Korean Journal of Radiology

The Korean Society of Radiology (대한영상의학회)
권호 표지
DOI QR코드

DOI QR Code

Generative Adversarial Network-Based Image Conversion Among Different Computed Tomography Protocols and Vendors: Effects on Accuracy and Variability in Quantifying Regional Disease Patterns of Interstitial Lung Disease

  • Hye Jeon Hwang (Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Hyunjong Kim (Robotics Program, Korea Advanced Institute of Science and Technology (KAIST)) ;
  • Joon Beom Seo (Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Jong Chul Ye (Kim Jaechul Graduate School of AI, Korea Advanced Institute of Science and Technology (KAIST)) ;
  • Gyutaek Oh (Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology (KAIST)) ;
  • Sang Min Lee (Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Ryoungwoo Jang (Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Jihye Yun (Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Namkug Kim (Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine) ;
  • Hee Jun Park (Coreline Soft, Co., Ltd) ;
  • Ho Yun Lee (Department of Radiology and Center for Imaging Science, Samsung Medical Center, School of Medicine, Sungkyunkwan University) ;
  • Soon Ho Yoon (Department of Radiology, Seoul National University Hospital, Seoul National University College of Medicine) ;
  • Kyung Eun Shin (Department of Radiology, Soonchunhyang University Bucheon Hospital) ;
  • Jae Wook Lee (Department of Radiology, Soonchunhyang University Bucheon Hospital) ;
  • Woocheol Kwon (Department of Radiology, Ewha Womans University Seoul Hospital) ;
  • Joo Sung Sun (Department of Radiology, Ajou University School of Medicine) ;
  • Seulgi You (Department of Radiology, Ajou University School of Medicine) ;
  • Myung Hee Chung (Department of Radiology, Bucheon St. Mary's Hospital, College of Medicine, The Catholic University of Korea) ;
  • Bo Mi Gil (Department of Radiology, Bucheon St. Mary's Hospital, College of Medicine, The Catholic University of Korea) ;
  • Jae-Kwang Lim (Department of Radiology, Kyungpook National University School of Medicine) ;
  • Youkyung Lee (Department of Radiology, Hanyang University Guri Hospital, Hanyang University College of Medicine) ;
  • Su Jin Hong (Department of Radiology, Hanyang University Guri Hospital, Hanyang University College of Medicine) ;
  • Yo Won Choi (Department of Radiology, Hanyang University Seoul Hospital, Hanyang University College of Medicine)
  • Received : 2023.01.31
  • Accepted : 2023.06.18
  • Published : 2023.08.01
Download PDF
⟨ Previous Next ⟩

Abstract

Objective: To assess whether computed tomography (CT) conversion across different scan parameters and manufacturers using a routable generative adversarial network (RouteGAN) can improve the accuracy and variability in quantifying interstitial lung disease (ILD) using a deep learning-based automated software. Materials and Methods: This study included patients with ILD who underwent thin-section CT. Unmatched CT images obtained using scanners from four manufacturers (vendors A-D), standard- or low-radiation doses, and sharp or medium kernels were classified into groups 1-7 according to acquisition conditions. CT images in groups 2-7 were converted into the target CT style (Group 1: vendor A, standard dose, and sharp kernel) using a RouteGAN. ILD was quantified on original and converted CT images using a deep learning-based software (Aview, Coreline Soft). The accuracy of quantification was analyzed using the dice similarity coefficient (DSC) and pixel-wise overlap accuracy metrics against manual quantification by a radiologist. Five radiologists evaluated quantification accuracy using a 10-point visual scoring system. Results: Three hundred and fifty CT slices from 150 patients (mean age: 67.6 ± 10.7 years; 56 females) were included. The overlap accuracies for quantifying total abnormalities in groups 2-7 improved after CT conversion (original vs. converted: 0.63 vs. 0.68 for DSC, 0.66 vs. 0.70 for pixel-wise recall, and 0.68 vs. 0.73 for pixel-wise precision; P < 0.002 for all). The DSCs of fibrosis score, honeycombing, and reticulation significantly increased after CT conversion (0.32 vs. 0.64, 0.19 vs. 0.47, and 0.23 vs. 0.54, P < 0.002 for all), whereas those of ground-glass opacity, consolidation, and emphysema did not change significantly or decreased slightly. The radiologists' scores were significantly higher (P < 0.001) and less variable on converted CT. Conclusion: CT conversion using a RouteGAN can improve the accuracy and variability of CT images obtained using different scan parameters and manufacturers in deep learning-based quantification of ILD.

Keywords

Acknowledgement

This work was supported by the Korea Medical Device Development Fund grant funded by the Korea government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, Republic of Korea, the Ministry of Food and Drug Safety) (Project Number: NTIS 1711138474).

