Reducing acquisition time in O-(2-¹⁸F-fluoroethyl)-L-tyrosine positron emission tomography (¹⁸F-FET PET) for malignant brain tumors: temporal stability of ordered subset expectation maximization (OSEM) and HYPER iterative algorithms and selection of reproducible radiomic features

Introduction

Malignant brain tumors include both primary and metastatic types. Gliomas make up about 80% of primary brain tumors, whereas metastatic brain tumors occur roughly 10 times more often than gliomas (1). Positron emission tomography (PET) is essential for cancer imaging, aiding initial diagnosis, surgical planning, and follow-up (2). Tumor uptake is evaluated visually and quantitatively, following European PET guidelines (3). Quantitative metabolic features from O-(2-¹⁸F-fluoroethyl)-L-tyrosine (¹⁸F-FET) PET, including standardized uptake value (SUV), tumor-to-brain ratio (TBR), and metabolic tumor volume (MTV) have been shown to offer important predictive and prognostic information (4-9). Moreover, radiomic features—such as first-order histogram metrics, gray-level run length matrix (GLRLM), and gray-level dependence matrix (GLDM)—provide numerous quantitative and noninvasive insights into tumor characteristics from medical images (10-12).

The clinical translation of quantitative imaging features critically depends on their repeatability and standardization. Among the factors affecting reproducibility in PET imaging, shorter acquisition time is an important clinical concern (13-16). Another important factor is the reconstruction algorithm, which is regularly updated to improve image quality (17-21). Although numerous studies have investigated these factors separately, comparisons of the temporal stability of different reconstruction algorithms under varying acquisition time remain limited. In clinical practice, acquisition times often vary due to patient movement or limited physical tolerance preventing completion of the full scan. Therefore, identifying features and algorithms that remain stable across different acquisition times is crucial to fully utilize clinical PET data.

According to the European Association of Nuclear Medicine (EANM), the European Association of Neuro-Oncology (EANO), and the Society of Nuclear Medicine and Molecular Imaging (SNMMI), clinical reading of static ¹⁸F-FET PET brain tumor images is recommended using summation images from 20 to 40 minutes post-injection (3,22). For patients with physical limitations, a practical approach is to reconstruct the data up to just before any movement, since PET scans in list mode allow images from any part of the data. However, shortened acquisition duration decreases signal-to-noise ratio (SNR) and image quality (23,24). Therefore, it is necessary to determine the minimum acquisition time for ¹⁸F-FET PET imaging of brain tumors that ensures diagnostic value.

The most widely used PET image reconstruction algorithm in clinics is ordered subset expectation maximization (OSEM), which improves image quality by iteratively updating pixels using maximum likelihood on divided subsets (25). Time-of-flight (TOF) systems have been introduced to reduce noise and enhance contrast recovery (26-28). A new Bayesian penalized likelihood reconstruction algorithm, HYPER Iterative with TOF (HIWT) from United Imaging Healthcare (Shanghai, China), iteratively seeks the maximum likelihood solution while incorporating noise control to achieve optimal convergence and effective noise suppression (29,30). The temporal stability of the quantitative features obtained by these reconstruction algorithms needs to be further investigated using clinical data.

In this study, we simultaneously examined the effects of acquisition duration and reconstruction algorithms on the repeatability of ¹⁸F-FET PET quantitative features in brain tumors, providing insights to optimize scan time and reconstruction methods, and to improve the identification of reliable imaging features across varying conditions. To our knowledge, this is the first integrated assessment of ¹⁸F-FET PET feature reproducibility.

Methods

Participants

A total of 76 patients with malignant brain tumors between 3 June 2019 and 9 January 2024 who underwent brain ¹⁸F-FET PET/magnetic resonance (MR) scanning followed by surgery were retrospectively enrolled in the study. The cases were excluded as follows: 7 patients without histological confirmation, 4 patients without ¹⁸F-FET avid findings, and 3 patients with significant motion artifacts on PET/MR images. Eventually, 62 patients were enrolled in the final cohort. Further detailed information about the inclusion and exclusion criteria and the participants is provided in Appendix 1. The study protocol was approved by the Ethics Committee of Xuanwu Hospital of Capital Medical University (No. 2023-044), and the requirement for individual consent for this retrospective analysis was waived. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

¹⁸F-FET PET imaging protocol

All ¹⁸F-FET PET acquisitions in this study were performed using a high-resolution 3 T hybrid PET/MR scanner (uPMR 790, United Imaging Healthcare) in three-dimensional (3D) mode for 20 minutes in list mode. PET image acquisition started approximately at 20 minutes post-injection. In this study, the starting acquisition time for the 62 patients was 20.17±0.43 minutes, with the variation primarily attributed to differences in patient setup time.

