Feasibility of Dedicated Breast Positron Emission Tomography Image Denoising Using a Residual Neural Network

Document Type : Original Article


1 Division of Clinical Radiology Service, Kyoto University Hospital, Kyoto, Japan

2 Department of Advanced Medical Imaging Research, Graduate School of Medicine Kyoto University, Kyoto , Japan

3 Department of Diagnostic Imaging and Nuclear Medicine, Graduate School of Medicine Kyoto University, Kyoto, Japan

4 Department of Breast Surgery, Graduate School of Medicine Kyoto University, Kyoto, Japan



Objective(s): This study aimed to create a deep learning (DL)-based denoising model using a residual neural network (Res-Net) trained to reduce noise in ring-type dedicated breast positron emission tomography (dbPET) images acquired in about half the emission time, and to evaluate the feasibility and the effectiveness of the model in terms of its noise reduction performance and preservation of quantitative values compared to conventional post-image filtering techniques.
Methods: Low-count (LC) and full-count (FC) PET images with acquisition durations of 3 and 7 minutes, respectively, were reconstructed. A Res-Net was trained to create a noise reduction model using fifteen patients’ data. The inputs to the network were LC images and its outputs were denoised PET (LC + DL) images, which should resemble FC images. To evaluate the LC + DL images, Gaussian and non-local mean (NLM) filters were applied to the LC images (LC + Gaussian and LC + NLM, respectively). To create reference images, a Gaussian filter was applied to the FC images (FC + Gaussian). The usefulness of our denoising model was objectively and visually evaluated using test data set of thirteen patients. The coefficient of variation (CV) of background fibroglandular tissue or fat tissue were measured to evaluate the performance of the noise reduction. The SUVmax and SUVpeak of lesions were also measured. The agreement of the SUV measurements was evaluated by Bland–Altman plots.
Results: The CV of background fibroglandular tissue in the LC + DL images was significantly lower (9.10 2.76) than the CVs in the LC (13.60  3.66) and LC + Gaussian images (11.51  3.56). No significant difference was observed in both SUVmax and SUVpeak of lesions between LC + DL and reference images. For the visual assessment, the smoothness rating for the LC + DL images was significantly better than that for the other images except for the reference images.
Conclusion: Our model reduced the noise in dbPET images acquired in about half the emission time while preserving quantitative values of lesions. This study demonstrates that machine learning is feasible and potentially performs better than conventional post-image filtering in dbPET denoising.