References

  1. Gay SE, Kazerooni EA, Toews GB, Lynch JP 3rd, Gross BH, Cascade PN, et al. Idiopathic pulmonary fibrosis: predicting response to therapy and survival. Am J Respir Crit Care Med 1998;157(4 Pt 1):1063-1072 https://doi.org/10.1164/ajrccm.157.4.9703022
  2. Muller NL. Clinical value of high-resolution CT in chronic diffuse lung disease. AJR Am J Roentgenol 1991;157:1163-1170 https://doi.org/10.2214/ajr.157.6.1950859
  3. Nishimura K, Izumi T, Kitaichi M, Nagai S, Itoh H. The diagnostic accuracy of high-resolution computed tomography in diffuse infiltrative lung diseases. Chest 1993;104:1149-1155 https://doi.org/10.1378/chest.104.4.1149
  4. Scatarige JC, Diette GB, Haponik EF, Merriman B, Fishman EK. Utility of high-resolution CT for management of diffuse lung disease: results of a survey of U.S. pulmonary physicians. Acad Radiol 2003;10:167-175 https://doi.org/10.1016/S1076-6332(03)80041-7
  5. Aziz ZA, Wells AU, Hansell DM, Bain GA, Copley SJ, Desai SR, et al. HRCT diagnosis of diffuse parenchymal lung disease: inter-observer variation. Thorax 2004;59:506-511 https://doi.org/10.1136/thx.2003.020396
  6. Collins CD, Wells AU, Hansell DM, Morgan RA, MacSweeney JE, du Bois RM, et al. Observer variation in pattern type and extent of disease in fibrosing alveolitis on thin section computed tomography and chest radiography. Clin Radiol 1994;49:236-240 https://doi.org/10.1016/S0009-9260(05)81847-1
  7. Best AC, Lynch AM, Bozic CM, Miller D, Grunwald GK, Lynch DA. Quantitative CT indexes in idiopathic pulmonary fibrosis: relationship with physiologic impairment. Radiology 2003;228:407-414
  8. Best AC, Meng J, Lynch AM, Bozic CM, Miller D, Grunwald GK, et al. Idiopathic pulmonary fibrosis: physiologic tests, quantitative CT indexes, and CT visual scores as predictors of mortality. Radiology 2008;246:935-940 https://doi.org/10.1148/radiol.2463062200
  9. Castellano G, Bonilha L, Li LM, Cendes F. Texture analysis of medical images. Clin Radiol 2004;59:1061-1069 https://doi.org/10.1016/j.crad.2004.07.008
  10. Delorme S, Keller-Reichenbecher MA, Zuna I, Schlegel W, Van Kaick G. Usual interstitial pneumonia. Quantitative assessment of high-resolution computed tomography findings by computer-assisted texture-based image analysis. Invest Radiol 1997;32:566-574
  11. Rodriguez LH, Vargas PF, Raff U, Lynch DA, Rojas GM, Moxley DM, et al. Automated discrimination and quantification of idiopathic pulmonary fibrosis from normal lung parenchyma using generalized fractal dimensions in high-resolution computed tomography images. Acad Radiol 1995;2:10-18 https://doi.org/10.1016/S1076-6332(05)80240-5
  12. Zavaletta VA, Bartholmai BJ, Robb RA. High resolution multidetector CT-aided tissue analysis and quantification of lung fibrosis. Acad Radiol 2007;14:772-787 https://doi.org/10.1016/j.acra.2007.03.009
  13. Iwasawa T, Takemura T, Okudera K, Gotoh T, Iwao Y, Kitamura H, et al. The importance of subpleural fibrosis in the prognosis of patients with idiopathic interstitial pneumonias. Eur J Radiol 2017;90:106-113 https://doi.org/10.1016/j.ejrad.2017.02.037
  14. Sverzellati N, Calabro E, Chetta A, Concari G, Larici AR, Mereu M, et al. Visual score and quantitative CT indices in pulmonary fibrosis: Relationship with physiologic impairment. Radiol Med 2007;112:1160-1172 https://doi.org/10.1007/s11547-007-0213-x
  15. Yoon RG, Seo JB, Kim N, Lee HJ, Lee SM, Lee YK, et al. Quantitative assessment of change in regional disease patterns on serial HRCT of fibrotic interstitial pneumonia with texture-based automated quantification system. Eur Radiol 2013;23:692-701 https://doi.org/10.1007/s00330-012-2634-8
  16. Jacob J, Bartholmai BJ, Rajagopalan S, Brun AL, Egashira R, Karwoski R, et al. Evaluation of computer-based computer tomography stratification against outcome models in connective tissue disease-related interstitial lung disease: a patient outcome study. BMC Med 2016;14:190
  17. Lee SM, Seo JB, Oh SY, Kim TH, Song JW, Lee SM, et al. Prediction of survival by texture-based automated quantitative assessment of regional disease patterns on CT in idiopathic pulmonary fibrosis. Eur Radiol 2018;28:1293-1300 https://doi.org/10.1007/s00330-017-5028-0