PET reconstruction

A total of 24 PET image reconstruction schemes were designed, comprising a combination of eight acquisition time windows (3, 5, 7, 10, 13, 15, 17, and 20 min) with three reconstruction algorithms [OSEM, OSEM with TOF (OWT), and HIWT]. Both OWT and HIWT utilized TOF information. All algorithms incorporated correction for the system’s point spread function (PSF). The reconstruction parameters were 20 subsets and 5 iterations for OSEM and OWT, and 0.28 regularization strength for HIWT. All PET images were reconstructed with a matrix size of 256×256, voxel dimensions of 1.2×1.2×1.4 mm, and a field of view (FOV) set to 30 cm. Attenuation correction was conducted using an ultrashort time-to-echo sequence, and the image data were further corrected for scatter and random coincidences. 3D T1-weighted (3D-T1) and T2-weighted fluid-attenuated inversion recovery (T2-FLAIR) magnetic resonance imaging (MRI) data without gadolinium contrast were also acquired concurrently with PET data. Images exhibiting artifacts that could compromise data quality were excluded from the analysis. MRI scanning parameters are detailed in Appendix 1.

Imaging analysis

Image quality was blindly evaluated by two experienced nuclear medicine physicians with 6 years of experience using a 5-point Likert scale. The evaluation was conducted in random order, with the physicians blinded to both the clinical data and the specific datasets to minimize bias. The scale ranged from 5 (excellent diagnostic quality) to 1 (non-diagnostic), with scores of 3, 4, and 5 deemed diagnostically valuable (24,31). Specifically, 5 indicated excellent diagnostic image quality, 4 indicated good quality, 3 indicated acceptable quality, 2 indicated suboptimal quality with limited additional clinical information, and 1 indicated non-diagnostic quality. Kappa statistic was used to evaluate the rating agreement between the two readers for the image quality assessment.

The SNR of the reconstructed PET images was determined as follows:

SNR=SUVmaxVOI−SUVmeanbgSUVsdbg[1]

where SUVmaxVOI represents the maximum tracer activity concentration within the lesion volume of interest (VOI), SUVmeanbg represents the average activity concentration in the background, and SUVsdbg represents the standard deviation of the background activity.

Quantitative common metabolic features and radiomics features were automatically calculated (as depicted in Appendix 1) for the VOI by the open-source software 3D Slicer (version 5.2.1; https://www.slicer.org/). In total, 8 common metabolic features [mean SUV (SUVmean), maximum SUV (SUVmax), mean TBR (TBRmean), maximum TBR (TBRmax), median SUV (SUVmedian), peak SUV (SUVpeak), MTV and total lesion uptake (TLU)] and 57 radiomics features [14 shape, 15 first-order, 11 GLDM, 8 GLRLM, 9 gray-level size zone matrix (GLSZM)] were extracted.

Statistics

In order to determine the impact of acquisition duration on ¹⁸F-FET PET quantitative features in brain tumors, relative percentage difference (RPD) represented by Δ was proposed. Δ in the SUVmean, SUVmax, TBRmean, TBRmax, MTV, and TLU were calculated as follows:

Δ=Ft−F20F20×100%[2]

where F_t represents the quantitative value at each duration of 3, 5, 7, 10, 13, 15, 17 minutes, and F₂₀ represents the quantitative value at the acquisition duration of 20 minutes.

The research flowchart for identifying screening features with excellent repeatability and stability is presented in Figure 1. First, PET image quality was rated using a Likert scale for qualitative assessment, and the qualified moments when all the image quality could be used for clinical diagnosis were selected. Then, feature repeatability between the three algorithms was evaluated by the intraclass correlation coefficient (ICC). Features with ICC >0.95 were considered as having excellent repeatability. The features with ICC >0.95 at all the qualified moments were included in the screening cohort for feature stability. Finally, a Friedman test was conducted for each algorithm to verify the stability across the qualified moments. Features with P>0.05 were considered of excellent stability. The ICC and Friedman statistical analysis were conducted in Python (Pingouin 0.5.3, https://pypi.org/project/pingouin/0.5.3/).

