*Result*: Low-count whole-body PET denoising with deep learning in a multicenter, multi-tracer and externally validated study.
Li W, Fang L, Li J. Exposure doses to technologists working in 7 PET/CT departments’. Dose Response. 2020;18(3):1559325820938288. https://doi.org/10.1177/1559325820938288 . (PMID: 10.1177/1559325820938288326949617350403)
McCann A, Vintró LL, Cournane S, Lucey J. Assessment of occupational exposure from shielded and unshielded syringes for clinically relevant positron emission tomography (PET) isotopes-a Monte Carlo approach using EGSnrc. J Radiol Prot. 2021. https://doi.org/10.1088/1361-6498/ac0df5 . (PMID: 10.1088/1361-6498/ac0df534161938)
Farkas J, Martin M, Nielsen C, Jennings SG. The effects on technologist occupational exposure in PET/CT departments when working with students at various levels of supervision. J Nucl Med Technol. 2020;48(3):214–7. https://doi.org/10.2967/jnmt.119.241398 . (PMID: 10.2967/jnmt.119.24139832312851)
Wen JC, Sai V, Straatsma BR, McCannel TA. Radiation-related cancer risk associated with surveillance imaging for metastasis from choroidal melanoma. JAMA Ophthalmol. 2013;131(1):56–61. https://doi.org/10.1001/JAMAOPHTHALMOL.2013.564 . (PMID: 10.1001/JAMAOPHTHALMOL.2013.56423307209)
Kessara A, Buyukcizmeci N, Gedik GK. Cancer risk estimation for patients undergoing whole-body PET/CT scans. Radiat Prot Dosimetry. 2023;199(6):509–18. https://doi.org/10.1093/RPD/NCAD040 . (PMID: 10.1093/RPD/NCAD04036856709)
Subramanian K, et al. Complex implementation factors demonstrated when evaluating cost-effectiveness and monitoring racial disparities associated with [18F] DCFPyL PET/CT in prostate cancer men. Sci Rep. 2023;13(1):8321. (PMID: 10.1038/s41598-023-35567-w3722139710205741)
Mason C, Gimblet GR, Lapi SE, Lewis JS. Novel tracers and radionuclides in PET imaging. Radiol Clin North Am. 2021;59(5):887. https://doi.org/10.1016/J.RCL.2021.05.012 . (PMID: 10.1016/J.RCL.2021.05.012343929259116257)
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(1):24–46. https://doi.org/10.1007/s12194-024-00780-3 . (PMID: 10.1007/s12194-024-00780-33831956310902118)
Balaji V, Song TA, Malekzadeh M, Heidari P, Dutta J. Artificial intelligence for PET and SPECT image enhancement. J Nucl Med. 2024;65(1):4–12. https://doi.org/10.2967/jnumed.122.265000 . (PMID: 10.2967/jnumed.122.2650003794538410755520)
Bousse A, et al. A review on low-dose emission tomography post-reconstruction denoising with neural network approaches. IEEE Trans Radiat Plasma Med Sci. 2024;8(4):333–47. https://doi.org/10.1109/TRPMS.2023.3349194 . (PMID: 10.1109/TRPMS.2023.33491943942980511486494)
Seith F, et al. Simulation of tracer dose reduction in 18F-FDG PET/MRI: effects on oncologic reading, image quality, and artifacts. J Nucl Med. 2017;58(10):1699–705. https://doi.org/10.2967/jnumed.116.184440 . (PMID: 10.2967/jnumed.116.18444028360205)
Gatidis S, Seith F, Schäfer JF, Christian la Fougère MD, Nikolaou K, Schwenzer NF. Towards tracer dose reduction in PET studies: simulation of dose reduction by retrospective randomized undersampling of list-mode data. Hell J Nucl Med. 2016;19(1):15–8. (PMID: 26929936)
Schaefferkoetter J, et al. Convolutional neural networks for improving image quality with noisy PET data. EJNMMI Res. 2020;10(1):105. https://doi.org/10.1186/S13550-020-00695-1 . (PMID: 10.1186/S13550-020-00695-1329556697505915)
Xu J, Gong E, Pauly J, Zaharchuk G. ‘200x low-dose PET reconstruction using deep learning’. arXiv preprint arXiv:1712.04119 , 2017.
