Background
Currently, the role of 2-[
18F]fluoro-2-deoxy-D-glucose ([
18F]FDG) positron emission tomography/computed tomography (PET/CT) imaging is still expanding in the diagnosis and follow-up of paediatric oncologic, infectious and inflammatory diseases [
1,
2]. This expansion brings a point of concern, since both the CT and the [
18F]FDG expose patients to ionizing radiation, which can cause radiation-induced effects later in life [
3]. The risk of radiation-induced carcinogenesis is higher in children, as they have a longer post-radiation exposure life expectancy compared to adults [
3]. To reduce this risk, the radiation dose of paediatric PET/CT scans should be as low as reasonably achievable (ALARA) [
4] with acceptable image quality and within a reasonable acquisition time. Therefore, optimization and harmonization of paediatric PET/CT imaging protocols are essential [
5].
For performing paediatric nuclear medicine procedures, both the Society of Nuclear Medicine and Molecular Imaging (SNMMI) and European Association of Nuclear Medicine (EANM) recommend the EANM paediatric dosage card (version 5.7.2016) [
6,
7] or the 2016 North American Consensus guidelines (NACG) [
7,
8]. Both these guidelines have, however, several shortcomings as they are derived from adult-based protocols [
9‐
12], and both focus on radiation dose without taking image quality into account [
9‐
13]. Moreover, the EANM paediatric dosage card recommends even higher administered activities per kilogram than the adult 2015 EANM [
18F]FDG guidelines [
14]. The optimized adult guidelines recommend a quadratic dosage regimen based on body mass to obtain sufficient constant image quality, which was reported by a study of de Groot et al. [
15] using signal-to-noise ratio as a surrogate for image quality. This study investigated the relationship between a patient-dependent parameter, for example, body mass, BMI, LBM, fat mass (defined by body mass minus the lean body mass) and body mass per body/length and the administered [
18F]FDG. Another option is to use a linear dosage regimen < 75 kg and a quadratic dosage regimen > 75 kg to compensate for the lower signal-to-noise ratio due to excessive attenuation in heavier patients. Optimizing image quality in these patients was investigated by several studies [
16‐
20].
In contrast to adult dose regimens, only a few studies have been published on optimizing administered activities in children. Accorsi et al. [
10] found that weight was the best patient-dependent indicator for the administered activity necessary to obtain constant sufficient image quality. A pilot study of van Gent et al.[
21] focussed on body weight showed that a linear relationship between body weight and administered activity results in a constant image quality. Other studies have reduced the administered [
18F]FDG activity per kilogram of body weight based on simulations of PET low-dose scans by reduction of count rates [
11,
22,
23].
The Paediatric Dosage Harmonization Working Group and the 2020 paediatric guideline stated that more data are needed for evidence-based optimisation of the current guidelines [
7,
24]. This can be achieved by dedicated paediatric studies, based on the methods used for the adult guidelines, for different PET systems and reconstruction methods to provide the needed data for updating the paediatric guidelines. The aim of this study, therefore, is to investigate the relation between patient-dependent parameters and [
18F]FDG PET image quality and to propose a dedicated paediatric dose regimen that provides a constant and clinical sufficient image quality.
Discussion
In this study, the relation between patient-dependent parameters and [18F]FDG PET image quality was investigated in order to propose a dedicated paediatric dose regimen that aims for constant and sufficient image quality. To the best of our knowledge, this study is the first that investigates the relationship between paediatric patient-dependent parameters and [18F]FDG PET image quality based on SNR. This study has demonstrated that body weight is the parameter with the greatest effect on [18F]FDG image quality in children. Another important finding was that SNRnorm values were significant higher in children and correlated more strongly with age than in adults. Furthermore, this is the first study providing insight into the differences between paediatric and adult dosage regimens. This insight emphasizes the need for a dedicated paediatric dosage regimen, especially for young children.
Liver deoxyglucose metabolism in children was shown to be significantly lower than in adults. These results confirm the results obtained in a paediatric study by Yeung et al.[
30]. The increase in [
18F]FDG uptake during growth may be caused by age related changes in liver volume [
34] and function of hepatocytes [
35,
36]. Furthermore, the significant changes in body size, body composition [
34,
37] and blood volume during growth [
37] could also account for an increase in [
18F]FDG uptake. Not only body size and age affected SUV measurements but also differences in uptake period, plasma glucose, recovery coefficient and partial volume artefacts [
38]. The contribution of these factors may be limited in this study due to the use of standardized protocols.
Despite lower SUVbsa mean values and concomitant lower liver [
18F]FDG uptake, small children showed high SNRnorm values, which implies that less activity is needed to obtain sufficient image quality (Fig.
1a). This is probably caused by less attenuation and scatter due to the smaller body sizes. In contrast to SNRnorm, SNR has almost no correlation with age (Fig.
