Visual Abstract
Abstract
Tumor metabolism is a hallmark of cancer, yet cellular heterogeneity within the tumor microenvironment presents a significant challenge, as bulk analysis masks the diverse metabolic profiles of individual cell populations. This complexity complicates our understanding of [18F]FDG uptake by distinct cell types in the tumor microenvironment. This study aims to investigate [18F]FDG uptake at the single-cell level in the lung of Kirsten rat sarcoma virus–driven cancer mouse models using the novel technique radio–flow cytometry (radioFlow). Methods: Two Kirsten rat sarcoma virus–driven lung cancer mouse models were injected with [18F]FDG for small-animal PET/CT and subsequent fluorescence-activated cell sorting of the lung. For radioFlow, the sorted cell fractions were then measured in a γ-counter and their radioactivity was normalized to the number of cells. Results: RadioFlow analysis of the lung tissue of both models showed a robust cell type–specific uptake pattern across experiments. Our key findings indicate that the [18F]FDG PET signal predominantly derives from immune cells (CD45+, F4/80−, 78.3% ± 6.6%; macrophage, 13.9% ± 4.3%), whereas tumor cells contributed only with 2.8% ± 1.0%, similar to the uptake of structural cells (CD45−; tumor cells, 5.0% ± 2.3%). Normalization showed that macrophages exhibited the highest glucose metabolism in both tumor models (57% ± 8%), followed by the remaining immune cells (27% ± 3%). Conclusion: These findings highlight the critical influence of immune cell metabolism on [18F]FDG imaging, emphasizing the need to account for immune contributions when interpreting [18F]FDG imaging in cancer.
Metabolic complexity as an underlying cause of tumor heterogeneity poses a major obstacle in treating cancer (1,2). The high metabolic activity of tumors is exploited to diagnose and monitor disease with modern molecular imaging technologies. In particular, tumor lesion glycolysis (TLG) as determined by [18F]FDG PET has emerged as an important diagnostic hallmark (3). [18F]FDG PET enables the detection of cancerous lesions throughout the entire body, and the derived TLG is a valuable tool predicting survival outcomes in various cancers, including lung cancer (3). However, despite its utility, [18F]FDG PET has limitations. First, the low specificity of [18F]FDG for tumor cells makes it difficult to differentiate between inflamed tissues, immune cells, and tumors. Second, the relatively poor spatial resolution (1–3 mm) of PET limits the appraisal of tumor heterogeneity and the contribution of the tumor microenvironment (TME). Although ex vivo autoradiography offers significantly a higher resolution, down to the micrometer scale, it still does not enable the visualization of single cells and is often prone to a poor signal-to-noise ratio. Lastly, cancer cells are thought to exhibit increased glucose uptake and rewire their metabolism, preferentially using glycolysis not only for energy production, even in the presence of oxygen, but also for generating metabolic building blocks for progression, including proliferation, survival, and maintenance (4,5). A recent study reported increased [18F]FDG uptake in giant cells, derived from radiation therapy on treatment response, which may lead to misinterpretation of therapy efficacy (6). Another study found elevated glucose metabolism in immune cells within a melanoma model, whereas tumor cells consumed more glutamine (7). These findings are also relevant to lung cancer, the second most prevalent cancer type in both sexes, which is associated with chronic inflammation of the lung parenchyma due to exposure to toxic cigarette smoke. Consequently, lung tumors are characterized by a strong infiltration of tumor-promoting immune cells, such as macrophages, into the TME. This infiltration concurrently inhibits the presence of antitumorigenic cells, such as cytotoxic T cells (CD8+), thereby promoting tumor progression and immune evasion (8,9). The TME is increasingly recognized as a critical contributor to tumor progression and heterogeneity, contributing significantly to therapeutic resistance. It is constantly remodeled through a complex array of interactions that include inflammatory cytokines and extracellular matrix components, as well as unevenly distributed oxygen and metabolites, such as glucose, lactate, and glutamine (10).
In the light of increasing numbers of immunotherapies in cancer treatment, there is a growing need to reassess the interpretation of quantitative imaging results by studying tumor and immunometabolism. For instance, whereas a high TLG in the therapy-naïve status is generally associated with a poor prognosis, the pseudoprogression, as detected by elevated [18F]FDG tumor uptake after immunotherapy, can be a confounding factor in assessing therapy response. An increase of [18F]FDG uptake in the tumor tissue on immunotherapy may therefore reflect the activation of the immune system and thus indicate a positive response to treatment (11–14).