Main Subjects

  1. Fletcher JW, Djulbegovic B, Soares HP, Siegel BA, Lowe VJ, Lyman GH, et al. Recommendations on the use of 18F-FDG PET in oncology. J Nucl Med. 2008; 49(3):480–508.
  2. Groheux D, Espié M, Giacchetti S, Hindié E. Performance of FDG PET/CT in the clinical management of breast cancer. Radiology. 2013; 266(2):388–405.
  3. Groheux D, Cochet A, Humbert O, Alberini JL, Hindié E, Mankoff D. 18F-FDG PET/CT for staging and restaging of breast cancer. J Nucl Med. 2016; 57(Suppl 1):17S-26S.
  4. Lee H, Lee DE, Park S, Kim TS, Jung SY, Lee S, et al. Predicting response to neoadjuvant chemotherapy in patients with breast cancer: Combined statistical modeling using clinicopathological factors and FDG PET/CT texture parameters. Clin Nucl Med. 2019; 44(1):21–9.
  5. Jo J, Chung HW, So Y, Yoo YB, Park KS, Nam SE, et al. FDG PET/CT to predict recurrence of early breast invasive ductal carcinoma. Diagnostics. 2022; 12(3):694.
  6. Avril N, Rosé CA, Schelling M, Dose J, Kuhn W, Bense S, et al. Breast imaging with positron emission tomography and fluorine-18 fluorodeoxyglucose: Use and J Clin Oncol. 2000; 18(20):3495–502.
  7. Kumar R, Chauhan A, Zhuang H, Chandra P, Schnall M, Alavi A. Clinicopathologic factors associated with false negative FDG-PET in primary breast cancer. Breast Cancer Res Treat. 2006; 98:267–74.
  8. Sueoka S, Sasada S, Masumoto N, Emi A, Kadoya T, Okada M. Performance of dedicated breast positron emission tomography in the detection of small and low-grade breast cancer. Breast Cancer Res Treat. 2021; 187:125–33.
  9. Nahmias C, Wahl LM. Reproducibility of standardized uptake value measurements determined by 18F-FDG PET in malignant tumors. J Nucl Med. 2008; 49(11):1804–8.
  10. Pan T, Einstein SA, Kappadath SC, Grogg KS, Lois Gomez C, Alessio AM, et al. Performance evaluation of the 5-Ring GE Discovery MI PET/CT system using the national electrical manufacturers association NU 2-2012 Standard. Med Phys. 2019; 46(7):3025–33.
  11. Karlberg AM, Sæther O, Eikenes L, Goa PE. Quantitative comparison of PET performance-siemens biograph mCT and EJNMMI Phys. EJNMMI Physics; 2016; 3(1):1-4.
  12. Yano F, Itoh M, Hirakawa H, Yamamoto S, Yoshikawa A, Hatazawa J. Diagnostic Accuracy of Positron Emission Mammo-graphy with 18F-fluorodeoxyglucose in Breast Cancer Tumor of Less than 20 mm in Size. Asia Ocean J Nucl Med Biol. 2019; 7(1):13–21.
  13. Satoh Y, Kawamoto M, Kubota K, Murakami K, Hosono M, Senda M, et al. Clinical practice guidelines for high-resolution breast PET, 2019 edition. Ann Nucl Med. 2021; 35(3):406–14.
  14. Thompson CJ, Murthy K, Picard Y, Weinberg IN, Mako R. Positron emission mammography (PEM): A promising technique for detecting breast cancer. IEEE Trans Nucl Sci. 1995; 42(4):1012–7.
  15. Miyake KK, Kataoka M, Ishimori T, Matsumoto Y, Torii M, Takada M, et al. A proposed dedicated breast pet lexicon: Standardization of description and reporting of radiotracer uptake in the breast. 2021; 11(7):1267.
  16. Miyake KK, Nakamoto Y, Togashi K. Current status of dedicated breast PET imaging. Curr Radiol Rep. 2016; 4:16.
  17. Miyake KK, Matsumoto K, Inoue M, Nakamoto Y, Kanao S, Oishi T, et al. Performance evaluation of a new dedicated breast PET scanner using NEMA NU4-2008 J Nucl Med. 2014; 55(7):1198–203.
  18. Satoh Y, Motosugi U, Imai M, Onishi H. Comparison of dedicated breast positron emission tomography and whole-body positron emission tomography/computed tomography images: a common phantom study. Ann Nucl Med. 2020; 34:119–27.
  19. Berg WA, Madsen KS, Schilling K, Tartar M, Pisano ED, Larsen LH, et al. Breast cancer: Comparative effectiveness of positron emission mammography and MR imaging in presurgical planning for the ipsilateral breast. Radiology. 2011; 258(1):59–72.
  20. Sasada S, Masumoto N, Goda N, Kajitani K, Emi A, Kadoya T, et al. Dedicated breast PET for detecting residual disease after neoadjuvant chemotherapy in operable breast cancer: A prospective cohort study. Eur J Surg Oncol. 2018; 44(4):444–8.
  21. Yuge S, Miyake KK, Ishimori T, Kataoka M, Matsumoto Y, Fujimoto K, et al. Reproducibility assessment of uptake on dedicated breast PET for noise discrimination. Ann Nucl Med. 2023: 37(2): 121-30.
  22. Buades A, Coll B, Morel JM. A non-local algorithm for image denoising. Proc - 2005 IEEE Comput Soc Conf Comput Vis Pattern Recognition, CVPR 2005. IEEE; 2005; 2:60–5.