  18. Maldonado F, Moua T, Rajagopalan S, Karwoski RA, Raghunath S, Decker PA, et al. Automated quantification of radiological patterns predicts survival in idiopathic pulmonary fibrosis. Eur Respir J 2014;43:204-212 https://doi.org/10.1183/09031936.00071812
  19. Chen-Mayer HH, Fuld MK, Hoppel B, Judy PF, Sieren JP, Guo J, et al. Standardizing CT lung density measure across scanner manufacturers. Med Phys 2017;44:974-985 https://doi.org/10.1002/mp.12087
  20. Gierada DS, Bierhals AJ, Choong CK, Bartel ST, Ritter JH, Das NA, et al. Effects of CT section thickness and reconstruction kernel on emphysema quantification relationship to the magnitude of the CT emphysema index. Acad Radiol 2010;17:146-156 https://doi.org/10.1016/j.acra.2009.08.007
  21. Kemerink GJ, Kruize HH, Lamers RJ, van Engelshoven JM. Density resolution in quantitative computed tomography of foam and lung. Med Phys 1996;23:1697-1708 https://doi.org/10.1118/1.597757
  22. Madani A, De Maertelaer V, Zanen J, Gevenois PA. Pulmonary emphysema: radiation dose and section thickness at multidetector CT quantification--comparison with macroscopic and microscopic morphometry. Radiology 2007;243:250-257 https://doi.org/10.1148/radiol.2431060194
  23. Kim J, Lee JK, Lee KM. Accurate image super-resolution using very deep convolutional networks. In: 2016 The IEEE Conference on Computer Vision and Pattern Recognition (CVPR); 2016 Jun 27-30; Las Vegas, USA. Danvers: The Institute of Electrical and Electronics Engineers, Inc; 2016. p. 1646-1654
  24. Kim H, Oh G, Seo JB, Hwang HJ, Lee SM, Yun J, et al. Multi-domain CT translation by a routable translation network. Phys Med Biol 2022;67:21
  25. Lee SM, Lee JG, Lee G, Choe J, Do KH, Kim N, et al. CT image conversion among different reconstruction kernels without a sinogram by using a convolutional neural network. Korean J Radiol 2019;20:295-303 https://doi.org/10.3348/kjr.2018.0249
  26. Choe J, Lee SM, Do KH, Lee G, Lee JG, Lee SM, et al. Deep learning-based image conversion of CT reconstruction kernels improves radiomics reproducibility for pulmonary nodules or masses. Radiology 2019;292:365-373 https://doi.org/10.1148/radiol.2019181960
  27. Raghu G, Collard HR, Egan JJ, Martinez FJ, Behr J, Brown KK, et al. An official ATS/ERS/JRS/ALAT statement: idiopathic pulmonary fibrosis: evidence-based guidelines for diagnosis and management. Am J Respir Crit Care Med 2011;183:788-824 https://doi.org/10.1164/rccm.2009-040GL
  28. Travis WD, Costabel U, Hansell DM, King TE Jr, Lynch DA, Nicholson AG, et al. An official American Thoracic Society/European Respiratory Society statement: update of the international multidisciplinary classification of the idiopathic interstitial pneumonias. Am J Respir Crit Care Med 2013;188:733-748 https://doi.org/10.1164/rccm.201308-1483ST
  29. Kim GB, Jung KH, Lee Y, Kim HJ, Kim N, Jun S, et al. Comparison of shallow and deep learning methods on classifying the regional pattern of diffuse lung disease. J Digit Imaging 2018;31:415-424 https://doi.org/10.1007/s10278-017-0028-9
  30. Choe J, Hwang HJ, Seo JB, Lee SM, Yun J, Kim MJ, et al. Content-based image retrieval by using deep learning for interstitial lung disease diagnosis with chest CT. Radiology 2022;302:187-197
  31. Taha AA, Hanbury A. Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool. BMC Med Imaging 2015;15:29
  32. Lee SB, Cho YJ, Hong Y, Jeong D, Lee J, Kim SH, et al. Deep learning-based image conversion improves the reproducibility of computed tomography radiomics features a phantom study. Invest Radiol 2022;57:308-317 https://doi.org/10.1097/RLI.0000000000000839
  33. Martensson G, Ferreira D, Granberg T, Cavallin L, Oppedal K, Padovani A, et al. The reliability of a deep learning model in clinical out-of-distribution MRI data: a multicohort study. Med Image Anal 2020;66:101714
  34. Reinke A, Eisenmann M, Tizabi MD, Sudre CH, Radsch T, Antonelli M, et al. Common limitations of performance metrics in biomedical image analysis. In: Medical Imaging with Deep Learning 2021; 2021 Jul 7-9; Lubeck, Germany. 2021
  35. Kim GHJ, Weigt SS, Belperio JA, Brown MS, Shi Y, Lai JH, et al. Prediction of idiopathic pulmonary fibrosis progression using early quantitative changes on CT imaging for a short term of clinical 18-24-month follow-ups. Eur Radiol 2020;30:726-734  https://doi.org/10.1007/s00330-019-06402-6