Figure 1 Research flowchart for identifying screening features with excellent repeatability and stability. ¹⁸F-FET, O-(2-¹⁸F-fluoroethyl)-L-tyrosine; HIWT, HYPER iterative with TOF; ICC, intraclass correlation coefficient; OSEM, ordered subset expectation maximization; OWT, OSEM with TOF; PET, positron emission tomography; TOF, time-of-flight.

Results

Effects of different acquisition times and reconstruction algorithms on PET image quality

The representative ¹⁸F-FET PET images of a 46-year-old woman with a left temporal IDH1-wild-type glioblastoma, reconstructed using the 24 designed schemes, clearly demonstrated the impact of the acquisition time and the algorithm on image quality (Figure 2A). The image quality improved along with the increase of acquisition duration (Figure 2B,2C). The weighted kappa of 0.81 indicated good agreement between readers on image quality. The algorithms with TOF achieved a higher image quality than the OSEM algorithm without TOF, reaching a Likert score of 3 sooner. Among the eight acquisition duration windows investigated in this work, all three algorithms reached a Likert score of 3 or higher at the three moments: 15, 17, and 20 minutes.

Figure 2 Representative PET images reconstructed by OSEM, OWT, and HIWT with acquisition duration from 3 to 20 minutes in a 46-year-old woman with a left temporal IDH1-wild-type glioblastoma (A) and the qualitative Likert scale rating results by two blinded experienced nuclear medicine physicians (B,C). Likert scale: 5, excellent diagnostic image quality; 4, good; 3, acceptable; 2, sub-optimal with limited clinical value; 1, non-diagnostic. HIWT, HYPER iterative with TOF; OSEM, ordered subset expectation maximization; OWT, OSEM with TOF; PET, positron emission tomography; TOF, time-of-flight.

In the first 10 minutes, HIWT demonstrated higher Likert scores attributable to its superior noise suppression capabilities, with its average SNR exceeding that of OWT by approximately 48.9% (see Appendix 1 for details). After 10 minutes, the SNR for OWT markedly improved by approximately 13.9%, whereas HIWT lost some boundary information, especially for lesions smaller than 10 mm in diameter, due to excessive smoothing. At the clinical standard acquisition duration of 20 minutes, OWT showed the best reconstruction performance.

Quantitative features comparison of PET images reconstructed by different schemes

SUVmean values exhibited minimal variation across acquisition durations for all three algorithms, with the RPD <1.1% after 15 minutes (Figure 3A and Table 1).

Figure 3 Change of SUVmean (A), SUVmax (B), TBRmean (C), TBRmax (D), MTV (E), and TLU (F) along with acquisition duration. HIWT, HYPER iterative with TOF; MTV, metabolic tumor volume; OSEM, ordered subset expectation maximization; OWT, OSEM with TOF; SUV, standardized uptake value; TBR, tumor-to-brain ratio; TLU, total lesion uptake; TOF, time-of-flight.

SUVmax values gradually decreased as the acquisition time increased for OSEM and OWT, with the highest SUVmax of 6.2 g/mL at 3 minutes due to low SNR and lowest SUVmax of 4.2 and 4.3 g/mL at 20 minutes
(Figure 3B and Table 1). When the acquisition time exceeded 15 minutes, SUVmax values tended to become usable, with an RPD of around 5%. For HIWT, SUVmax values fluctuated with an RPD <3.6% throughout, which indicated its superior noise filtering capability, enabling it to achieve stable SUVmax values in a shorter time compared to the previous two algorithms.

TBRmean values remained largely stable, with RPD below 1.2% after 15 minutes across all three algorithms (Figure 3C and Table 1).

For OSEM and OWT, TBRmax values gradually decreased as acquisition time increased, showing high RPD of 60.4% and 58.9% at 3 minutes. After 15 minutes, TBRmax values became usable, with RPDs around 5% (Figure 3D and Table 1). In contrast, HIWT exhibited fluctuating TBRmax values with RPDs of 9.1% at 3 minutes and 3.5% at 20 minutes, demonstrating its superior noise reduction capability compared to the other two algorithms.

