This research project retrospectively included data acquired for clinical indications. Ethics approval was obtained from the Ethics in Research Committee of the Research Division, School of Medicine, Universidad Nacional Autónoma de México (UNAM) under the reference number SR1132014. This study was performed in accordance with the ethical standards of the 1964 Declaration of Helsinki and its later amendments. In addition, written informed consent was obtained from each child’s parents or guardians, permitting the use of data for both clinical and research purposes.

Whole-body [18F]FDG-PET/CT scans were acquired on a Siemens Biograph Truepoint 64 (Siemens Medical Systems, USA). Verification of the cross-calibration between the PET scanner and the dose calibrator was periodically performed by using a uniform phantom filled by a 18F solution as part of the quality control of the scanner. Each subject received a weight-adapted dose of 6.95 ± 1.87 MBq/kg of [18F]FDG intravenously after fasting for at least 6 h; patients < 2 years old fasted for approximately 4 h. Patients that required sedation for clinical imaging had nothing to eat or drink after midnight before the examination. In all cases, the patients fasted during the 1-h [18F]FDG uptake period. When required, anesthesia was administered during the scan acquisition, after the biodistribution period of the radiopharmaceutical.

Once the radiopharmaceutical was administered, patients remained in the preparation room and rested for approximately 70 min (range 60 to 79 min). A transmission CT scan (35 mAs, 120 kVs, slice thickness of 3 mm) was used for attenuation correction, followed by a PET emission scan for 2 min/bed position from the skull to the proximal third of the thigh. PET images were reconstructed with an ordered subset expectation maximization algorithm 2D (3–4 iterations, 15 subsets, Gaussian filter of 8–10 mm). The voxel size of the image was 4.07 × 4.07 × 3 mm3.

The imaging database of the PET/CT Unit of the School of Medicine at the UNAM, Research Division, was used as the source of analyzed data in the present research with whole-body [18F]FDG-PET/CT studies performed on pediatric patients younger than 18 years of age between January 2010 and February 2019 included. A senior M.D. specialist (NE, 18 years experience) reviewed the clinical information of each patient in order to select the sample of subjects for the present analysis.

The following cases were excluded from the study: patients with a history or presence of a neurological disease, central nervous system (CNS) involvement by neoplasms or infections, surgery or biopsy at brain level, diabetes, vasculitis, and prior or current treatment with chemotherapy or intrathecal radiotherapy. Additionally, examinations were excluded for incomplete clinical data, images that did not cover the entire brain, presence of artifacts, uptake times > 80 min, follow-up studies of the same patient, images with SUVmean outlier values based on a robust regression, and outlier removal test [40] with a Q = 5%.

Brain images for the selected patients were extracted from the whole-body PET/CT images using the PMOD v.3.806 software (PMOD Technologies LLC). Once the brain images were obtained, spatial normalization and co-registration procedures were performed through SPM12 v. 9.11 (Institute of Neurology, London, UK) [41]. Normalized images were segmented using the Hammers N30R83 atlas implemented in PMOD [42] to obtain the 3D volume of interest (VOI) of 83 cortical and subcortical brain structures and were scaled to their SUV using the injected dose and the weight of each patient. The average SUV (SUVmean) was obtained in each structure for all subjects. In addition, an average FDG uptake brain map (in terms of SUVmean) was created for each age group through all the subject’s data (freely available).

A linear regression analysis and standard deviation of the \(SU_\) values by age was performed to investigate the relationship of the [18F]FDG uptake across the pediatric age ranges. To evaluate the effect of age on [18F]FDG uptake in each region, an ANOVA test was performed with multiple Tukey–Kramer post hoc comparisons between \(SU_\) values and age groups. Multiple linear and non-linear models were proposed for each region to model the age effect on regional glycolytic brain metabolism. The models considered were linear, quadratic, cubic, exponential (\(y=_^\)), and power law (\(y=A*^+C^\)) at 95% prediction interval. The optimal model for each brain region was determined through the lowest Akaike information criterion (AIC), root mean square error (RMSE) value, and coefficient of determination (\(^\)) closest to unity.

The sex effect on the [18F]FDG brain uptake in each region was evaluated using a two-way ANOVA test and post hoc multiple comparisons (Bonferroni), comparing the \(SU_\) of males and females in each age group.

In addition, for each age group, the regional \(SU_\) normalized to pseudo-references was evaluated using previously reported regions [26, 28, 31]: cerebellum, brainstem, and the whole brain. Relative SUV (\(SU_\)) was defined as the ratio of the SUVmean to the \(SU_\) of the pseudo-reference.

The effect of laterality on the [18F]FDG uptake was evaluated from the values of the 37 paired brain structures in each age group, and the structures whose absolute percentage difference of SUVmean between left- and right-sided was more than 5% (\(\left|\Delta _\right|\)> 5%) were evaluated. First, a normality test (Kolmogorov Smirnov, n > 50 and Shapiro Wilk, n < 50) was performed for this evaluation. Subsequently, once the type of distribution of the data was known, the corresponding statistical tests were performed (paired t-test or Wilcoxon’s test).

In order to avoid systematic bias in the [18F]FDG uptake semi-quantification due to the large number of patients with Hodgkin’s lymphoma in this sample, an unpaired t-test with Welch’s correction was performed to compare the mean regional \(SU_\) values of this group of patients to other included patients.

All statistical analyses were carried out using GraphPad Prism v. 8.0.1 and R Studio v.1.4.1717 (R-Project.org) with p < 0.05 (α = 5%) considered significant for all inference testing.