Korean Journal of Radiology

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

DOI QR Code

Bone Suppression on Chest Radiographs for Pulmonary Nodule Detection: Comparison between a Generative Adversarial Network and Dual-Energy Subtraction

  • Kyungsoo Bae (Department of Radiology, Institute of Health Sciences, Gyeongsang National University School of Medicine) ;
  • Dong Yul Oh (Monitor Corporation) ;
  • Il Dong Yun (Division of Computer and Electronic System Engineering, Hankuk University of Foreign Studies) ;
  • Kyung Nyeo Jeon (Department of Radiology, Institute of Health Sciences, Gyeongsang National University School of Medicine)
  • Received : 2021.02.19
  • Accepted : 2021.08.17
  • Published : 2022.01.01
Download PDF
⟨ Previous Next ⟩

Abstract

Objective: To compare the effects of bone suppression imaging using deep learning (BSp-DL) based on a generative adversarial network (GAN) and bone subtraction imaging using a dual energy technique (BSt-DE) on radiologists' performance for pulmonary nodule detection on chest radiographs (CXRs). Materials and Methods: A total of 111 adults, including 49 patients with 83 pulmonary nodules, who underwent both CXR using the dual energy technique and chest CT, were enrolled. Using CT as a reference, two independent radiologists evaluated CXR images for the presence or absence of pulmonary nodules in three reading sessions (standard CXR, BSt-DE CXR, and BSp-DL CXR). Person-wise and nodule-wise performances were assessed using receiver-operating characteristic (ROC) and alternative free-response ROC (AFROC) curve analyses, respectively. Subgroup analyses based on nodule size, location, and the presence of overlapping bones were performed. Results: BSt-DE with an area under the AFROC curve (AUAFROC) of 0.996 and 0.976 for readers 1 and 2, respectively, and BSp-DL with AUAFROC of 0.981 and 0.958, respectively, showed better nodule-wise performance than standard CXR (AUAFROC of 0.907 and 0.808, respectively; p ≤ 0.005). In the person-wise analysis, BSp-DL with an area under the ROC curve (AUROC) of 0.984 and 0.931 for readers 1 and 2, respectively, showed better performance than standard CXR (AUROC of 0.915 and 0.798, respectively; p ≤ 0.011) and comparable performance to BSt-DE (AUROC of 0.988 and 0.974; p ≥ 0.064). BSt-DE and BSp-DL were superior to standard CXR for detecting nodules overlapping with bones (p < 0.017) or in the upper/middle lung zone (p < 0.017). BSt-DE was superior (p < 0.017) to BSp-DL in detecting peripheral and sub-centimeter nodules. Conclusion: BSp-DL (GAN-based bone suppression) showed comparable performance to BSt-DE and can improve radiologists' performance in detecting pulmonary nodules on CXRs. Nevertheless, for better delineation of small and peripheral nodules, further technical improvements are required.

Keywords

Acknowledgement

This work was supported by a fund of Research Promotion Program, Gyeongsang National University Hospital, 2019.