MTV values gradually decreased as the acquisition time increased for OSEM and OWT, with the highest MTV of 37.9 and 38.9 mL at 3 minutes and lowest MTV of 35.8 and 36.1 mL at 20 minutes (Figure 3E and Table 1). MTV values tended to become usable with an RPD <5% when the acquisition time exceeded 15 minutes. For HIWT, MTV values tended to become usable with an RPD <0.9% when the acquisition time exceeded 15 minutes, which revealed the excellent stability of MTV in terms of acquisition duration after 15 minutes.

TLU values gradually decreased as the acquisition time increased for OSEM and OWT, with the highest TLU of 77.7 and 80.3 g at 3 minutes and lowest TLU of 70.2 and 72.3 g at 20 minutes (Figure 3F and Table 1). TLU values tended to become usable with an RPD of around 5% when the acquisition time exceeded 15 minutes. For HIWT, TLU values tended to become usable with an RPD <0.9% when the acquisition time was longer than 15 minutes.

Repeatability of PET common metabolic features in different reconstruction frameworks

Feature repeatability between the three algorithms was evaluated by the ICC. The ICC values of the eight common metabolic features over 3 to 20 minutes were all above 0.9, indicating good repeatability, except for SUVmax and TBRmax at 3 and 5 minutes, which showed ICCs of 0.726, 0.860, 0.810, and 0.890, respectively (Figure 4A).

Figure 4 ICC values of the eight common metabolic features with acquisition duration from 3 to 20 minutes (A), SUVmax of OSEM, OWT, and HIWT at 3 minutes (B) and 5 minutes (C), and Friedman test of reproducible common metabolic features between the three moments of 15, 17, and 20 minutes for each algorithm (D), with quantitative colors indicating the P value of the Friedman test. ***, P<0.001 (Friedman test); HIWT, HYPER iterative with TOF; ICC, intraclass correlation coefficient; MTV, metabolic tumor volume; OSEM, ordered subset expectation maximization; OWT, OSEM with TOF; SUV, standardized uptake value; TBR, tumor-to-brain ratio; TLU, total lesion uptake; TOF, time-of-flight.

The SUVmax at 3 and 5 minutes were compared between OSEM, OWT, and HIWT (Figure 4B,4C). Friedman test was performed in pairs between the three algorithms. The P value between OSEM and HIWT, as well as between OWT and HIWT, was less than 0.001, indicating a significant difference. Low SNR at 3 and 5 minutes and superior noise reduction capabilities of HIWT contributed to the difference.

With ICC greater than 0.95, all of the eight features with excellent repeatability were selected at 15, 17, and 20 minutes when the images could be used for clinical diagnosis. In order to verify the temporal stability of the features, a Friedman test was carried out between the three moments for each algorithm (Figure 4D). Only the P value of 0.595 of MTV for HIWT was higher than 0.05, which represented tolerable stability.

Repeatability of PET radiomics features in different reconstruction frameworks

Among the 57 radiomics features, 30 demonstrated an ICC greater than 0.95 at 15 and 17 minutes, indicating excellent repeatability. At the standard acquisition time of 20 minutes, 36 radiomics features exhibited an ICC greater than 0.95, suggesting that more features with high repeatability can be obtained (Figure 5A).

Figure 5 The distribution of ICC of the radiomics features at 15, 17, and 20 minutes (A), the number of radiomics features with P>0.05 for each reconstruction algorithm (B), and word cloud of the radiomics features with excellent stability (C), with more prominent features indicating higher frequency of occurrence across the three algorithms. HIWT, HYPER iterative with TOF; ICC, intraclass correlation coefficient; OSEM, ordered subset expectation maximization; OWT, OSEM with TOF; TOF, time-of-flight.

The 30 radiomics features with ICC values greater than 0.95 at 15, 17, and 20 minutes were selected. The temporal stability of these radiomics features was further verified using Friedman test between the three moments of 15, 17, and 20 minutes for each algorithm. There were 3, 2, and 18 radiomics features with P>0.05 for OSEM, OWT, and HIWT, respectively, indicating excellent stability (Figure 5B and Table 2). The common radiomics feature of the three algorithms with P>0.05 was firstorder_10Percentile, which had both excellent repeatability and stability (Figure 5C and Table 2).

