Introduction
Preclinical positron emission tomography (PET) imaging has become a crucial tool for the development and evaluation of radiolabeled tracers and therapeutic drugs and can facilitate faster translation from bench to bedside [
1,
2]. Preclinical PET cameras have been in constant evolution to offer improved spatial and temporal resolution and sensitivity. Nowadays, the advantages of functional and anatomical imaging are combined within hybrid cameras where computerized tomography (CT), magnetic resonance (MR) or fluorescent imaging technologies are added to the PET scanners [
3,
4]. Compared to CT, MRI offers better soft tissue resolution and no radiation exposure. However, preclinical and clinical development of PET/MR scanners came with their own specific technical hurdles due to the strong magnetic field that might affect PET functionality, making the combination of the PET with MRI components more challenging than with CT. Furthermore, as MRI does not provide information on electron density, the attenuation coefficient map is more difficult to obtain and can result in lower imaging resolution [
2].
To evaluate the potential and in vivo behavior of radiotracers, ex vivo biodistribution is generally considered the method of choice. In such studies, the radiotracer is administered to animals, and at predetermined time points tissues of interest are collected, weighed, and counted for radioactivity. Percentage of injected dose (or percentage of activity) per gram of tissue (%ID/g or %IA/g) is then calculated providing information about the biodistribution and in vivo kinetics of the PET tracer. A large number of animals is usually used for a limited number of time points, which is not in line with the ethical aim to reduce animal use. Unlike ex vivo biodistribution, preclinical PET imaging is offering the possibility for longitudinal studies, where the same animals can be used as their own control at multiple and early time points. This fully complies with current ethics on animal experiments and the 3Rs guidelines (Reduction, Replacement, Refinement) [
5‐
8]. Furthermore, PET imaging allows quantitative and visual whole body assessment of tracer uptake giving insights on the heterogeneity of its distribution often not possible with ex vivo biodistribution. Despite the increased availability of preclinical PET cameras in research centers, PET imaging is mostly used as a qualitative visualization tool and not as a replacement for ex vivo quantification of biodistribution [
9]. The review of Kuntner and Stout [
10] summarized the crucial parameters that are needed to obtain reliable and reproducible quantitative PET data in small animals studies: i. the camera itself (i.e., size, type, and number of detectors), ii. the physical properties of the PET isotope used (i.e., positron range), and iii. the animal models and animal handling. Moreover, preclinical PET studies still lack standardized protocols that lead to unreliable data making inter-study comparison more challenging.
In recent publications, the reliability and reproducibility of preclinical data acquired using phantoms and various cameras in different imaging centers were evaluated. The key factors that resulted in variability were, apart from the type of camera used, the differences in animal handling and image acquisition protocols and analysis, and variability between users [
11‐
13].
PET imaging is rarely used for the quantification of radiotracer biodistribution and data that allow comparison between in vivo and ex vivo uptake are very scarce. The influence of the partial volume effect (PVE) is often mentioned as a cause of quantification inaccuracy. PVE results from the limited spatial resolution of preclinical cameras which impairs accurate measurement of activity concentrations in small regions surrounded by other organs or background activity. Consequently, an underestimation of tracer uptake, especially in, e.g., small tumors, orthotopic or metastatic models, organs or brain sub-regions is observed [
10].
In this study, we evaluated the performance of a Mediso nanoScan PET/CT and PET/MR scanner for tracer quantification in subcutaneous tumors and selected organs by direct comparison of PET imaging with ex vivo biodistribution. We evaluated the critical parameters involved in reliable PET quantification in vitro as well as in vivo. Firstly, phantoms filled with the most commonly used PET radionuclides 11C, 68Ga, 18F, and 89Zr were used in order to evaluate the performance of the cameras with respect to linearity, recovery coefficient, PVE and spill-over effect. Secondly, breast cancer tumor-bearing mice were used for quantification of [18F]FDG and [89Zr]Zr-DFO-NCS-trastuzumab uptake in tumors and selected organs (brain, kidney and liver) by PET imaging as well as by ex vivo biodistribution. Congruency between PET quantification and ex vivo biodistribution was evaluated using Bland–Altman plots.