Jang SI, et al. Spach transformer: spatial and channel-wise transformer based on local and global self-attentions for PET image denoising. IEEE Trans Med Imaging. 2023;43(6):2036–49. https://doi.org/10.1109/TMI.2023.3336237 . (PMID: 10.1109/TMI.2023.3336237)
Pan S, et al. Full-dose whole-body PET synthesis from low-dose PET using high-efficiency denoising diffusion probabilistic model: PET consistency model. Med Phys. 2024;51(8):5468–78. https://doi.org/10.1002/MP.17068 . (PMID: 10.1002/MP.170683858851211321936)
Liu Z, et al. Swin transformer: Hierarchical vision transformer using shifted windows. In: Proceedings of the IEEE International Conference on Computer Vision. 2021. pp. 9992–10002. https://doi.org/10.1109/ICCV48922.2021.00986 .
Wang YR, et al. Low-count whole-body PET/MRI restoration: an evaluation of dose reduction spectrum and five state-of-the-art artificial intelligence models. Eur J Nucl Med Mol Imaging. 2023;50(5):1337–50. https://doi.org/10.1007/S00259-022-06097-W . (PMID: 10.1007/S00259-022-06097-W3663361410387227)
Xie H, et al. Dose-aware diffusion model for 3D low-dose PET: multi-institutional validation with reader study and real low-dose data. arXiv preprint arXiv:2405.12996 , 2024.
Yu B, et al. Robust whole-body PET image denoising using 3D diffusion models: evaluation across various scanners, tracers, and dose levels. Eur J Nucl Med Mol Imaging. 2025. https://doi.org/10.1007/S00259-025-07122-4 . (PMID: 10.1007/S00259-025-07122-44135047212860873)
Liu Q, et al. A personalized deep learning denoising strategy for low-count PET images. Phys Med Biol. 2022;67(14):145014. (PMID: 10.1088/1361-6560/ac783d)
Rogasch JMM, Hofheinz F, van Heek L, Voltin C-A, Boellaard R, Kobe C. Influences on PET quantification and interpretation. Diagnostics. 2022;12(2):451. (PMID: 10.3390/diagnostics12020451352045428871060)
Bonardel G, et al. Clinical and phantom validation of a deep learning based denoising algorithm for F-18-FDG PET images from lower detection counting in comparison with the standard acquisition. EJNMMI Phys. 2022;9(1):36. (PMID: 10.1186/s40658-022-00465-z355438949095795)
Quak E, Weyts K, Jaudet C, Prigent A, Foucras G, Lasnon C. Artificial intelligence-based 68Ga-DOTATOC PET denoising for optimizing 68Ge/68Ga generator use throughout its lifetime. Front Med. 2023. https://doi.org/10.3389/FMED.2023.1137514 . (PMID: 10.3389/FMED.2023.1137514)
Weyts K, et al. Artificial intelligence-based PET denoising could allow a two-fold reduction in [18F]FDG PET acquisition time in digital PET/CT. Eur J Nucl Med Mol Imaging. 2022;49(11):3750–60. https://doi.org/10.1007/S00259-022-05800-1 . (PMID: 10.1007/S00259-022-05800-1355939259399218)
Mehranian A, et al. Deep learning–based time-of-flight (ToF) image enhancement of non-ToF PET scans. Eur J Nucl Med Mol Imaging. 2022;49(11):3740–9. https://doi.org/10.1007/S00259-022-05824-7/FIGURES/6 . (PMID: 10.1007/S00259-022-05824-7/FIGURES/6355070599399038)
Chaudhari AS, et al. Low-count whole-body PET with deep learning in a multicenter and externally validated study. NPJ Digit Med. 2021;4(1):127. (PMID: 10.1038/s41746-021-00497-2344266298382711)
He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2015. pp. 770–778. https://doi.org/10.1109/CVPR.2016.90 .