Stable-isotope infusions with 13C-labeled substrates for metabolomic efflux studies revealed that tumors use a diverse range of nutrients to supply key metabolic pathways, including the tricarboxylic acid cycle. However, only a handful of studies have reported cell sorting in combination with radiotracers to delineate cell-type–specific contributions to metabolic flux in the tumor (7,15–19). Thus, little is known about the uptake of metabolic radiotracers, including [18F]FDG, by different cell types within the TME, which, however, may hold critical information about the course of a pathology.
To address these challenges, we require a detailed and dynamic appraisal of the metabolic status in cancer. We therefore resort to analyze the metabolic fingerprint in a preclinical lung cancer model using the combination of oncogenic Kirsten rat sarcoma virus (K-RAS) with an inflammatory burden imposed by a p53/A20 deficiency (antiinflammatory tumor necrosis factor alpha–induced protein 3) (20,21). We advance the analysis of tumor and immune cell metabolism with a combination of molecular imaging and an optimized radioactive fluorescence-activated cell sorting tool (radioFlow). In this study, we investigate the metabolic contribution of various cellular subtypes within the diseased lung by quantifying [18F]FDG uptake at the single-cell level.
MATERIALS AND METHODS
Animals
Animal husbandry and all animal procedures as described below were approved by the Austrian Federal Ministry of Education, Science and Research (GZ 66.009/0157-V/3b/2018, GZ 2021-0.406.848). Study procedures were conducted in accordance with the European Community’s Council Directive of September, 22, 2010 (2010/63/EU), and data reported in this study are in compliance with the Animal Research: Reporting of In Vivo Experiments guidelines 2.0 (22).
In Vivo Imaging and Quantification
Three weeks after tumor induction, mice (age, 10–12 wk) were fasted overnight (10–12 h) and anesthetized using 1.5%–2% isoflurane with oxygen (1.5–2 L/min), and the lateral tail vein was cannulated for subsequent radiotracer injection (30 ± 3 MBq of [18F]FDG for radioFlow analyses or 10 ± 4 MBq for imaging only in a total volume of 100 μL). [18F]FDG was prepared using a fully automated cassette-based synthesizer (FASTlab; GE HealthCare) within the clinical routine production at the Vienna General Hospital, Austria. Animals were awake during the 50-min radiotracer distribution, which was followed by a 10-min static PET scan. After the PET scan was completed, a 5-min CT scan was performed for anatomic guidance and attenuation correction. Images were acquired using a Siemens Inveon small-animal PET/SPECT/CT system (Siemens Preclinical Solutions). Multimodal (small-animal PET/CT) rigid-body image coregistration and biomedical image quantification were performed using the image analysis software PMOD 3.8 (PMOD Technologies).
Ex Vivo RadioFlow Analysis of Lung Tissue
Successively, cell suspensions of the tumor-bearing lungs were prepared by mechanical and enzymatical digestions (tumor dissociation kit, mouse [#130-096-730]; Miltenyi Biotec) for 40 min at 37°C with subsequent straining and lysis of erythrocytes. Cells were pretreated for Fc blocking (CD16/32, 1:200, #14-0161-85; eBiosciences) and labeled ex vivo with 1 of 3 fluorescent monoclonal antibody (mAb) cocktails (see supplemental materials; available at http://jnm.snmjournals.org). Cell sorting was conducted via a FACSAria fusion flow cytometer (Becton Dickinson), and fluorescence data were analyzed via FlowJo (FlowJo LLC). The radioactivity of the sorted cell fractions was measured in a γ-counter (Wizard2; PerkinElmer Inc.). Depending on the used antibody cocktail, the following cell subtypes were sorted: tumor cells: tdTom+; macrophages [MΦ] F4/80+ or MΦ2, CD206+; remaining immune cells: CD45+ F4/80−; structural cells: CD45− tdTom− (mAb cocktail 1); tumor cells: tdTom+; macrophages: F4/80+, helper T cells: CD4+; cytotoxic T cells: CD8+ (mAb cocktail 2); tumor cells: tdTom+; macrophages: F4/80+, B cells: CD19+; T cells: CD3+ (mAb cocktail 3). It is important to note that the respective sorting scheme significantly influences the relative distribution of the cells. Therefore, these 3 panels were chosen to depict the total cell composition of the lung tissue. Counts per minute were further normalized to the fraction’s cell number and expressed as percentages. The results of single-cell fractions are illustrated relative to the total counts of all cell fractions. The total cell number of the sorted cells was calculated within the cell-sorting system.