1b). This means that, despite the quite large spread (range 4.3–8.1), the currently used dosage regimen already provides a constant image quality throughout our patient population.
Body weight was the patient-dependent parameter with the highest coefficient of determination for both fit models for SNRnorm. The fit parameters (α = 2.23 and d = 0.46) of the preferred model are in line with the fit parameters (α = 3.2 and d = 0.52) obtained by van Gent et al. [
21] in a paediatric pilot study of 20 patients. The fit of the preferred model was not significantly different from the fits of BMI, BWH and BSA, even though body mass was used to define the optimal dose regimen as it is more practical than BMI, BWH and BSA. Although the selected model explains 81% of variability in SNRnorm among patients, the remaining 19% was not explained. This variability might be caused by differences in patient size, not covered by body weight. Unknown inhomogeneities within the liver VOIs might also cause variability of SNRnorm. Nevertheless, our results are consistent with earlier adult [
18F]FDG studies of de Groot et al.[
15] and Menezes et al.[
39]. These studies determined body weight with the highest coefficient of determination with SNRnorm with, respectively, R
2’s of 0.77 (OSEM 3D + PSF + TOF) and 0.86 (OSEM 3D + PSF) using comparable scanners to ours. Body weight was also identified as the best single predictor for image quality by Accorsi et al. [
10] at a comparable R
2 of 0.86 using a different camera, reconstruction method and the noise equivalent count rate density (NECRD) as measure of image quality. The NECRD is derived from the noise equivalent count rate (NECR) method [
19], which is considered to be more objectively related to SNR, since it is not affected by possible differences in liver metabolism and reconstruction methods [
40‐
42]. However, at that moment NECRD was not validated with a visual assessment, and therefore, the estimated sufficient image quality and proposed dosage regimen could be unreliable [
43].
In contrast to the studies mentioned above, this study directly compared the models of children and adults. It was not possible to find a universal dosing regimen linking paediatric and adult protocols (Fig.
3). This insight emphasizes the need for a dedicated paediatric dosage regimen, especially for young children. These results are in agreement with the EANM adult guidelines [
14], which recommends using a quadratic dosage regimen, especially for patients > 75 kg, as this compensates for the lower image quality caused by substantial attenuation when using a linear dosage regimen [
15]. In addition, they recommend that the linear dosage regimen is appropriate to use for patients < 75 kg, but this is not supported by data or references.
Our proposed paediatric [
18F]FDG dosage regimen (acquisition time 3 mbp) showed a reduction of the amount of administered activity and effective dose of 41% (NACG) and 63% (EANM) in a child of 30 kg (Fig.
4). Our findings broadly support the work of other studies in this area, which also found reductions of 40–50% [
10,
11,
22,
23], although they used other dosage regimens, scanners and reconstruction methods.
A limitation of our study is the absence of raw data due to the retrospective approach. Therefore, NECR analysis of the raw data to support the SNR data was not possible. The NECR data are more objective since they are not affected by differences in liver deoxyglucose metabolism and reconstruction parameters. An earlier adult study of Menezes et al. [
39] determined that both SNR and NECR showed the best correlation with body weight. The clinical SNR analysis could be replicated in a phantom study by de Groot et al. [
15] and this showed that the effects of liver glucose metabolism and reconstruction parameters are either rare or not as influential compared to attenuation effects. Another limitation is that we included only 85 adult patients against 102 children. However, we tried to create as much as possible equal patient distribution and adult patients below 55 kg and children above 95 kg are rare in our patient population. Furthermore, it should be pointed out that our study has been primary concerned with optimizing radiation dose from [
18F]FDG rather than from both [
18F]FDG and CT. The low-dose CT protocols of our scanners have already been optimized to reduce radiation dose (Child of 30 kg: DLP ± 75 mGy∙cm and 1.4 mSv [
44]) and to maintain sufficient image quality as with phantom studies in the past [
26‐
28].
The results of this study are only valid for Siemens Biograph mCT scanners with an OSEM 3D + PSF + TOF reconstruction. However, the objective method applied in this study can be used for further investigation for other scanners and reconstruction methods to obtain evidence-based recommendations for different types of scanners in the guidelines. Recently, three paediatric PET/magnetic resonance imaging (PET/MR) studies concerning the newest generation large field of view PET scanners with solid-state silicon photomultipliers and higher TOF resolution showed already that even more dose reduction is possible due to their higher sensitivity [
22,
45,
46], especially when MR replaces CT for localization and attenuation correction [
47]. Another promising tool is the new block sequential regularized expectation maximization (BSREM/Q.Clear) reconstruction software [
48‐
50] that generates images with higher image quality, which allows dose reduction. In the future, another step in dose reduction will be taken with the implementation of artificial intelligence (AI) technology in nuclear medicine imaging. AI technology offers a wide range of application opportunities for low-dose PET scans for example AI imaging [
51,
52], AI reconstruction [
53] and AI post-reconstruction image enhancement [
54].
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.