Imaging Science in Dentistry

Korean Academy of Oral and Maxillofacial Radiology (대한영상치의학회)
권호 표지
DOI QR코드

DOI QR Code

Synthesis of T2-weighted images from proton density images using a generative adversarial network in a temporomandibular joint magnetic resonance imaging protocol

  • Chena, Lee (Department of Oral and Maxillofacial Radiology, Yonsei University College of Dentistry) ;
  • Eun-Gyu, Ha (Department of Oral and Maxillofacial Radiology, Yonsei University College of Dentistry) ;
  • Yoon Joo, Choi (Department of Oral and Maxillofacial Radiology, Yonsei University College of Dentistry) ;
  • Kug Jin, Jeon (Department of Oral and Maxillofacial Radiology, Yonsei University College of Dentistry) ;
  • Sang-Sun, Han (Department of Oral and Maxillofacial Radiology, Yonsei University College of Dentistry)
  • Received : 2022.07.14
  • Accepted : 2022.09.24
  • Published : 2022.12.31
Download PDF
⟨ Previous Next ⟩

Abstract

Purpose: This study proposed a generative adversarial network (GAN) model for T2-weighted image (WI) synthesis from proton density (PD)-WI in a temporomandibular joint(TMJ) magnetic resonance imaging (MRI) protocol. Materials and Methods: From January to November 2019, MRI scans for TMJ were reviewed and 308 imaging sets were collected. For training, 277 pairs of PD- and T2-WI sagittal TMJ images were used. Transfer learning of the pix2pix GAN model was utilized to generate T2-WI from PD-WI. Model performance was evaluated with the structural similarity index map (SSIM) and peak signal-to-noise ratio (PSNR) indices for 31 predicted T2-WI (pT2). The disc position was clinically diagnosed as anterior disc displacement with or without reduction, and joint effusion as present or absent. The true T2-WI-based diagnosis was regarded as the gold standard, to which pT2-based diagnoses were compared using Cohen's ĸ coefficient. Results: The mean SSIM and PSNR values were 0.4781(±0.0522) and 21.30(±1.51) dB, respectively. The pT2 protocol showed almost perfect agreement(ĸ=0.81) with the gold standard for disc position. The number of discordant cases was higher for normal disc position (17%) than for anterior displacement with reduction (2%) or without reduction (10%). The effusion diagnosis also showed almost perfect agreement(ĸ=0.88), with higher concordance for the presence (85%) than for the absence (77%) of effusion. Conclusion: The application of pT2 images for a TMJ MRI protocol useful for diagnosis, although the image quality of pT2 was not fully satisfactory. Further research is expected to enhance pT2 quality.

Keywords

Acknowledgement

This work was funded by a National Research Foundation of Korea (NRF) grant funded by the Korean government(MSIT)(No. 2022R1A2B5B01002517).