Further methods and technical details can be found in the supplemental materials. All data not included in the article or the supplemental materials are available on request.
RESULTS
In Vivo Imaging and Quantification of [18F]FDG in Tumor-Bearing and Wild-Type (WT) Lungs
In our study, we investigated potential differences in the uptake of [18F]FDG in tumor-bearing lungs of KPr (n = 3) and KPAr mice (n = 4), comparing the results to those from healthy control lungs (WT mice, n = 5). Representative small-animal PET/CT images of all 3 cohorts are depicted in Figure 1A.
In vivo and ex vivo analyses of healthy (WT) and tumor-bearing (KPr and KPAr) lungs. Representative [18F]FDG small-animal PET/CT images of thorax (upper left, axial; upper right, sagittal; lower left, coronal). Images 50–60 min post tracer administration. Regions corresponding to lungs, heart, and tumor lesions are highlighted (A). (B) Image-based quantification of lung volume, n ≥ 3; (C) ex vivo determination of lung weights, n ≥ 3; (D) [18F]FDG whole-lung uptake, n ≥ 3, (E) metabolic tumor volume (MTV), n ≥ 24; and (F) TLG (SUVmean* MTV), n ≥ 24. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Our analysis included CT-based measurements of lung volumes, which were significantly higher in KPAr mice (1.0 ± 0.16 cm3), compared with WT (0.45 ± 0.06 cm3, P = 0.0001) and KPr mice (0.60 ± 0.14 cm3, P = 0.0029) (Fig. 1B). Analysis of ex vivo lung weights revealed high intragroup variability and no significant differences; however, the lungs exhibited a trend similar to that of the in vivo volume results. Lung weights ranged from 0.13 to 0.15 g for WT mice (healthy lungs, n = 3), from 0.2 to 1 g for KPr mice (n = 12), and from 0.3 to 1.4 g for KPAr mice (n = 16) (Fig. 1C). These findings were supported by PET imaging and detailed quantification of [18F]FDG, which highlighted the distinct metabolic activity in tumor-bearing lungs. Small-animal PET/CT imaging showed a significantly higher tracer uptake in the whole lung for KPAr mice with an SUVmean of 1.74 ± 0.4 g/mL compared with 1.05 ± 0.2 g/mL for KPr mice (P = 0.0185) and 0.8 ± 0.2 g/mL for WT mice (P = 0.0012) (Fig. 1D). KPAr mice also had a higher number of metabolic tumor lesions, resulting in significantly higher metabolic tumor volume in their lungs (0.007 ± 0.001 cm3, n = 40) than in KPr lungs (0.003 ± 0.001 cm3, P = 0.0029, n = 24) (Fig. 1E). The lesion-to-liver ratio gave similar results for [18F]FDG uptake in KPr and KPAr lesions (Supplemental Fig. 1). The TLG of KPAr lesions was significantly increased with a mean TLG of 0.018 (range, 0.003–0.07) compared with a mean TLG of 0.004 for the KPr lesions (range, 0.0007–0.014) (Fig. 1F). Overall, the most prominent lung uptake of [18F]FDG was found in the KPAr mice, reflecting the increased tumor burden, which was also indicated by the larger lung volume.
Ex Vivo RadioFlow Assessment of [18F]FDG Uptake in Different Cell Populations Within the Lung
To further explore the in vivo [18F]FDG uptake in distinct cell populations, we established the technique radioFlow, which combines cell sorting with radioactivity detection. This approach enabled us to evaluate [18F]FDG uptake and glucose metabolism within the examined lung tissue and TME on a cellular level. The key results are illustrated in Figure 2; individual values are provided in Tables 1 and 2.
