Background
The European Association of Nuclear Medicine (EANM) procedure guidelines for whole-body fluorodeoxyglucose positron-emission tomography(/computed tomography) (FDG-PET(/CT)) scans for tumour imaging [
1] provide a standardization for the administration of FDG and the acquisition and reconstruction of an FDG whole-body PET scan. Together with quality control standards, these guidelines ensure that the measured FDG tumour uptake is, within certain limits, independent of the system used or the centre where the study is performed. Furthermore, it was discussed that the linear relationship between the patient’s body mass and administered FDG dose would result in a more uniform image quality between patients compared to a constant administered dose. However, it is well known from clinical practice that even after adherence to the guidelines, the image quality of whole-body FDG-PET scans decreases for obese patients, which can result in false-negative PET scans. Thus, a different relationship between dose and body mass or a different patient-dependent parameter, e.g. body mass index (BMI) or lean mass, might be required to obtain an even more constant image quality [
2,
3]. Research on image quality has been performed at various institutes [
4,
5], also with a special focus on obese patients [
6,
7]. Using the signal-to-noise ratio (SNR) in the liver as a measure for image quality, the use of a higher FDG dose per kilogram of body mass (typically 6 to 10 MBq/kg) or longer PET acquisition times per bed position for patients with high body mass were suggested [
5,
8]. However, the optimal relationship has not yet been determined. For children, an optimized dose regimen using an exponential relation between body mass and FDG dose was suggested [
9].
Thus, the aim of this study was to optimize the administered FDG dose as a function of a patient-dependent parameter, providing whole-body FDG-PET images of a more constant quality.
Since the publication of [
1], more PET cameras have been equipped with a time of flight (TOF) option [
10‐
12]. In addition, reconstructions using a position-dependent point spread function (PSF) have also been introduced [
13]. It is possible that these new reconstruction options change the optimal relation between the patient-dependent parameter and the administered FDG dose. Therefore, these different reconstruction methods were taken into account in this study.
Discussion
The current EANM guidelines [
1] advise a linear relation between FDG dose and body mass. The data in Figure
2C,E,G were obtained using a linear relation between the patient’s body mass and the FDG dose, but the scan time per bed position was varied for different classes of body mass. However, this adaption in the scan time can be corrected for, based on Equation
2. The results of this correction are shown in Figure
2D,F,H. From these figures, it is clearly seen that the EANM guidelines result in a decreasing SNR
L with increasing body mass for the Biograph mCT for all reconstructions used in this study. The regression lines in Figure
2A,B were obtained using the data of all patients scanned on the Biograph TruePoint in the first part of the study. However, patients with a body mass above 90 kg received a higher FDG dose per kilogram of body mass for every kilogram over 90 kg (see Table
1). This higher dose might compensate for the decline of SNR
L for heavier patients. However, if two separate regression analyses are performed for these two classes of body mass, it turns out that for both groups the SNR
L decreases significantly with body mass (
p < 0.001 and
p = 0.03, respectively). So even though patients with a body mass above 90 kg were administered a higher FDG dose per kilogram of body mass, the SNR
L also decreases with body mass for this group on the Biograph TruePoint. The first part of the study indicated that for both cameras a quadratic relation between FDG dose and body mass should result in a more constant SNR
L and that this is valid for all the reconstruction methods used. Validation of this new dose regimen showed that a quadratic dose regimen actually results in a more constant SNR
L. These results are consistent with clinical observations, indicating that the use of the SNR as measured in the liver is a good parameter to investigate the relation between body parameters and image quality. The reason that the liver was chosen is that this is the only organ that has a relative homogeneous uptake of FDG. However, SNR
L also reflects variability in physiological uptake. Since heterogeneity of liver uptake was the only exclusion criterion for this study, it is expected that our results can be extrapolated to all FDG whole-body scans.