Discussion

Although the quantitative features of FET PET imaging have demonstrated significant value in improving characterization and prognostication, the widespread application of quantitative analysis in clinical settings has been hindered by the lack of comprehensive studies on the reproducibility and standardization of these features. Previous research has primarily focused on the impact of either acquisition time or reconstruction algorithm on the reproducibility of quantitative features, with limited investigation into the temporal stability across different algorithms. In clinical practice, reducing image acquisition time remains a critical concern. Therefore, selecting algorithms that demonstrate greater temporal stability under varying acquisition time conditions is of significant value for optimizing PET imaging protocols. Features that are reproducible across algorithms and stable over time facilitate unified quantitative analysis of clinical data with varying acquisition times or reconstruction algorithms. This is especially important in scenarios where patients are unable to comply with the recommended guideline duration, thereby enabling optimal utilization of valuable clinical PET imaging data. To our knowledge, this is the first study to simultaneously investigate the impact of acquisition duration and reconstruction algorithms on ¹⁸F-FET PET image reconstruction results in brain tumors. Besides common metabolic features such as SUV, TBR, MTV, and TLU, radiomics features were also extracted to evaluate the reconstruction image.

Efforts to reduce PET acquisition time have been significant. Jannusch et al. demonstrated that ¹⁸F-FDG PET acquisition times in breast cancer can be reduced to below 10 minutes and lesion SUVmax group comparison between all PET acquisition times revealed no significant differences (32). Another study found that reducing frame duration in ¹⁸F-FET brain scans with PET/MR led to a significant increase in SUVmax, whereas SUV_A50, with lower deviation, was more suitable for comparing PET studies with varying counts (33). In this paper, the shortest acquisition time of 15 minutes, when the image quality meets the clinical diagnostic requirements, was identified. SUVmax slowly decreased after 15 minutes with an RPD of around 5% or less. MTV was identified with excellent temporal stability for the HIWT algorithm. Additionally, this work uniquely evaluated the temporal stability of radiomic parameters for each algorithm, an aspect not addressed in prior research. The HIWT algorithm identified 18 features with excellent temporal stability, whereas the OWT algorithm identified 2 and the OSEM algorithm identified 3.

Several studies have demonstrated that different reconstruction algorithms impact PET image quality and quantitative parameters. In particular, TOF technology offers significant advantages by enhancing lesion detection, improving quantification accuracy, and reducing image noise (19,20,26). Ayati et al. found that OSEM and Bayesian penalized likelihood algorithms produced highly correlated quantitative and volumetric parameters on gallium (⁶⁸Ga) PSMA-11 (⁶⁸Ga-PSMA-11) PET/computed tomography (CT), with SUVmean showing small differences and serving as a predictive biomarker (34). In this work, the temporal stability of the algorithms was uniquely evaluated. We identified 8 common metabolic features and 30 radiomics features with excellent repeatability across the three algorithms, and their temporal stability was further assessed across varying acquisition durations for each algorithm.

This study still had several limitations. The first limitation is the low number of patients, and further investigations with a larger dataset are warranted to validate the findings of this study. Second, various reconstruction settings—including the number of iterations and subsets, PSF inclusion, and post-filtering—have been reported to affect quantitative PET images (35,36). A preliminary investigation is provided in Appendix 1, but further study on their impact is still needed. Third, partial volume correction (PVC), a technique that addresses signal contamination between adjacent tissues due to the limited spatial resolution of PET imaging, is crucial for improving image quality and the accuracy of quantitative measurements (37,38). Image post-processing of PVC should be considered for the complexity of the brain tumor environment.

Conclusions

This study shows that ¹⁸F-FET PET imaging with an acquisition time of 15 minutes or longer provides sufficient diagnostic image quality for malignant brain tumors. We identified 8 common metabolic features and 30 radiomic features with excellent repeatability across three reconstruction algorithms (OSEM, OWT, and HIWT) at 15, 17, and 20 minutes. Furthermore, HIWT demonstrated superior temporal stability, detecting 18 stable radiomic features, compared to 3 for OSEM and 2 for OWT. These findings provide practical guidance for optimizing ¹⁸F-FET PET protocols by reducing acquisition time, choosing algorithms with better stability, and focusing on reliable features for tumor quantification.

Acknowledgments

None.