(A and D) Key results of radioFlow analyses. Sorted cell type distribution within lungs of WT, KPr, and KPAr mice, expressed in percent from all sorted cell populations. (B and E) In vivo [18F]FDG distribution in sorted cells from whole lung of KPr and KPAr mice, expressed as percentage of total radioactivity in percentage of counts per minute (%CPM). (C and F) In vivo [18F]FDG uptake normalized to sorted cell number from whole lung of KPr and KPAr mice (%CPM/105 cells); n ≥ 3. CPM measured in γ-counter. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Cocktail 1 Cell Sorting and RadioFlow Results
Cocktail 2 Cell Sorting and RadioFlow Results
To depict the total cellular composition, we sorted tumor cells, residual and structural cells, total MΦ or MΦ2 (CD206+)—MΦ1 (CD86+) could not be detected—as well as the remaining immune cells. Sorting analysis revealed that, in both disease models, the cell fraction with the highest cell number consisted of immune cells. In contrast, WT mice exhibited a lower proportion of immune cells, whereas structural cells dominated. Tumor cell counts were comparable for KPr and KPAr mice. MΦ2 were found to be lower in KPr mice than in KPAr mice (Fig. 2A). Analysis of the relative [18F]FDG distribution among the cell fractions showed the highest uptake in the immune cell fraction, with a significant difference between the 2 models (P = 0.0002). MΦ2 also showed a prominent [18F]FDG uptake and significant differences between KPr and KPAr mice (P = 0.0001). Interestingly, the [18F]FDG uptake of the tumor cell was in the same range as that of the structural cell and did not differ between the groups (Fig. 2B). We could not perform radioFlow analysis on the healthy lungs of the WT mice because the radioactivity in the sorted cell fractions was below the limit of quantification. However, it resulted in a decrease in the total uptake of immune cells while concurrently increasing the uptake in MΦ2. The general ratio between tumor and structural cells remained unaffected by the normalization (Fig. 2C). These data indicate that the [18F]FDG signal from PET imaging in these 2 cancer models originates predominantly from immune cells rather than from cancer cells, with MΦ2 identified as the immune cell population with the highest glucose metabolism.
To explore various T cell populations, we implemented mAb cocktail 2. In this panel, cytotoxic T cells and T helper cells were sorted along with tumor cells and MΦ. Since this sorting scheme more accurately represents the TME (excluding B cells from tumor-infiltrating lymph nodes and structural cells from healthy tissue or vascularization), tumor cells were found to be the highest fraction followed by MΦ. T cells exhibited the lowest numbers. Consistent with the results of the sorting panel using mAb cocktail 1, the number of MΦ was significantly higher in the KPAr mice (P = 0.0312), whereas the other cell types did not show significant differences in the cell distribution (Fig. 2D). Furthermore, the highest [18F]FDG uptake was found in the MΦ, also showing a significantly (P = 0.0004) higher uptake compared with the MΦ in KPr lungs and to tumor cells. The [18F]FDG uptake of the other cell types showed no significant differences (Fig. 2E). Normalization of [18F]FDG cell uptake revealed minimal differences in uptake compared with the nonnormalized data for the KPAr mice, yet MΦ still exhibited the highest uptake. Normalization of the KPr cell uptake shows that MΦ are the most glycolytic cell type (Fig. 2F).
The third sorting panel provided further differentiation of immune cells into MΦ, T cells, and B cells. Once again, MΦ exhibited the highest [18F]FDG uptake after normalization (Supplemental Fig. 2C; Supplemental Table 1). When comparing ex vivo and in vitro [18F]FDG uptake in the different cell types, we observed the same pattern (Supplemental Fig. 5A).
In summary, the radioFlow results show that the main [18F]FDG signal in both tumor models does not originate from tumor cells but from immune cells. The normalized results reveal that the highest glucose consumption was found in MΦ, more specifically in MΦ2.
Ex Vivo Autoradiography and Immunofluorescence Staining of Lung Tissue
Consistent with the in vivo imaging quantification, autoradiography results calculated as the percentage of injected dose per tissue slide show higher values for KPAr lungs (0.019 ± 0.003, n = 6) compared with KPr lungs (0.014 ± 0.002, n = 6) or WT lungs (0.006 ± 0.002, n = 5, P = 0.0077) (Fig. 3A). Qualitative comparison of the autoradiography signal with the tdTom signal indicates that the highest radioactive signal is localized around the tumor cores (Fig. 3B). Furthermore, comparison of immunofluorescence staining reveals that MΦ (F4/80+), although also infiltrating the tumor, are mainly situated in the surrounding regions rather than within the tumor core (Fig. 3C).