References

  1. Tomas X, Pomes J, Berenguer J, Quinto L, Nicolau C, Mercader JM, et al. MR imaging of temporomandibular joint dysfunction: a pictorial review. Radiographics 2006; 26: 765-81. https://doi.org/10.1148/rg.263055091
  2. Navallas M, Inarejos EJ, Iglesias E, Cho Lee GY, Rodriguez N, Anton J. MR imaging of the temporomandibular joint in juvenile idiopathic arthritis: technique and findings. Radiographics 2017; 37: 595-612. https://doi.org/10.1148/rg.2017160078
  3. Tokuda O, Harada Y, Shiraishi G, Motomura T, Fukuda K, Kimura M, et al. MRI of the anatomical structures of the knee: the proton density-weighted fast spin-echo sequence vs the proton density-weighted fast-recovery fast spin-echo sequence. Br J Radiol 2012; 85: e686-93. https://doi.org/10.1259/bjr/99570113
  4. Lustig M, Donoho D, Pauly JM. Sparse MRI: the application of compressed sensing for rapid MR imaging. Magn Reson Med 2007; 58: 1182-95. https://doi.org/10.1002/mrm.21391
  5. Hollingsworth KG. Reducing acquisition time in clinical MRI by data undersampling and compressed sensing reconstruction. Phys Med Biol 2015; 60: R297-322. https://doi.org/10.1088/0031-9155/60/21/R297
  6. Shin Y, Yang J, Lee YH. Deep generative adversarial networks: applications in musculoskeletal imaging. Radiol Artif Intell 2021; 3: e200157.
  7. Hwang JJ, Jung YH, Cho BH, Heo MS. Very deep super-resolution for efficient cone-beam computed tomographic image restoration. Imaging Sci Dent 2020; 50: 331-7. https://doi.org/10.5624/isd.2020.50.4.331
  8. Yang G, Yu S, Dong H, Slabaugh G, Dragotti PL, Ye X, et al. DAGAN: deep de-aliasing generative adversarial networks for fast compressed sensing MRI reconstruction. IEEE Trans Med Imaging 2018; 37: 1310-21.
  9. Mori M, Fujioka T, Katsuta L, Kikuchi Y, Oda G, Nakagawa T, et al. Feasibility of new fat suppression for breast MRI using pix2pix. Jpn J Radiol 2020; 38: 1075-81. https://doi.org/10.1007/s11604-020-01012-5
  10. Gao F, Xu X, Yu J, Shang M, Li X, Tao D. Complementary, heterogeneous and adversarial networks for image-to-image translation. IEEE Trans Image Process 2021; 30: 3487-98. https://doi.org/10.1109/TIP.2021.3061286
  11. Xia Y, Zhang L, Ravikumar N, Attar R, Piechnik SK, Neubauer S, et al. Recovering from missing data in population imaging - cardiac MR image imputation via conditional generative adversarial nets. Med Image Anal 2021; 67: 101812.
  12. Wang Z, Bovik AC, Sheikh HR, Simoncelli EP. Image quality assessment: from error visibility to structural similarity. IEEE Trans Image Process 2004; 13: 600-12. https://doi.org/10.1109/TIP.2003.819861
  13. Mason A, Rioux J, Clarke SE, Costa A, Schmidt M, Keough V, et al. Comparison of objective image quality metrics to expert radiologists' scoring of diagnostic quality of MR images. IEEE Trans Med Imaging 2020; 39: 1064-72. https://doi.org/10.1109/TMI.2019.2930338
  14. Boulanger M, Nunes JC, Chourak H, Largent A, Tahri S, Acosta O, et al. Deep learning methods to generate synthetic CT from MRI in radiotherapy: a literature review. Phys Med 2021; 89: 265-81. https://doi.org/10.1016/j.ejmp.2021.07.027
  15. Park JW, Song HH, Roh HS, Kim YK, Lee JY. Correlation between clinical diagnosis based on RDC/TMD and MRI findings of TMJ internal derangement. Int J Oral Maxillofac Surg 2012; 41: 103-8. https://doi.org/10.1016/j.ijom.2011.09.010
  16. Goncalves FG, Serai SD, Zuccoli G. Synthetic brain MRI: review of current concepts and future directions. Top Magn Reson Imaging 2018; 27: 387-93. https://doi.org/10.1097/RMR.0000000000000189
  17. Tanenbaum LN, Tsiouris AJ, Johnson AN, Naidich TP, DeLano MC, Melhem ER, et al. Synthetic MRI for clinical neuroimaging: results of the magnetic resonance image compilation (MAGiC) prospective, multicenter, multireader trial. AJNR Am J Neuroradiol 2017; 38: 1103-10. https://doi.org/10.3174/ajnr.A5227
  18. Lee SM, Choi YH, Cheon JE, Kim IO, Cho SH, Kim WH, et al. Image quality at synthetic brain magnetic resonance imaging in children. Pediatr Radiol 2017; 47: 1638-47. https://doi.org/10.1007/s00247-017-3913-y
  19. Lee C, Choi YJ, Jeon KJ, Han SS. Synthetic magnetic resonance imaging for quantitative parameter evaluation of temporomandibular joint disorders. Dentomaxillofac Radiol 2021; 50: 20200584. https://doi.org/10.1259/dmfr.20219002
  20. Benzakoun J, Deslys MA, Legrand L, Hmeydia G, Turc G, Hassen WB, et al. Synthetic FLAIR as a substitute for FLAIR sequence in acute ischemic stroke. Radiology 2022; 303: 153-9. https://doi.org/10.1148/radiol.211394
  21. Kim KH, Do WJ, Park SH. Improving resolution of MR images with an adversarial network incorporating images with different contrast. Med Phys 2018; 45: 3120-31. https://doi.org/10.1002/mp.12945
  22. Yu B, Zhou L, Wang L, Shi Y, Fripp J, Bourgeat P. Ea-GANs: edge-aware generative adversarial networks for cross-modality MR image synthesis. IEEE Trans Med Imaging 2019; 38: 1750-62. https://doi.org/10.1109/tmi.2019.2895894
  23. Kim S, Jang H, Hong S, Hong YS, Bae WC, Kim S, et al. Fat-saturated image generation from multi-contrast MRIs using generative adversarial networks with Bloch equation-based autoencoder regularization. Med Image Anal 2021; 73: 102198.