  23. Teoh EJ, McGowan DR, Macpherson RE, Bradley KM, Gleeson F V. Phantom and clinical evaluation of the Bayesian penalized likelihood reconstruction algorithm Q. Clear on an LYSO PET/CT system. J Nucl Med. 2015; 56(9):1447–52.
  24. Satoh Y, Imai M, Hirata K, Asakawa Y, Ikegawa C, Onishi H. Optimal relaxation parameters of dynamic row-action maximum likelihood algorithm and post-smoothing filter for image reconstruction of dedicated breast PET. Ann Nucl Med. 2021; 35:608–16.
  25. Schaefferkoetter J, Yan J, Ortega C, Sertic A, Lechtman E, Eshet Y, et al. Convolutional neural networks for improving image quality with noisy PET data. EJNMMI Res. 2020; 10(1):105.
  26. Weyts K, Lasnon C, Ciappuccini R, Lequesne J, Corroyer-Dulmont A, Quak E, 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.
  27. Mehranian A, Wollenweber SD, Walker MD, Bradley KM, Fielding PA, Su KH, et al. Image enhancement of whole-body oncology [18F]-FDG PET scans using deep neural networks to reduce noise. Eur J Nucl Med Mol Imaging. 2022; 49(2):539–49.
  28. Spuhler K, Serrano-Sosa M, Cattell R, DeLorenzo C, Huang C. Full-count PET recovery from low-count image using a dilated convolutional neural network. Med Phys. 2020; 47(10):4928–38.
  29. Zhang K, Zuo W, Chen Y, Meng D, Zhang L. Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising. IEEE Trans Image Process. 2017; 26(7): 3142–55.
  30. Liu P, Fang R. Wide Inference Network for Image denoising via learning pixel-distribution Prior. arXiv.2017; 1707.05414.
  31. Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. 32nd Int Conf Mach Learn ICML 2015. 2015; 448-56.
  32. He K, Zhang X, Ren S, Sun J. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. Proc IEEE Int Conf Comput Vis. 2015; 1026–34.
  33. Kingma DP, Ba JL. Adam: A method for stochastic optimization. 3rd Int Conf Learn Represent ICLR 2015 - Conf Track Proc. 2015. p. 1–15.
  34. Watanabe M, Nakamoto Y, Nakamoto R, Ishimori T, Saga T, Togashi K. Qualitative and quantitative assessment of nonlocal means reconstruction algorithm in a flexible PET scanner. Am J Roentgenol. 2021; 216(2):486–93.
  35. Gong K, Guan J, Liu CC, Qi J. PET image denoising using a deep neural network through fine tuning. IEEE Trans Radiat Plasma Med Sci. 2018; 3(2):153–61.
  36. Teymurazyan A, Riauka T, Jans HS, Robinson D. Properties of noise in positron emission tomography images recon-structed with filtered-backprojection and row-action maximum likelihood J Digit Imaging. 2013; 26:447–56.
  37. Kim JH, Ahn IJ, Nam WH, Chang Y, Ra JB. Post-filtering of PET image based on noise characteristic and spatial sensitivity distribution. IEEE Nucl Sci Symp Conf Rec. IEEE. 2013; pp: 1–3.
  38. Wang T, Sun M, Hu K. Dilated deep residual network for image denoising. Proc - Int Conf Tools with Artif Intell ICTAI. IEEE. 2017; pp: 1272–9.
  39. Tian C, Xu Y, Fei L, Wang J, Wen J, Luo N. Enhanced CNN for image denoising. CAAI Trans Intell Technol. 2019; 49(1):17–23.
  40. Zhou L, Schaefferkoetter JD, Tham IWK, Huang G, Yan J. Supervised learning with cyclegan for low-dose FDG PET image denoising. Med Image Anal. 2020; 65:101770.
  41. Panda A, Naskar R, Rajbans S, Pal S. A 3D wide residual network with perceptual loss for brain MRI image denoising. In2019 10th International Conference on Computing, Communication and Networking Techno-logies (ICCCNT). IEEE. 2019; pp: 1-7.
  42. Gholizadeh-Ansari M, Alirezaie J, Babyn P. Deep learning for low-dose CT denoising using perceptual loss and edge detection layer. J Digit Imaging. Journal of Digital Imaging; 2020; 33:504–15.
  43. Ataei S, Alirezaie J, Babyn P. Low dose ct denoising using dilated residual learning with perceptual loss and structural dissimilarity. In2020 IEEE 5th Middle East and Africa Conference on Biomedical Engineering (MECBME). IEEE. 2020; pp: 1-5.
  44. Shreyamsha Kumar BK. Image denoising based on non-local means filter and its method noise thresholding. Signal, Image Video Process. 2013; 7:1211–27.
  45. Zhang X, Hou G, Jianhua M, Yang W, Lin B, Xu Y, et al. Denoising MR images using non-local means filter with combined patch and pixel similarity. PLoS One. 2014; 9(6): e100240.
  46. Barrett HH, Wilson DW, Tsui BMW. Noise properties of the EM algorithm. I. Theory. Phys Med Biol. 1994; 39(5):833–46.