Footnote

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-649/dss

Funding: This work was supported by the National Key Research and Development Program of China (No. 2022YFC2406900 to J.L.); and the Huizhi Ascent Project of Xuanwu Hospital (No. HZ2021ZCLJ005 to J.L.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-649/coif). Y.H. is employed by the Central Research Institute, United Imaging Healthcare Group (a for-profit entity). This study received no financial or technical support from his employer or any commercial organization. J.L. reports the funding from the National Key Research and Development Program of China (No. 2022YFC2406900) and the Huizhi Ascent Project of Xuanwu Hospital (No. HZ2021ZCLJ005). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was approved by the Ethics Committee of Xuanwu Hospital of Capital Medical University (No. 2023-044), and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. DeAngelis LM. Global consequences of malignant CNS tumours: a call to action. Lancet Neurol 2019;18:324-5. [Crossref] [PubMed]
2. Salaün PY, Abgral R, Malard O, Querellou-Lefranc S, Quere G, Wartski M, et al. Good clinical practice recommendations for the use of PET/CT in oncology. Eur J Nucl Med Mol Imaging 2020;47:28-50. [Crossref] [PubMed]
3. Law I, Albert NL, Arbizu J, Boellaard R, Drzezga A, Galldiks N, la Fougère C, Langen KJ, Lopci E, Lowe V, McConathy J, Quick HH, Sattler B, Schuster DM, Tonn JC, Weller M. Joint EANM/EANO/RANO practice guidelines/SNMMI procedure standards for imaging of gliomas using PET with radiolabelled amino acids and [18F]FDG: version 1.0. Eur J Nucl Med Mol Imaging 2019;46:540-57.
4. Hua T, Zhou W, Zhou Z, Guan Y, Li M. Heterogeneous parameters based on (18)F-FET PET imaging can non-invasively predict tumor grade and isocitrate dehydrogenase gene 1 mutation in untreated gliomas. Quant Imaging Med Surg 2021;11:317-27. [Crossref] [PubMed]
5. Thiele F, Ehmer J, Piroth MD, Eble MJ, Coenen HH, Kaiser HJ, Schaefer WM, Buell U, Boy C. The quantification of dynamic FET PET imaging and correlation with the clinical outcome in patients with glioblastoma. Phys Med Biol 2009;54:5525-39. [Crossref] [PubMed]
6. Lohmann P, Stoffels G, Ceccon G, Rapp M, Sabel M, Filss CP, Kamp MA, Stegmayr C, Neumaier B, Shah NJ, Langen KJ, Galldiks N. Radiation injury vs. recurrent brain metastasis: combining textural feature radiomics analysis and standard parameters may increase (18)F-FET PET accuracy without dynamic scans. Eur Radiol 2017;27:2916-27. [Crossref] [PubMed]
7. Lindemann M, Oteiza A, Martin-Armas M, Guttormsen Y, Moldes-Anaya A, Berzaghi R, Bogsrud TV, Bach-Gansmo T, Sundset R, Kranz M. Glioblastoma PET/MRI: kinetic investigation of [18F]rhPSMA-7.3, [18F]FET and [18F]fluciclovine in an orthotopic mouse model of cancer. Eur J Nucl Med Mol Imaging 2023;50:1183-94.
8. Lohmeier J, Bohner G, Siebert E, Brenner W, Hamm B, Makowski MR. Quantitative biparametric analysis of hybrid (18)F-FET PET/MR-neuroimaging for differentiation between treatment response and recurrent glioma. Sci Rep 2019;9:14603. [Crossref] [PubMed]
9. Tanaka A, Sekine T, Ter Voert EEGW, Zeimpekis KG, Delso G, de Galiza Barbosa F, Warnock G, Kumita SI, Veit Haibach P, Huellner M. Reproducibility of Standardized Uptake Values Including Volume Metrics Between TOF-PET-MR and TOF-PET-CT. Front Med (Lausanne) 2022;9:796085. [Crossref] [PubMed]