Discussion
In this study, we evaluated the performance of preclinical PET imaging quantification versus ex vivo biodistribution to assess whether and under which conditions PET imaging is able to replace ex vivo biodistribution. This is of particular interest in longitudinal studies with xenograft-bearing mice where accurate tumor uptake quantification via PET imaging would drastically reduce the number of animals used for tracer evaluation. For this purpose, we compared the uptake derived from the PET/CT and the PET/MR images with ex vivo biodistribution in N87 tumor-bearing mice with the most commonly used PET radionuclide, 18F (as in [18F]FDG). To further compare both quantification methods, we evaluated in the same animal model, mice injected with a long-lived radionuclide matching the biological half-life time of monoclonal antibodies: 89Zr as in [89Zr]Zr-DFO-NCS-trastuzumab, a well-known HER2 targeting monoclonal antibody.
Before performing in vivo studies, we evaluated the nanoScan PET/CT and PET/MR using the preclinical NEMA NU 4-2008 phantom in a standard approach and obtained results in line with others (Figs.
1,
2 and Additional file
1: S1–S2) [
12,
14,
16,
18,
23].
18F outperformed all radionuclides with the highest recovery coefficient (80%), while
11C and
89Zr had a lower but comparable recovery and
68Ga the lowest (54%) as could be expected from physical properties,
ß+ energies and range in water, of those radionuclides [
24]. In the case of
18F and
89Zr, further explored in the in vivo studies, there are many reasons for lower RCs for
89Zr in comparison with
18F including the following: (i)
89Zr possesses a lower signal to noise ratio compared to
18F which can be due to lower injected activities and the fact that
89Zr has a lower positron branching fraction (23%) compared to
18F (97%), (ii) the positron range of
89Zr is larger than
18F (respective mean range in water 1.2 and 0.6 mm) leading to a lower effective spatial resolution and thus lower RCs [
16].
The unique aspect of the phantom studies apart from evaluating both a PET/CT and PET/MR from the same provider with four radionuclides (
11C,
68Ga,
18F and
89Zr) was that we evaluated the RCs in the five cylinders of the phantom based on ROIs sizes matching the real contours of the cylinders (Fig.
2) and not only using ROI sizes twice their actual size as is performed in the standard NEMA quality control of the cameras (as represented in Additional file
1: Figure S1). This explains why RCs in Additional file
1: Figure S1 were higher than in Fig.
2. Our method was intended to reflect the approach used later in vivo where tumors were also delineated based on the exact contours of the organ.
The number of organs that can accurately be segmented and thus quantified with PET imaging is more limited than with ex vivo biodistribution. Even though subcutaneous tumor xenograft can usually be very well delineated, they may present various shapes and few studies have been comparing tumor uptake derived from PET imaging directly to ex vivo biodistribution. As tumors were the main organ of interest, they were delineated based on CT or MR to define ROIs and this method proved indeed to be reliable.
To obtain ROIs, we defined a systematic approach without the use of (semi-) quantifying tools for all organs to be able to compare with ex vivo biodistribution. The eventual choice of predefined quantifying tools should always be well justified as they can lead to large differences in assessed uptake that can limit inter-study comparisons. In our systematic approach, in comparison with ex vivo biodistribution, such tools were thus not used. Available tools for PET quantification and their influence on assessed uptake in organs have been nicely explored in the comprehensive paper from Mannheim et al. [
11]. As an example, they compared [
18F]FDG uptake in left ventricles by two different persons using fixed but different thresholding strategies, and this led to different uptake values, proving that this was not a reliable analysis method.