Ex vivo autoradiography and immunofluorescence of lung tissue. (A) Quantification of [18F]FDG uptake per lung slide (20 μm) used for autoradiography, n ≥ 5. Representative immunofluorescence (left; 20 μm) and vicinal autoradiography slide (right; 20 μm) of KPAr lung. (B) Immunofluorescence signal of tdTom+ in red. (C) Representative immunofluorescence staining of KPAr lung tissue slides (10 μm). Left: tdTom+ in red; middle: MΦ (F4/80+) in green; right: merged image of tumor cells and MΦ. %ID = percentage injected dose.
DISCUSSION
The field of preclinical imaging has traditionally focused primarily on the evaluation of newly developed radiotracers, often limited to xenograft models in immunosuppressed mice, which fail to reflect the complexity of the tumor in many aspects, including the TME (23). With the advent of immunotherapy and new animal models designed for this treatment (24), alongside the emergence of tracers for immunoimaging and the potential for image-guided therapies, there is growing recognition of the pivotal role immune cells play in shaping specific phenotypes, which can significantly vary in their metabolic status. In a recent study, a metabolic switch toward a more glycolytic phenotype was shown on radiation therapy and during the development of giant cells in tumor-bearing mice (6). Because of the low spatial resolution of PET, it remains unclear which cells cause the increased uptake on immune stimulation (11–14,25). All these aspects may potentially limit the prognostic value of [18F]FDG and TLG for monitoring therapy response. In general, there is limited knowledge about which cell types contribute to the [18F]FDG signal in any organ or tumor, underscoring the need to investigate [18F]FDG uptake at the single-cell level.
In both K-RAS–driven mouse models, which mirror a tumor mutation highly prevalent in human lung cancer, we observed that [18F]FDG is mainly consumed by immune cells rather than tumor cells. When comparing the percentage of counts per minute normalized to the cell number, MΦ exhibited the highest uptake of [18F]FDG. Further characterization revealed that MΦ belonged to the MΦ2 subpopulation, which is known to exhibit antiinflammatory and protumoral function (26). Overall, the KPAr mice showed a higher inflammatory burden (elevated interleukin-6 levels, unpublished data) and, consequently, an increased TLG and total lung [18F]FDG uptake, which could also be associated with the higher infiltration of MΦ2 and not necessarily with higher tumor burden. In detail, the radioactivity values for the normalized cell counts were similar in KPAr and KPr mice, indicating that MΦ2 consume comparable amounts of [18F]FDG in both models and are therefore metabolically equivalent. In vitro analyses showed no significant difference in [18F]FDG uptake between the different MΦ subtypes (Supplemental Fig. 5B), suggesting that anti- and protumoral MΦ cannot be distinguished by their glucose metabolism. We further validated our findings on tumor slices subjected to immunofluorescence staining to colocalize the tumor cells with low radioactive intensity, as determined by autoradiography, and confirmed significant MΦ infiltration. Two additional studies support our findings, both reporting elevated glucose metabolism in the immune cell compartment, with the latest study showing similar outcomes for MΦ in a colon carcinoma model (7,15).
No differences were observed when comparing [18F]FDG uptake in CD4 and CD8 T cells, regardless of their activation status (Supplemental Fig. 5C). However, comparing T cells to MΦ, we found a 10-times higher uptake in the MΦ (Supplemental Fig. 5B), suggesting that glucose metabolism might be one of the main metabolic pathways in MΦ, whereas T cells may also rely on other energy sources and biosynthetic intermediates (27,28).
