Donghwi Hwang1,2,3, Seung Kwan Kang1,2,3,4, Kyeong Yun Kim1,2,4, Hongyoon Choi2, Jae Sung Lee5,6,7,8,9. 1. Department of Biomedical Sciences, Seoul National University College of Medicine, Seoul, South Korea. 2. Department of Nuclear Medicine, Seoul National University College of Medicine, 103 Daehak-ro, Jongno-gu, Seoul, 03080, South Korea. 3. Artificial Intelligence Institute, Seoul National University, Seoul, South Korea. 4. Brightonix Imaging Inc., Seoul, South Korea. 5. Department of Biomedical Sciences, Seoul National University College of Medicine, Seoul, South Korea. jaes@snu.ac.kr. 6. Department of Nuclear Medicine, Seoul National University College of Medicine, 103 Daehak-ro, Jongno-gu, Seoul, 03080, South Korea. jaes@snu.ac.kr. 7. Artificial Intelligence Institute, Seoul National University, Seoul, South Korea. jaes@snu.ac.kr. 8. Brightonix Imaging Inc., Seoul, South Korea. jaes@snu.ac.kr. 9. Institute of Radiation Medicine, Medical Research Center, Seoul National University College of Medicine, Seoul, South Korea. jaes@snu.ac.kr.
Abstract
PURPOSE: This study aims to compare two approaches using only emission PET data and a convolution neural network (CNN) to correct the attenuation (μ) of the annihilation photons in PET. METHODS: One of the approaches uses a CNN to generate μ-maps from the non-attenuation-corrected (NAC) PET images (μ-CNNNAC). In the other method, CNN is used to improve the accuracy of μ-maps generated using maximum likelihood estimation of activity and attenuation (MLAA) reconstruction (μ-CNNMLAA). We investigated the improvement in the CNN performance by combining the two methods (μ-CNNMLAA+NAC) and the suitability of μ-CNNNAC for providing the scatter distribution required for MLAA reconstruction. Image data from 18F-FDG (n = 100) or 68 Ga-DOTATOC (n = 50) PET/CT scans were used for neural network training and testing. RESULTS: The error of the attenuation correction factors estimated using μ-CT and μ-CNNNAC was over 7%, but that of scatter estimates was only 2.5%, indicating the validity of the scatter estimation from μ-CNNNAC. However, CNNNAC provided less accurate bone structures in the μ-maps, while the best results in recovering the fine bone structures were obtained by applying CNNMLAA+NAC. Additionally, the μ-values in the lungs were overestimated by CNNNAC. Activity images (λ) corrected for attenuation using μ-CNNMLAA and μ-CNNMLAA+NAC were superior to those corrected using μ-CNNNAC, in terms of their similarity to λ-CT. However, the improvement in the similarity with λ-CT by combining the CNNNAC and CNNMLAA approaches was insignificant (percent error for lung cancer lesions, λ-CNNNAC = 5.45% ± 7.88%; λ-CNNMLAA = 1.21% ± 5.74%; λ-CNNMLAA+NAC = 1.91% ± 4.78%; percent error for bone cancer lesions, λ-CNNNAC = 1.37% ± 5.16%; λ-CNNMLAA = 0.23% ± 3.81%; λ-CNNMLAA+NAC = 0.05% ± 3.49%). CONCLUSION: The use of CNNNAC was feasible for scatter estimation to address the chicken-egg dilemma in MLAA reconstruction, but CNNMLAA outperformed CNNNAC.
PURPOSE: This study aims to compare two approaches using only emission PET data and a convolution neural network (CNN) to correct the attenuation (μ) of the annihilation photons in PET. METHODS: One of the approaches uses a CNN to generate μ-maps from the non-attenuation-corrected (NAC) PET images (μ-CNNNAC). In the other method, CNN is used to improve the accuracy of μ-maps generated using maximum likelihood estimation of activity and attenuation (MLAA) reconstruction (μ-CNNMLAA). We investigated the improvement in the CNN performance by combining the two methods (μ-CNNMLAA+NAC) and the suitability of μ-CNNNAC for providing the scatter distribution required for MLAA reconstruction. Image data from 18F-FDG (n = 100) or 68 Ga-DOTATOC (n = 50) PET/CT scans were used for neural network training and testing. RESULTS: The error of the attenuation correction factors estimated using μ-CT and μ-CNNNAC was over 7%, but that of scatter estimates was only 2.5%, indicating the validity of the scatter estimation from μ-CNNNAC. However, CNNNAC provided less accurate bone structures in the μ-maps, while the best results in recovering the fine bone structures were obtained by applying CNNMLAA+NAC. Additionally, the μ-values in the lungs were overestimated by CNNNAC. Activity images (λ) corrected for attenuation using μ-CNNMLAA and μ-CNNMLAA+NAC were superior to those corrected using μ-CNNNAC, in terms of their similarity to λ-CT. However, the improvement in the similarity with λ-CT by combining the CNNNAC and CNNMLAA approaches was insignificant (percent error for lung cancer lesions, λ-CNNNAC = 5.45% ± 7.88%; λ-CNNMLAA = 1.21% ± 5.74%; λ-CNNMLAA+NAC = 1.91% ± 4.78%; percent error for bone cancer lesions, λ-CNNNAC = 1.37% ± 5.16%; λ-CNNMLAA = 0.23% ± 3.81%; λ-CNNMLAA+NAC = 0.05% ± 3.49%). CONCLUSION: The use of CNNNAC was feasible for scatter estimation to address the chicken-egg dilemma in MLAA reconstruction, but CNNMLAA outperformed CNNNAC.
Authors: Vincent Keereman; Yves Fierens; Tom Broux; Yves De Deene; Max Lonneux; Stefaan Vandenberghe Journal: J Nucl Med Date: 2010-05 Impact factor: 10.057
Authors: Ciprian Catana; Andre van der Kouwe; Thomas Benner; Christian J Michel; Michael Hamm; Matthias Fenchel; Bruce Fischl; Bruce Rosen; Matthias Schmand; A Gregory Sorensen Journal: J Nucl Med Date: 2010-09 Impact factor: 10.057
Authors: Axel Martinez-Möller; Michael Souvatzoglou; Gaspar Delso; Ralph A Bundschuh; Christophe Chefd'hotel; Sibylle I Ziegler; Nassir Navab; Markus Schwaiger; Stephan G Nekolla Journal: J Nucl Med Date: 2009-03-16 Impact factor: 10.057