10. Li X, Cheng Y, Han X, Cui B, Li J, Yang H, Xu G, Lin Q, Xiao X, Tang J, Lu J. Exploring the association of glioma tumor residuals from incongruent [18F]FET PET/MR imaging with tumor proliferation using a multiparametric MRI radiomics nomogram. Eur J Nucl Med Mol Imaging 2024;51:779-96.
11. Galldiks N, Langen KJ. Amino acid PET in neuro-oncology: applications in the clinic. Expert Rev Anticancer Ther 2017;17:395-7. [Crossref] [PubMed]
12. Ahrari S, Zaragori T, Zinsz A, Hossu G, Oster J, Allard B, et al. Clinical impact of an explainable machine learning with amino acid PET imaging: application to the diagnosis of aggressive glioma. Eur J Nucl Med Mol Imaging 2025;52:1989-2001. [Crossref] [PubMed]
13. Kruzhilov I, Kudin S, Vetoshkin L, Sokolova E, Kokh V. Whole-body PET image denoising for reduced acquisition time. Front Med (Lausanne) 2024;11:1415058. [Crossref] [PubMed]
14. Pilz J, Hehenwarter L, Zimmermann G, Rendl G, Schweighofer-Zwink G, Beheshti M, Pirich C. Feasibility of equivalent performance of 3D TOF [18F]-FDG PET/CT with reduced acquisition time using clinical and semiquantitative parameters. EJNMMI Res 2021;11:44.
15. Duan H, Baratto L, Hatami N, Liang T, Levin CS, Khalighi MM, Iagaru A. Reduced Acquisition Time per Bed Position for PET/MRI Using (68)Ga-RM2 or (68)Ga-PSMA-11 in Patients With Prostate Cancer: A Retrospective Analysis. AJR Am J Roentgenol 2022;218:333-40. [Crossref] [PubMed]
16. Niederkohr RD, Hayden SP, Hamill JJ, Jones JP, Schaefferkoetter JD, Chiu E. Reproducibility of FDG PET/CT image-based cancer staging and standardized uptake values with simulated reduction of injected FDG dose or acquisition time. Am J Nucl Med Mol Imaging 2021;11:428-42.
17. Hashimoto F, Onishi Y, Ote K, Tashima H, Reader AJ, Yamaya T. Deep learning-based PET image denoising and reconstruction: a review. Radiol Phys Technol 2024;17:24-46. [Crossref] [PubMed]
18. Yang FJ, Ai SY, Wu R, Lv Y, Xie HF, Dong Y, Meng QL, Wang F. Impact of total variation regularized expectation maximization reconstruction on the image quality of (68)Ga-PSMA PET: a phantom and patient study. Br J Radiol 2021;94:20201356. [Crossref] [PubMed]
19. Mehranian A, Kotasidis F, Zaidi H. Accelerated time-of-flight (TOF) PET image reconstruction using TOF bin subsetization and TOF weighting matrix pre-computation. Phys Med Biol 2016;61:1309-31. [Crossref] [PubMed]
20. Shekari M, Ghafarian P, Ahangari S, Ay MR. Quantification of the impact of TOF and PSF on PET images using the noise-matching concept: clinical and phantom study. Nucl Sci Tech 2017;28:167.
21. Liu L, Liu H, Xu S, Zhang S, Tao Y, Mok GSP, Chen Y. The Impact of Total Variation Regularized Expectation Maximization Reconstruction on (68)Ga-DOTA-TATE PET/CT Images in Patients With Neuroendocrine Tumor. Front Med (Lausanne) 2022;9:845806. [Crossref] [PubMed]
22. Vander Borght T, Asenbaum S, Bartenstein P, Halldin C, Kapucu O, Van Laere K, Varrone A, Tatsch KEuropean Association of Nuclear Medicine (EANM). EANM procedure guidelines for brain tumour imaging using labelled amino acid analogues. Eur J Nucl Med Mol Imaging 2006;33:1374-80. [Crossref] [PubMed]
23. Morimoto M, Kudomi N, Maeda Y, Kobata T, Oishi A, Matsumoto K, Monden T, Iwasaki T, Mitamura K, Norikane T, Yamamoto Y, Nishiyama Y. Effect of quantitative values on shortened acquisition duration in brain tumor (11)C-methionine PET/CT. EJNMMI Phys 2021;8:34. [Crossref] [PubMed]
24. Sonni I, Baratto L, Park S, Hatami N, Srinivas S, Davidzon G, Gambhir SS, Iagaru A. Initial experience with a SiPM-based PET/CT scanner: influence of acquisition time on image quality. EJNMMI Phys 2018;5:9. [Crossref] [PubMed]