The fit of SNRnorm to different patient-dependent parameters shows that the fit with body mass has the highest R2, followed by mass per length (the difference of R2 is not significant for both cameras). Fits of SNRnorm with lean body mass, fat mass and BMI turned out to have lower values of R2, sometimes significantly so. Based on these findings and the fact that body mass is the easiest patient-dependent parameter to use, the choice for body mass was made. It may be surprising that fits of SNRnorm with lean body mass, fat mass and BMI have lower values of R2 than the fit with body mass as one would assume that body shape should influence image quality. Apparently, in our population, this is only a minor effect since body mass already explains between 77% and 93% of the variability in SNRnorm. The remaining error in the fit of SNRnorm with body mass showed no trend with the patient’s body mass (data not shown), indicating that including additional powers of body mass will not improve the value of R2 of the fit of SNRnorm with body mass. Of course, the effect of other body parameters may be hidden in the remaining unexplained variance. However, a multi-variate fit would overcomplicate the application of improved dose regimes and probably reduce inter-subject variability only marginally. Since the fit of SNRnorm with body mass has the highest R2 and is easily obtained and a very practical parameter to use, only the relation between SNRnorm and body mass was considered further.
Analyzing the data, we determined that a quadratic relation between the patient’s body mass and administered dose should result in a more constant SNR
L, i.e. an image quality that is less dependent of the patient’s body mass. From Table
3, it is seen that for both TOF and non-TOF systems, the fitting parameter
d is approximately, but not exactly equal to, 1. The obtained relationship between the patient’s body mass and the administered FDG dose (Equation
6) is therefore not exactly quadratic. The standard deviation of the relative error distribution of the fits varies, depending on the reconstruction used, from 7.3% to 15.3%. To determine the influence of this deviation of the parameter
d of the value of 1, the best fit between SNR
norm and body mass for
d = 1 is displayed in Figure
4 too. It is seen from Figure
4 that for three out of the four cases the fit with
d = 1, which would result in an exact quadratic relation between the patient’s body mass and the FDG dose, lies within the 95% CI of the best fit. Only for the Biograph mCT, OSEM3D + PSF + TOF reconstruction, the fit for
d = 1 lies partly outside the 95% CI of the best fit. However, this deviation is small. Therefore, it was reasonable to state that the optimal relation between the patient’s body mass and the FDG dose is quadratic.
Equation 6 was based on the assumption that the SNR scales with the square root of the measured counts (Equation
2), i.e. that
N is proportional to
A, where
A is the amount of activity at the time of administration. This assumption was checked for OSEM3D, OSEM3D + PSF and OSEM3D + PSF + TOF reconstructions by performing scans of different acquisition times of a cylindrical phantom uniformly filled with
68Ge (data not shown). The activity in the phantom was low enough to avoid dead time effects. The fit of the SNR with the scan time showed that SNR is proportional to
t
n
, where the parameter
n varied from 0.44 to 0.47, depending on the reconstruction used. However, it turns out that if these exact values are used in the analysis, the values of
d do not change significantly. For both cameras, the waiting period between the administration of FDG and the start of the PET acquisition was standardized (1 h) and the total scan times were relatively short compared to the half-life of FDG, which makes it reasonable to assume that
N is proportional to
A. On the Biograph mCT, different scan times per bed position were used for different classes of body mass. This introduces a difference in the time between FDG administration and the start of the acquisition of the PET slice containing the liver for patients with different body masses. However, this difference will be only a couple of minutes, which is short compared to the half-life of FDG. Therefore, the use of Equation
2 was justified.
As we have shown, the product of the dose and the time per bed position determines the SNR
L (Equation
6). Therefore, the acquisition time per bed position can, in principle, be varied without influencing the image quality, as long as the administered FDG dose is changed accordingly. However, this study assumes that other effects which influence image quality such as dead time correction as well as random and scatter coincidences are small. Therefore, one should be careful when increasing the FDG dose in favour of a shorter acquisition time per bed position because dead time effects can influence the image quality negatively. In this study the noise equivalent count rate (NECR) remained in the linear range with respect to the administered FDG dose. However, if higher doses are applied, which is, e.g. the case in the USA [
16], this is no longer the case, and dead time effects become more important. For very heavy patients and without adapting the scan time, the quadratic dose regimen results in very high levels of administered dose and system dead time should then be monitored. In addition to the decrease of dead time effects, the use of longer acquisition time per bed position for obese patients also has the benefit of reducing the radiation burden for both the patient and the technician.
