Using the NEMA NU 4 PET Image Quality Phantom in Multipinhole Small-Animal SPECT

DISCUSSION

The aim of this study was to investigate the usefulness of the NU 4 IQ phantom and parameters to evaluate the image quality performance of a multipinhole small-animal SPECT system.

It was observed that smaller rods need more iterations to obtain maximum RC_rod, but %STD_RC increased rapidly as a function of the number of iterations (Fig. 1). Application of postreconstruction spatial filtering reduced %STD_RC considerably but also led to a smaller RC_rod. This trade-off between recovery of small structures and image noise, influenced by the number of iterations and spatial filtering, should be considered in the reconstruction settings for small-animal experiments. The optimum settings will of course depend on issues such as the size of the structures to be imaged, scan duration, total activity, and the signal-to-background ratio for these structures.

Somewhat unexpectedly, for small numbers of iterations, the use of scatter correction led to smaller activity recovery, especially for the 1-mm-diameter rod, and to a smaller extent for the 2 mm-diameter rod, as shown in Figure 2A. In contrast, %STD_RC was not influenced using scatter correction, as shown in Figure 2B. The outlier in %STD_RC for the 1-mm rod with scatter correction and 6 iterations can be explained by noting that according to NEMA NU 4, the line profile through the rod over which the average activity concentration is calculated is positioned such as to yield maximum total activity along the line. In cases of high noise, the implication is that these profiles will not always be calculated over exactly the same transverse pixel positions. Indeed, for this outlier, the line profile was shifted by 1 pixel (x) and by 2 pixels (y) relative to the profiles of the other data points in Figure 2B.

%STD_unif was not affected by scatter correction and shows the same trends as %STD_RC. %STD_unif increases as a function of the number of iterations and decreases when postreconstruction spatial filtering is applied (Fig. 3).

SOR was found to decrease as a function of the number of iterations (Fig. 4), showing that cold regions in hot environments can be recovered only by using many iterations, as has previously been described for PET image reconstruction (17). Furthermore, SOR_wat was much larger than SOR_air—a finding that is clearly attributed to the fact that photon scatter in water is stronger than in air. It is interesting to note the differences in scatter-related spillover effects between SPECT and PET. Related to the detection of photon pairs forming lines of response in PET, spillover is to a large extent caused by scatter outside the cold compartments, causing the corresponding lines of response to be falsely attributed to these compartments, independently of their density and scatter efficiency. This observation was made in small-animal ¹⁸F-PET (18), leading to almost equal values for SOR_air and SOR_wat. Because of single-photon detection in SPECT, however, the activity measured in cold regions should ideally be proportional to the residual photon scatter originating in the cold region itself, as explains the observed difference between SOR_air and SOR_wat. Scatter correction had a positive effect on image quality because SOR decreased, for both the water-filled and the air-filled compartments. It is, however, unknown why this decrease was larger for air than for water. A detailed comparison of the present results with those for a state-of-the art small-animal PET scanner can be made using information provided by Visser et al. (19).

The effect of the activity level was especially seen in the image noise, which increased rapidly for decreasing activity (Fig. 5). On the other hand, RC_rod and SOR did not systematically change as a function of activity, although these parameters were fluctuating around their average value for lower activities (Supplemental Fig. 1 shows data for RC_rod).

The scan duration of 1 h and activity of 10 MBq, considered typical for WB SPECT mouse scans, are larger than the values prescribed for small-animal NEMA PET (20 min and 3.7 MBq of ¹⁸F, respectively) because WB SPECT has lower sensitivity than WB PET. Shorter scan times and lower activity, however, could be used when scanning isolated organs or tumors because reducing the field of view in multipinhole SPECT generally increases effective sensitivity.

As expected, the effective WB sensitivity of 117 cps/MBq is smaller than the centered-point-source sensitivity of 700 cps/MBq as presented elsewhere (2). The difference between these values reflects the combined effect of photon attenuation and partial coverage of the total phantom activity in individual bed positions. Effective WB sensitivity is related to the choice of phantom. The NU 4 IQ phantom clearly is somewhat larger than a typical mouse but much smaller than a rat. Instead, one could also use the NEMA NU 4 2008 counting rate and scatter fraction phantoms (radioactive line source in cold polyethylene cylinder (3)), for which differently sized mouse- and ratlike sizes have been defined.

The phantom experiments provided quantitative and objective data on the effect of different reconstruction and acquisition settings for a range of activity levels. Depending on the specific imaging goals, these image quality parameters can guide the researcher in the choice for optimum imaging settings in real small-animal experiments, for example, smoothing of images with low noise levels, suppression of scattered photons for better quantification, or detection of small organ structures. As such, the WB mouse scans (Fig. 6 and Supplemental Fig. 3) illustrated the difference between the various settings resulting in how much detail can be distinguished and the amount of image noise. Especially for smaller structures such as the small bony components of the sternum or the ribs, the differences were clearly visualized. The same characteristics as in these mouse images can be noted in the phantom images of Figure 6 and Supplemental Figure 4.

The U-SPECT-II scanner can be equipped with different collimators. Besides the one used in the present study (pinhole and bore diameter of 1.0 and 98 mm, respectively), collimators with pinhole diameters of 0.35 and 0.6 mm and a bore diameter of 44 mm are commercially available (2). However, because of the use of scanning-focus mode, the maximum diameter of objects to be completely scanned in these collimators is 28 mm, prohibiting complete WB scans of the NU 4 IQ phantom (outer diameter, 33.5 mm) since the outer part of the phantom would collide with the collimator wall when the opposite side is moved into the center of the field of view.

The present paper has dealt with the image quality aspect of small-animal SPECT. In a broader sense, one could discuss how a future NEMA “NU 5” standard for SPECT should be defined and to what extent it would differ from NEMA NU 4 for PET. In short, we believe that image quality parameters could be defined in the same way as for PET, and the NU 4 IQ phantom would be appropriate for WB mouse or rat scans. However, for scanning in high-spatial-resolution mode—for example, single-organ or single-tumor imaging—it should be discussed whether an additional phantom with smaller internal structures and smaller outer diameter would be appropriate.

Spatial resolution as defined by NEMA NU 4 (point source, reconstructed using filtered backprojection) cannot easily be adapted to multipinhole SPECT, because the iterative reconstruction algorithms used do not yield univocal results for point or line sources, and filtered backprojection is normally not available. Therefore, spatial resolution expressed as recovery coefficients for small rods in an image quality phantom would be more appropriate. However, because recovery coefficients depend on reconstruction settings, trade-offs with, for example, noise and spillover should be considered. For comparison of different vendors or collimators, the accompanying image quality parameters should be specified along with the recovery coefficients.

Count losses and scatter fraction could be measured in multipinhole SPECT using the same (mouselike and ratlike) phantoms as in NEMA NU 4. However, we have found that count loss is not of much importance in view of the highly linear relation between counting rate and source activity, even for high activities (Supplemental Fig. 2).

Finally, as already mentioned, sensitivity measured using a point source as in NEMA NU 4 is not appropriate for multipinhole SPECT. Instead, standardized, larger phantoms should be defined.