Limitations of this study include the possibility that [18F]FDG may not represent all changes in the metabolic activity on cell activation. Furthermore, the underlying metabolism after its phosphorylation and trapping remains largely unknown. Radiometabolite analysis (29), metabolomics, or transcriptomics after cell sorting would be complementing techniques to study glucose utilization. Additionally, we sorted the entire diseased lungs rather than single tumor lesions. However, our approach and data illustration of cell counts and total counts per minute alongside normalized values provide a comprehensive representation of the PET images. To ensure accurate coverage of the entire cell composition and corresponding [18F]FDG uptake, we applied 3 different mAb cocktails covering the entire cell populations of the lung. By analyzing total tumor cell counts, we account for all lesions. Our autoradiography and immunofluorescence data suggest that the plethora of MΦ resides at the tumor lesion borders, which could introduce a systematic error in single-lesion analysis. Unlike previous studies (7,15,16,30), we also evaluated technical aspects including limit of quantification, Cherenkov effects, and radiation dose (supplemental materials) to ensure robust analysis.
In general, we presume that the metabolic fingerprint within the TME may differ between tumors (intertumoral heterogeneity) as well as different stages of tumor progression. Additionally, metabolite competition in the TME has been reported as a critical factor influencing tumor growth (26,31).
Our experiments offer a highly relevant and novel approach to investigating the metabolic dependencies between tumor and stromal cells in vivo. The variable metabolic tumor volume and the radioFlow data might indicate metabolic and growth heterogeneity. To account for heterogeneity between lesions, image-guided lesion radioFlow analyses should be performed. However, longer dissection time and lower tissue mass necessitate significantly higher initial activities to achieve detection above the limit. Our findings suggest that infiltrating immune cells might also affect imaging in other tumor entities, especially in more aggressive forms. RadioFlow holds significant potential for elucidating metabolic changes within the TME across various tumor types and stages.
CONCLUSION
We offer a distinct single-cell perspective on [18F]FDG uptake by developing and optimizing a combination of molecular imaging and radioactive fluorescence-activated cell sorting (radioFlow) using a robust and reliable method for qualitatively and quantitatively assessing radiotracer uptake in specific tissue cell types after in vivo PET imaging. Hence, we illustrated the potential of the radioFlow method in combination with other molecular technique (metabolomics, RNAseq, etc.) to characterize the glycolytic fingerprint in vivo of major cell types in a tumor using 2 K-RAS–driven lung cancer mouse models. Our findings reveal that immune cells, rather than tumor cells, are the primary drivers of [18F]FDG uptake in the examined tumor models. This underscores the importance of accounting for immune cell metabolism when interpreting [18F]FDG imaging in relation to disease stage and therapy. We emphasize that there is no question regarding the value of [18F]FDG imaging, and we believe its impact could be further enhanced with a deeper understanding of the metabolism within the TME.
DISCLOSURE
Emilio Casanova and Herwig Moll were supported by the Austrian Science Fund (FWF) [10.55776/P32900, 10.55776/P33430, 10.55776/P36728, and 10.55776/DOC59 to Emilio Casanova; PAT 5733623 to Herwig Moll] and the grant “City of Vienna Fund for Innovative, Interdisciplinary Cancer Research”. No other potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: Which cells preferentially uptake and metabolize [18F]FDG in the TME?
PERTINENT FINDINGS: This preclinical imaging study and ex vivo single-cell analysis revealed that [18F]FDG uptake in 2 K-RAS lung cancer models is driven by immune cells and particularly MΦ2.
IMPLICATIONS FOR PATIENT CARE: These findings and future work may impact the understanding of [18F]FDG metabolism and, therefore, its prognostic value, especially during immunotherapy, with the advances in scanner technologies and using radiomics analyses.
ACKNOWLEDGMENTS
We gratefully thank Victoria Weissenböck, Anna Zacher, Johann Stanek, Lara Breyer, Günther Hofbauer, Daniela Laimer-Gruber, Paulina Susek, Theresa Patsch, Ralitsa Zhivova, Mahshid Eslami, and [18F]FDG producers for their technical support, and Maximilian Krisch and Marius Ozenil for student trainings. Thomas Wanek’s PIL management was indispensable and greatly appreciated. We express our gratitude to Thomas Weichhart for the insightful discussions and for sharing relevant literature. Additionally, we thank Florentina Porsch and Christoph Binder for their initial support of the flow cytometry analyses. Andreas Walter (Aalen University) and Anna Obenauf (Research Institute Molecular Pathology, IMP) are acknowledged for their melanoma model, which was used to establish radioFlow. The graphical abstract was created with BioRender.com.
Footnotes
Published online Jan. 16, 2025.
- © 2025 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication September 15, 2024.
- Accepted for publication December 16, 2024.