References

  1. Geitung JT, Skjaerstad LM, Gothlin JH. Clinical utility of chest roentgenograms. Eur Radiol 1999;9:721-723
  2. Toomes H, Delphendahl A, Manke HG, Vogt-Moykopf I. The coin lesion of the lung. A review of 955 resected coin lesions. Cancer 1983;51:534-537
  3. Del Ciello A, Franchi P, Contegiacomo A, Cicchetti G, Bonomo L, Larici AR. Missed lung cancer: when, where, and why? Diagn Interv Radiol 2017;23:118-126
  4. Austin JH, Romney BM, Goldsmith LS. Missed bronchogenic carcinoma: radiographic findings in 27 patients with a potentially resectable lesion evident in retrospect. Radiology 1992;182:115-122
  5. de Groot PM, Carter BW, Abbott GF, Wu CC. Pitfalls in chest radiographic interpretation: blind spots. Semin Roentgenol 2015;50:197-209
  6. Shah PK, Austin JH, White CS, Patel P, Haramati LB, Pearson GD, et al. Missed non-small cell lung cancer: radiographic findings of potentially resectable lesions evident only in retrospect. Radiology 2003;226:235-241
  7. Oda S, Awai K, Funama Y, Utsunomiya D, Yanaga Y, Kawanaka K, et al. Detection of small pulmonary nodules on chest radiographs: efficacy of dual-energy subtraction technique using flat-panel detector chest radiography. Clin Radiol 2010;65:609-615
  8. Li F, Engelmann R, Doi K, MacMahon H. Improved detection of small lung cancers with dual-energy subtraction chest radiography. AJR Am J Roentgenol 2008;190:886-891
  9. Szucs-Farkas Z, Patak MA, Yuksel-Hatz S, Ruder T, Vock P. Single-exposure dual-energy subtraction chest radiography: detection of pulmonary nodules and masses in clinical practice. Eur Radiol 2008;18:24-31
  10. Kelcz F, Zink FE, Peppler WW, Kruger DG, Ergun DL, Mistretta CA. Conventional chest radiography vs dual-energy computed radiography in the detection and characterization of pulmonary nodules. AJR Am J Roentgenol 1994;162:271-278
  11. Kashani H, Varon CA, Paul NS, Gang GJ, Van Metter R, Yorkston J, et al. Diagnostic performance of a prototype dual-energy chest imaging system ROC analysis. Acad Radiol 2010;17:298-308
  12. Kuhlman JE, Collins J, Brooks GN, Yandow DR, Broderick LS. Dual-energy subtraction chest radiography: what to look for beyond calcified nodules. Radiographics 2006;26:79-92
  13. Gasarev M, Kuleev R, Khan A, Rivera AR, Khattak AM. Deep learning models for bone suppression in chest radiographs. Proceedings of the 2017 IEEE Conference on Computational Interlligence in Bioinformatics and Computational Biology (CIBCB); 2017 Aug 23-25; Manchester, UK: IEEE; 2017; p. 1-7
  14. Yang W, Chen Y, Liu Y, Zhong L, Qin G, Lu Z, et al. Cascade of multi-scale convolutional neural networks for bone suppression of chest radiographs in gradient domain. Med Image Anal 2017;35:421-433
  15. Freedman MT, Lo SC, Seibel JC, Bromley CM. Lung nodules: improved detection with software that suppresses the rib and clavicle on chest radiographs. Radiology 2011;260:265-273
  16. Schalekamp S, van Ginneken B, Meiss L, Peters-Bax L, Quekel LG, Snoeren MM, et al. Bone suppressed images improve radiologists' detection performance for pulmonary nodules in chest radiographs. Eur J Radiol 2013;82:2399-2405
  17. Hong GS, Do KH, Lee CW. Added value of bone suppression image in the detection of subtle lung lesions on chest radiographs with regard to reader's expertise. J Korean Med Sci 2019;34:e250
  18. Ogul H, Ogul BB, Agildere AM, Bayrak T, Sumer E. Eliminating rib shadows in chest radiographic images providing diagnostic assistance. Comput Methods Programs Biomed 2016;127:174-184
  19. Schalekamp S, van Ginneken B, Koedam E, Snoeren MM, Tiehuis AM, Wittenberg R, et al. Computer-aided detection improves detection of pulmonary nodules in chest radiographs beyond the support by bone-suppressed images. Radiology 2014;272:252-261
  20. Oh DY, Yun ID. Learning bone suppression from dual energy chest X-rays using adversarial networks. Arxiv.org Web site. https://arxiv.org/abs/1811.02628. Published November 5, 2018. Accessed November 19, 2020
  21. Kang E, Min J, Ye JC. A deep convolutional neural network using directional wavelets for low-dose X-ray CT reconstruction. Med Phys 2017;44:e360-e375
  22. Hong GS, Do KH, Son AY, Jo KW, Kim KP, Yun J, et al. Value of bone suppression software in chest radiographs for improving image quality and reducing radiation dose. Eur Radiol 2021;31:5160-5171
  23. Chakraborty DP, Zhai X. Analysis of data acquired using ROC paradigm and its extensions. Mran.microsoft Web site. https://mran.microsoft.com/snapshot/2015-05-08/web/packages/RJafroc/vignettes/RJafroc.pdf. Accessed May 13, 2021
  24. van Ginneken B, Stegmann MB, Loog M. Segmentation of anatomical structures in chest radiographs using supervised methods: a comparative study on a public database. Med Image Anal 2006;10:19-40
  25. Suzuki K, Abe H, MacMahon H, Doi K. Image-processing technique for suppressing ribs in chest radiographs by means of massive training artificial neural network (MTANN). IEEE Trans Med Imaging 2006;25:406-416
  26. Li F, Engelmann R, Pesce L, Armato SG 3rd, Macmahon H. Improved detection of focal pneumonia by chest radiography with bone suppression imaging. Eur Radiol 2012;22:2729-2735
  27. Zarshenas A, Liu J, Forti P, Suzuki K. Separation of bones from soft tissue in chest radiographs: anatomy-specific orientation-frequency-specific deep neural network convolution. Med Phys 2019;46:2232-2242
  28. Oda S, Awai K, Suzuki K, Yanaga Y, Funama Y, MacMahon H, et al. Performance of radiologists in detection of small pulmonary nodules on chest radiographs: effect of rib suppression with a massive-training artificial neural network. AJR Am J Roentgenol 2009;193:W397-W402
  29. Zhou Z, Zhou L, Shen K. Dilated conditional GAN for bone suppression in chest radiographs with enforced semantic features. Med Phys 2020;47:6207-6215
  30. Endo K, Kaneko A, Horiuchi Y, Kasuga N, Ishizaki U, Sakai S. Detectability of pulmonary nodules on chest radiographs: bone suppression versus standard technique with single versus dual monitors for visualization. Jpn J Radiol 2020;38:676-682
  31. Szucs-Farkas Z, Schick A, Cullmann JL, Ebner L, Megyeri B, Vock P, et al. Comparison of dual-energy subtraction and electronic bone suppression combined with computer-aided detection on chest radiographs: effect on human observers' performance in nodule detection. AJR Am J Roentgenol 2013;200:1006-1013
  32. Li F, Engelmann R, Pesce LL, Doi K, Metz CE, Macmahon H. Small lung cancers: improved detection by use of bone suppression imaging--comparison with dual-energy subtraction chest radiography. Radiology 2011;261:937-949
  33. Miyoshi T, Yoshida J, Aramaki N, Matsumura Y, Aokage K, Hishida T, et al. Effectiveness of bone suppression imaging in the detection of lung nodules on chest radiographs: relevance to anatomic location and observer's experience. J Thorac Imaging 2017;32:398-405 
  34. Li F, Hara T, Shiraishi J, Engelmann R, MacMahon H, Doi K. Improved detection of subtle lung nodules by use of chest radiographs with bone suppression imaging: receiver operating characteristic analysis with and without localization. AJR Am J Roentgenol 2011;196:W535-W541