 resulted in a more accurate attenuation correction throughout the field of view of the small-animal PET scanner but reduced the apparent magnitude of ¹⁸F-fallypride BP_ND by 14% (Table 1; Fig. 3).

Mass effects can potentially influence the outcome measure in small-animal PET studies with ¹⁸F-fallypride and other tracers. We have earlier shown by calculation that the occupancy of dopamine D_2/3 receptors is less than 5% in mice receiving 10 MBq of ¹⁸F-fallypride (2). In the present study, we made a more systematic test of possible mass effects by comparing BP_ND estimates in groups of mice, with ¹⁸F-fallypride specific activity declining by a factor of 3; we found no associated decline in specific binding, indicating a lack of mass effects within the range of specific activity (140–50 GBq/μmol) encountered in our laboratory. Much more problematic with respect to small-animal PET quantitation is the difficulty in obtaining adequate spatial normalization for objects as small as the mouse brain. This should be remedied in the future, with the advent of hybrid small-animal PET/CT and small-animal PET/MRI but remains for the present a manual task requiring considerable skill and practice. Because BP_ND estimates are highly vulnerable to the misplacement of the VOIs, especially for the reference region, the manual registration is a potential source of experimental error or bias. However, our inter- and intraoperator covariances in the estimation of ¹⁸F-fallypride BP_ND (10%) were of the same scale as the variance with a subpopulation of 14 mice, indicating that skillful operators can obtain adequately reproducible quantitation.

In our earlier ¹⁸F-fallypride small-animal PET studies, we found a mean striatal BP_ND of 10 in saline-treated mice; these animals, scanned in groups of 2–4 with a zoom factor of 1.3, had all received a radiochemical dose of 10 MBq of ¹⁸F-fallypride. In the course of the present specific activity study with 14 mice, we administered only 6 MBq of ¹⁸F-fallypride and found a mean BP_ND of only 6. For the entire present population of 32 mice, there was a highly significant positive correlation between the injected dose and BP_ND. Because the magnitude of BP_ND is, in general, a steady-state ratio between the high binding and reference region activities, there is no a priori reason to expect dose effects, so long as receptor occupancy is minimal. Because mass effects are excluded, we had to find an alternative explanation for this unexpected phenomenon. Initial considerations focused on the potentially biasing effects of spill-in from the cranium, which can be quite significant at late ¹⁸F-fallypride circulation times (2). Defluorination of certain gaseous anesthetics in rodent liver microsomes proceeds as a zero-order process (13,14), which might have predicted mass effects on the rate of accumulation in bone of ¹⁸F-fluoride derived from the hepatic metabolism of ¹⁸F-fallypride. However, the time–activity curves from the petrosal bone did not reveal any differential labeling as a function of specific activity.

The relatively heavy acrylic holder is a source of scatter, which biased BP_ND estimates, as noted in the “Discussion” section. The application of a scatter-correction algorithm attenuated by 20% the slope of the relationship between individual injected dose and apparent BP_ND. The remaining 80% of the correlation must arise from other factors, because the scatter-corrected striatal AUC in the high-dose mice exceeded that in the low-dose mice, whereas the mean cerebellum AUCs did not significantly differ, and the petrosal uptake had only a trend toward increased signal (Fig. 5). Using a rat-sized phantom in the Inveon DPET scanner, Kemp et al. found an increasing mean relative counting rate error with increasing radioactivity (15), especially when the total radioactivity in the field of view exceeded 74 MBq for a rat-sized phantom. Given that the Octamouse is larger than the rat phantom, we can predict that the cutoff value should be lower, perhaps 50 MBq.

When total radioactivity in the field of view exceeds a certain limit, dead-time correction (which is applied globally) alters the contrast between regions of high and low radioactivity, thus plausibly accounting for the increase in apparent BP_ND. Indeed, our experiments with 10 MBq per mouse resulted in total activity as high as 80 MBq within the Octamouse. To avoid faulty dead-time correction, future Octamouse studies will be limited to a total of 40 MBq (8 × 5 MBq), this change having the additional advantage of reducing the maximum required injection volume to 160 μL per mouse.

We have considered some of the many possible sources of error and bias in BP_ND estimation by small-animal PET. In addition, it must be considered that the linear graphical Logan analysis used in the present study (9) is inherently vulnerable to noise effects, which led, in our hands, to underestimated voxelwise maps of ¹¹C-raclopride BP_ND in the pig brain (16) and ¹⁸F-MPPF BP_ND in the rat brain (3). Quantitation of ¹⁸F-fallypride BP_ND in mice is vulnerable to effects with opposing biases, perhaps none so important as the effects of partial volume; the size of the mouse striatum (50 mm³) is such that small-animal PET estimates of ¹⁸F-fallypride BP_ND are less than half of the true value obtained by dissection of the mouse brain ex vivo (2). In the 2 earlier mouse small-animal PET studies with ¹⁸F-fallypride, the BP_ND estimates were 13 (17) and 3.2 (18); this difference likely reflects the high resolution of the quad-HIDAC PET (17) used by Honer et al. (<1 mm), and the relatively lower resolution of the Advanced Technology Laboratory Animal Scanner (National Institutes of Health) and Siemens Focus 120 scanner used by Skinbjerg et al. (1.6 mm) (18). Rat small-animal PET studies found ¹⁸F-fallypride BP_ND of 7 (19) and 8 (20), or a striatum-to-cerebellum ratio of 8 (21), whereas Tantawy et al., using a full-compartmental analysis of 2-h dynamic recordings, found a BP_ND in rat striatum of 13 (22). Thus, the increment in striatal volume between rats and mice does not seem to consistently result in improved quantitation of BP_ND, although a direct comparison of mouse and rat has not yet been made.

Whereas monkey PET studies found striatum-to-cerebellum ratios ranging from 3 to 9, depending on the time frame (6,21,23), prolonged recordings ¹⁸F-fallypride gave BP_ND estimates of 27 (24) and 23 (25). Thus, results in the monkey compare well with those in the human brain (26–28), as expected for the primate striatum, which is 10- to 20-fold larger than in rodents. Disregarding the possible attainment of submillimeter small-animal PET resolution, more accurate BP_ND estimates in mouse striatum might be obtainable through the use of a partial-volume correction (29).