A quadratic dose regimen may be considered less practical than a linear relation, although this should not be a problem when using automated dispensing units or a lookup table. Alternatively, one could also linearize the quadratic relation in parts or use a linear dose regimen while adapting the acquisition time per bed position with the patient’s body mass to mimic the quadratic relation according to Equation
6. The latter approach was used on the Biograph mCT, resulting in the data of Figure
2C,E,G.
The value of SNRacc was in this study only determined for the OSEM3D reconstruction of the Biograph TruePoint camera. Because noise has a different structure in PSF and PSF + TOF reconstructions compared to OSEM3D reconstructions, the values of the SNR of different reconstruction methods are not directly comparable, and the value of SNRacc has to be determined separately for different reconstruction methods. The analysis in this paper was performed on two Siemens cameras. Apparently, for the systems tested, the quadratic relation is valid for OSEM3D with or without PSF or PSF + TOF reconstructions. This does not automatically mean that the obtained relationship is also valid for cameras of different manufacturers or for different reconstruction methods. Ideally, for these cameras and reconstruction methods, one should repeat the analysis of the SNRL. Nevertheless, we would not be surprised if our results hold for other systems too.
Under this assumption, the new, quadratic dose
A
q
(MBq) for a patient with body mass
m (kg) depends on the old, linear dose
A
lin (MBq) for this patient through Equation
8 (see Additional file
1):
(8)
where m
T
is the maximum body mass up to which the image quality is considered acceptable in the linear dose regimen.
The results of the simulations were in line with clinical observations and confirm that at FDG activities as recommended by EANM (approximately in the range of 180 of 260 MBq for a patient with a body mass of 75 kg) [
1], image quality, expressed as SNR, remained fairly constant across patients with different body masses when the amount of FDG dose administered was proportional with the square of the patient’s body mass. Simulations suggest that this relationship can be primarily explained by the increased attenuation with increasing tissue mass as dead time and contribution of random coincidences and scattered photons were not included. Although the latter omissions can be considered as a limitation of the simulations, it should be noted that typical FDG activities administered in Europe [
1] are much lower than those applied in, for example, the USA [
16,
17]. Consequently, PET/CT studies in Europe are usually operated in the linear part of the NECR curve, where the contributions of dead time and random coincidences have a much smaller effect on NECR than at higher FDG activities where the NECR curve becomes flat [
4]. SNR in patients can be influenced by physiology as well. Changes in plasma clearance, obesity and/or plasma glucose levels can have an effect on the biodistribution of FDG and thus SNR. However, this study shows that these effects are either rare or not as important as the effect of attenuation. For example, by simply simulating effects of attenuation on image quality (SNR), we were able to closely replicate the clinical findings, i.e. a linear relation between the patient’s body mass and administered FDG dose was not sufficient to achieve uniform image quality across patients. By proportionally scaling the FDG dose with the square of the body mass, a more uniform image noise level as a function of (simulated) patient body mass can be achieved. For patients heavier than approximately 120 kg, the simulations indicate that an FDG dose proportional to a higher power of the body mass is needed to obtain the same SNR as for lighter patients. However, patients in this range of body mass are rare in our settings and thus rare in our analysis. Therefore, we cannot compare in full these specific simulation results to clinical data.
The lowest patient’s body mass in this study was 45 kg. It remains to be shown whether our results can be extrapolated to lower body masses. In addition, it also important to note that the results obtained in this paper are only valid for adults. For children, one should therefore use current international guidelines such as those from the EANM [
18] or optimized dose regimes such as [
9].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
EHdG analyzed the data, performed the statistical analysis and drafted the manuscript. NP analyzed the data. RB designed and performed the simulations. NRLW provided the patient data. ATMW participated in the design of the study and revised the manuscript. JAvD designed the study, performed the statistical analysis and revised the manuscript. All authors read and approved the final manuscript.