Background
Neuroendocrine tumors (NETs) comprise a heterogeneous group of tumors originating from neuroendocrine cells and involving multiple organs, particularly the gastrointestinal tract and lungs [
1]. Marked expression of somatostatin receptor (SSTR) subtypes (SSTR1-SSTR5) is the main feature of NET cells, with overexpression of SSTR2 [
2]. PET imaging with
68 Ga‐DOTA‐Tyr3‐Thr8‐octreotide (
68 Ga‐DOTATATE) has higher affinity for SSTR-positive tissue (0.2 ± 0.04 nM) than other SSTR imaging agents [
3,
4].
Total-body PET detector crystals with a size of 2.76 × 2.76 × 18.0 mm
3 coupled to silicon photomultipliers (SIPMs) shows ultrahigh sensitivity and spatial resolution [
5]. Recently, a series of studies reported that low injection activity of
18F-FDG or a shorter scan time of total-body PET/CT imaging could still be feasible for good image quality [
6‐
9]. Additionally, influence factors of image quality were varied and complicated, such as PET equipment, activity of the tracer, acquisition time, and patient-specific photon attenuation, particularly for large body mass. In terms of conventional PET equipment, increasing injection activity and prolonging acquisition time to some extent might improve image quality. However, the probability of radiation-related injury and later-occurring effects, and the possibility of motion artifacts inevitably increased [
10]. As such, fixed acquisition time in clinical practice may result in significant differences in image quality between patients with varied body mass.
Regarding constant and acceptable image quality, we previously investigated the influence of patient size on image quality, and proposed a dose regimen based on body mass index (BMI, kg/m
2), demonstrating the feasibility of constant image quality for
18F-FDG total-body PET/CT [
11]. In addition, adjusting the duration time per bed based on scanner sensitivity and patient-specific attenuation might acquire uniform image noise or homogenous image [
12]. For
68 Ga-DOTATATE imaging, body mass (BM) was regarded as the strongest correlation with image quality [
13]. However, the injection activity and acquisition time of
68 Ga-DOTATATE vary in the literature, with reduced comparability of image quality between different studies. The regimen of a variable-acquisition time of
68 Ga-DOTATATE PET/CT for an acceptable and constant image has not been investigated thus far. Therefore, the aim of the present study was to propose a variable acquisition time regimen to balance the influence of scanners and BM and to obtain homogeneous image quality for patients with NETs.
Discussion
The 2017 European Association of Nuclear Medicine (EANM) procedure guidelines recommend that the administered activity of
68 Ga-DOTA-conjugated peptide ranges from 100 to 200 MBq, varying based on the PET system and patient size [
17]. Although that recommendation provides a reference for clinical practice, such a broad injection range creates certain challenges with respect to the comparability and repeatability of images. In previous research, we established a convenient patient-specific injection regimen of
18F-FDG for repeatable and constant imaging. Thus, considering the influence of the total-body PET system and body mass on image quality, the present study proposes a personal variable acquisition time regimen to gain constant image quality and avoid extending acquisition time based on objective and subjective image quality evaluation in the development of
68 Ga-DOTATATE PET/CT imaging. Next, the variable acquisition time regimen was validated and assessed for agreement with objective and subjective image quality evaluations in a validation cohort. The method we designed for a variable acquisition regimen might be instructive for other PET systems.
Image quality is commonly evaluated by the metric SNR or noise equivalent count rate (NECR). The SNR is calculated as the value of the square root of the product of system sensitivity, injected activity, and acquisition time [
13], whereas the NECR is calculated as the ratio of the square of the true events to the total of true events, random events, and scatter coincidences [
18]. A constant value of SNR and NECR can overcome the limitation of patient-specific attenuation to render uniform image quality. To achieve a constant image, acquisition time should vary from different patient size. For example, a typical patient with 100 kg should scan more time than a patient with 50 kg to achieve target SNR. One patient scanning by a related lower sensitivity detector should prolong acquiring time than by higher sensitive detector. In addition, to achieve target SNR, low activity of tracer should have longer scanning time than high activity. In developing cohort, the SNR
L increased with increasing acquisition time within 1 min (Fig.