The correlation between tumor volume derived from organ delineation based on CT or MR and tumor weight assessed during biodistribution was high (
R2 > 0.8 for all possible combinations, (tracer and cameras)) showing that the two methods of assessing the tumor weight are comparable and minimally affecting the %IA/g values. PET assessed uptake in tumors (in %IA/g) was consistently lower than ex vivo determined uptake with comparable biases for
18F and
89Zr for both cameras (Table
1). Underestimation of the PET in comparison with ex vivo biodistribution has been previously reported, using different systems and analysis methods and was attributed mostly to the inherent limited resolution of the systems, variation between ROI delineation methods and PVE [
25,
26]. Tatsumi et al. [
27] observed a recovery ratio of 80 ± 20% between PET/CT and ex vivo assessed activity concentrations (in Bq/mL) in tumor-bearing rats injected with [
18F]FDG. In brain and kidney, they reported a lower recovery of 40 and 60%, respectively, and they attributed it to the smaller organ sizes compared with tumors.
In our study, the bias for the ratio PET/ex vivo in tumors was slightly lower for
89Zr compared to
18F which could be explained by the lower recovery coefficient of
89Zr compared to
18F caused by the difference in physical properties of both radionuclides (see before). However, 95% limits of agreement were better in the
89Zr study which could be due to the relatively high uptake of the
89Zr tracer in N87 tumors (~ 15–25%IA/g) when compared to [
18F]FDG (~ 2–4%IA/g). Only a tendency for a lower PET/ex vivo uptake ratio with smaller tumors was observed in the
18F study with the PET/CT (Additional file
1: Figure S3A), and thus PVE was not identified to influence results in tumors.
An excellent bias for the PET/ex vivo ratios was obtained for the brain with the animals injected with [
18F]-FDG, and a very narrow 95% agreement interval for the PET/CT (Table
1). These results suggest that the brain is a suitable organ for quantitative imaging of
18F tracers like [
18F]FDG. However, this was not the case for
89Zr-radiolabeled-mAb, trastuzumab, where a high discrepancy between the two quantification methods was observed. This is likely attributed to the fact that the level of brain penetration of
89Zr-labeled mAbs is low (close to ~ 0%IA/g) with a lack of specific targeting, resulting in low organ to background ratios and relatively large biases of the ratios due to the small values. This aspect should be taken in consideration when evaluating antibodies for brain targeting [
28].
For PET quantification, the lack of standardized protocols and the variability between animal handling, analysis and users has been reported recently [
11]. In addition, our results suggest that quantification reliability also depends on the particular organ that is analyzed and how users define, select, and draw ROIs. For example, kidney and tumor were in our study always drawn based on anatomical images (CT or MRI) and not based on the PET signal, which in our opinion increases variability due to visual artefacts and scaling bias of the PET image. The soft tissue resolution of MR is better than CT allowing better delineation of organs that could influence the quantitative data derived from PET images. The overall comparison of tracer quantification by both cameras with ex vivo biodistribution especially in tumors gave very similar results suggesting that tumor delineation on CT is equally good as MR. In addition, the difference in attenuation and scatter correction algorithms in both scanners did not significantly affect the quality of the data [
14].
To the best of our knowledge no preclinical study compared PET/CT and PET/MR scanners with ex vivo biodistribution for various radionuclides. This is most probably due to the fact that preclinical PET/MR systems were introduced recently in comparison with single PET and PET/CT scanners. Most studies in the past have estimated the outcome of in vivo studies solely based on the performance of the PET in phantoms which does not always predict the performance in animals [
14,
18,
29‐
31].
It is also important to note that while ex vivo biodistribution is considered the gold standard for quantifying tracer biodistribution, in practice this technique might appear not perfectly standardized leading to different results between research centers. Technical details on how biodistribution is performed are often missing in publications, while this is important for standardized sampling. Critical details to be reported: is blood removed from tissues, is the organ collected entirely, and are fat, skin or other surrounding tissues properly removed? Next to this also the counting of radioactivity should be standardized, e.g., taking care of geometric effects and counting saturation (dead time).
Our manuscript described real-life issues related to uptake quantification and offers solutions on how to perform preclinical studies in a systematic way. Furthermore, this study provides a direct comparison of ex vivo biodistribution with PET/CT and PET/MR cameras from the same provider. PET appears to be a reliable quantification method to assess tumor uptake in xenografted mice, for which on top of the aforementioned parameters, an evaluation per organ and per radiotracer is necessary for future preclinical studies and comparison between them.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.