25. Lantos J, Mittra ES, Levin CS, Iagaru A. Standard OSEM vs. regularized PET image reconstruction: qualitative and quantitative comparison using phantom data and various clinical radiopharmaceuticals. Am J Nucl Med Mol Imaging 2018;8:110-8.
26. Ter Voert EEGW, Veit-Haibach P, Ahn S, Wiesinger F, Khalighi MM, Levin CS, Iagaru AH, Zaharchuk G, Huellner M, Delso G. Clinical evaluation of TOF versus non-TOF on PET artifacts in simultaneous PET/MR: a dual centre experience. Eur J Nucl Med Mol Imaging 2017;44:1223-33. [Crossref] [PubMed]
27. Surti S. Update on time-of-flight PET imaging. J Nucl Med 2015;56:98-105. [Crossref] [PubMed]
28. Vandenberghe S, Mikhaylova E, D'Hoe E, Mollet P, Karp JS. Recent developments in time-of-flight PET. EJNMMI Phys 2016;3:3. [Crossref] [PubMed]
29. Xie H, Lyu Y. System and method for medical imaging. US patent 11,900,602 B2, 2024 Feb.
30. Sui X, Tan H, Yu H, Xiao J, Qi C, Cao Y, Chen S, Zhang Y, Hu P, Shi H. Exploration of the total-body PET/CT reconstruction protocol with ultra-low (18)F-FDG activity over a wide range of patient body mass indices. EJNMMI Phys 2022;9:17. [Crossref] [PubMed]
31. Lenga L, Czwikla R, Wichmann JL, Leithner D, Albrecht MH, Booz C, Arendt CT, Yel I, D'Angelo T, Vogl TJ, Martin SS. Dual-energy CT in patients with colorectal cancer: Improved assessment of hypoattenuating liver metastases using noise-optimized virtual monoenergetic imaging. Eur J Radiol 2018;106:184-91. [Crossref] [PubMed]
32. Jannusch K, Lindemann ME, Bruckmann NM, Morawitz J, Dietzel F, Pomykala KL, Herrmann K, Bittner AK, Hoffmann O, Mohrmann S, Umutlu L, Antoch G, Quick HH, Kirchner J. Towards a fast PET/MRI protocol for breast cancer imaging: maintaining diagnostic confidence while reducing PET and MRI acquisition times. Eur Radiol 2023;33:6179-88. [Crossref] [PubMed]
33. Wampl S, Rausch I, Traub-Weidinger T, Beyer T, Gröschl M, Cal-González J. Quantification accuracy of neuro-oncology PET data as a function of emission scan duration in PET/MR compared to PET/CT. Eur J Radiol 2017;95:257-64. [Crossref] [PubMed]
34. Ayati N, McIntosh L, Buteau J, Alipour R, Pudis M, Daw N, Jackson P, Hofman MS. Comparison of quantitative whole body PET parameters on [68Ga]Ga-PSMA-11 PET/CT using ordered Subset Expectation Maximization (OSEM) vs. bayesian penalized likelihood (BPL) reconstruction algorithms in men with metastatic castration-resistant prostate cancer. Cancer Imaging 2024;24:57.
35. Shiri I, Rahmim A, Ghaffarian P, Geramifar P, Abdollahi H, Bitarafan-Rajabi A. The impact of image reconstruction settings on 18F-FDG PET radiomic features: multi-scanner phantom and patient studies. Eur Radiol 2017;27:4498-509. [Crossref] [PubMed]
36. Hosseini SA, Shiri I, Hajianfar G, Bahadorzadeh B, Ghafarian P, Zaidi H, Ay MR. Synergistic impact of motion and acquisition/reconstruction parameters on (18) F-FDG PET radiomic features in non-small cell lung cancer: Phantom and clinical studies. Med Phys 2022;49:3783-96. [Crossref] [PubMed]
37. Mandeville JB, Efthimiou N, Weigand-Whittier J, Hardy E, Knudsen GM, Jørgensen LM, Chen YI. Partial volume correction of PET image data using geometric transfer matrices based on uniform B-splines. Phys Med Biol 2024;
38. Erlandsson K, Buvat I, Pretorius PH, Thomas BA, Hutton BF. A review of partial volume correction techniques for emission tomography and their applications in neurology, cardiology and oncology. Phys Med Biol 2012;57:R119-59. [Crossref] [PubMed]

(English Language Editor: J. Jones)