2a). The SNR
norm showed a strong correlation with body mass (less than 150 kg) with a determination coefficient (R square) of 0.63, slightly lower than that in a previous study [
13]. We speculate that the range of body mass and the sample size might have contributed to this difference.
Based on the excellent intra- and inter-agreement agreement (all kappa > 0.85), the mean threshold SNR
L was 11.2 for acceptable image quality of
68 Ga-DOTATATE total-body PET/CT. Compared with
18F-FDG images we previously analyzed, the threshold SNR of 14.0 was slightly higher. The difference was consistent with a previous study in which the acceptable SNR
L was 6.2 for whole-body
68 Ga-DOTATATE PET/CT, but
18F-FDG studies have revealed a higher SNR of 10 [
19]. Compared with
18F-FDG with liver SUVmean of 2.6, the accumulation degree of
68 Ga-DOTATATE in liver was higher with liver SUVmean of 8.4 [
11]. We hypothesize that difference might be related to the percentage of biological distribution in the liver. In addition, the coefficient of variation, representing image noise, is recommended to be 15% as a reference maximum noise level for clinical
18F-FDG PET image interpretation [
20]. In the present study, both the CV of D2 in the development cohort and R4 in the validation cohort were less than 10%. A recent study [
9] showed that 5.5 times noise reduction of a 194-cm FOV PET compared to a 30-cm FOV digital PET with the same total examination time for scanning a 2-m-long phantom, and the noise reduction became 1.5 times when the same acquisition time per bed was performed.
In the validation cohort, all 90 lesions were detected in all acquisition time PET images of all subgroups. The higher percentage of G1 and G2 patients with marked SSTR expression (18/19 for the development cohort and 47/57 for the validation cohort) than G3 or NEC patients enrolled in this study might lead to bias in the results. Based on the proposed variable time regimen, the mean time was 2.99 ± 0.91 min, ranging from 2.18 to 6.35 min. Compared to the fixed acquisition protocol of 10 min, the mean acquisition time decreased by 70.1%, ranging from 36.5% to 78.2%. The variable acquisition time regimen based on a constant SNRL of 11.2 of total-body scan can eliminated the influence of BM, and provide more consistent image quality.
In this study, SNR was selected for evaluation of image quality because there was relatively homogeneous uptake of
68 Ga-DOTATATE by the liver, which was easily influenced by several circumstances. Previous studies have found less uptake of
68 Ga-DOTATATE by the liver, spleen, and thyroid after treatment initiation in patients with than without somatostatin analog treatment [
21]. One prospective study performed
68 Ga-DOTATATE imaging one day before and one day after injection of lanreotide, and no evidence of decreased uptake in the tumor, but a higher tumor-to-liver ratio, was obtained [
22]. To avoid the influence of treatment, our study excluded 6 patients with continuous treatment with octreotide and one patient with high-intensity focused ultrasound in the liver within 1 week. Additionally, the dynamic distribution between the development cohort regimen and low-activity regimen exhibited equivalent trends for the liver, pancreas, kidney, and spleen over time, which might eliminate the influence of different doses on biological distribution (Fig. S
2).
In addition, the image quality could also be improved using the reconstructed method of PSF and TOF. Previous study reported that the TOF could obtain more contract information than that without TOF information [
23]. Although the image reconstructed with PSF correction slowed the iterative convergence, it could provide a more uniform background and increased SNR than that without PSF correction. Previous study showed that the sufficient image quality could be acquired for low activity objects and a shorter acquisition time when the image constructed by PSF and TOF [
24]. In this study, the sufficient image quality might be contributed by the combination of ultra-high sensitivity of total body detector and the image reconstructed by TOF and PSF.
The findings of this study have to be considered in light of several limitations. First, the variable acquisition time regimen was established based on retrospective image reconstructions for 19 full-activity cases. The small sample size may result in confounding bias that may influence the reliability of the proposed regimen. Second, further investigation focusing on more organ SNRs, such as the spleen, kidney and additional metrics for image quality are needed. Third, we provide a method to realize personalized duration time; however, the acquisition time regimen was established only on total body PET. Thus, the proposed method should be referred to and rebuilt for other PET systems according to those characteristics.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.