Singh G, Mittal A, Aggarwal N. ResDNN: deep residual learning for natural image denoising. IET Image Process. 2020;14(11):2425–34. https://doi.org/10.1049/iet-ipr.2019.0623 . (PMID: 10.1049/iet-ipr.2019.0623)
Katsari K, et al. Artificial intelligence for reduced dose <sup>18</sup>F-FDG PET examinations: a real-world deployment through a standardized framework and business case assessment. EJNMMI Phys. 2021;8(1):25. https://doi.org/10.1186/S40658-021-00374-7 . (PMID: 10.1186/S40658-021-00374-7336876027943690)
Dadgar M, Verstraete A, Maebe J, D’Asseler Y, Vandenberghe S. Assessing the deep learning based image quality enhancements for the BGO based GE omni legend PET/CT. EJNMMI Phys. 2024;11(1):86. (PMID: 10.1186/s40658-024-00688-23941263311484998)
*Further Information*
*Background: Positron Emission Tomography (PET) is a powerful diagnostic tool, but its availability, high cost and radiation burden limit its accessibility. Deep learning-based denoising offers a potential solution by enabling low-count PET scans, reducing tracer dose or scan time without compromising diagnostic utility. However, clinical validation of such approaches across different scanner technologies and radiotracers remains limited.
Methods: We conducted a multicenter, blinded evaluation of NUCLARITY, a deep learning-based denoising software, using PET data from three European hospitals. Data included 65 scans acquired with [1⁸F]FDG, [1⁸F]PSMA, [<sup>68</sup> Ga]PSMA, and [<sup>68</sup> Ga]DOTATATE on GE and Siemens systems not seen during model training. Low-count scans (50% simulated) were denoised and compared to full-count clinical scans. Image quality was assessed using RMSE, PSNR, and SSIM. Six nuclear physicians evaluated diagnostic image quality (DIQ), diagnostic confidence (DC), and lesion detection across six anatomical regions. Lesion quantification was compared using SUVmean, SUVmax, and MTV.
Results: Low-count enhanced (LCE) scans showed improved quantitative image quality metrics compared to unenhanced low-count scans (higher PSNR/SSIM, lower RMSE). Across 243 lesions, SUVmean and SUVmax showed high concordance between standard-count (SC) and LCE scans (CCC = 1.00 and 0.99, respectively). Diagnostic image quality and confidence were slightly lower on LCE versus SC scans, but only one reader indicated a clear preference for SC. Sensitivity and specificity for lesion detection in LCE scans were 99% and 99%, respectively, with interscan agreement exceeding inter-reader variability.
Conclusions: This is the first blinded, multicenter reader study evaluating a PET denoising algorithm in a European clinical setting across multiple tracers, incorporating unseen scanner technologies. The denoising algorithm demonstrated robust generalizability and preserved diagnostic accuracy on 50% count data. These findings support the clinical adoption of deep learning-based PET denoising to reduce dose or scan time for four commonly used tracers.
(© 2025. The Author(s).)*
*Declarations. Ethical approval and consent to participate: The study was conducted in accordance with the Declaration of Helsinki and used fully anonymized, retrospective data in compliance with the GDPR. Based on the nature of the data and existing data sharing agreements, the participating centers did not require formal ethics approval or informed consent under their institutional policies. Consent for publication: All authors have read and approved the manuscript. Competing interests: S.DK., T.B. and M.L. are currently employed at nuclivision and hold shares in the company. S.U. and S.V. act as advisors for nuclivision and hold shares in the company.*