Multi-parameter ultrasound imaging for musculoskeletal tissues based on a physics informed generative adversarial network
Pengxin Wang(王鹏鑫)1, Heyu Ma(马贺雨)1, Tianyu Liu(刘天宇)1, Chengcheng Liu(刘成成)1,2,†, Dan Li(李旦)3, and Dean Ta(他得安)2,4
1 Institute of Biomedical Engineering & Technology, Academy for Engineering and Technology, Fudan University, Shanghai 200433, China; 2 State Key Laboratory of Integrated Chips and Systems, Fudan University, Shanghai 201203, China; 3 Department of Electronic Engineering, School of Information Science and Technology, Fudan University, Shanghai 200438, China; 4 Department of Biomedical Engineering, School of Information Science and Technology, Fudan University, Shanghai 200438, China
Abstract Full waveform inversion (FWI) has showed great potential in the detection of musculoskeletal disease. However, FWI is an ill-posed inverse problem and has a high requirement on the initial model during the imaging process. An inaccurate initial model may lead to local minima in the inversion and unexpected imaging results caused by cycle-skipping phenomenon. Deep learning methods have been applied in musculoskeletal imaging, but need a large amount of data for training. Inspired by work related to generative adversarial networks with physical informed constrain, we proposed a method named as bone ultrasound imaging with physics informed generative adversarial network (BUIPIGAN) to achieve unsupervised multi-parameter imaging for musculoskeletal tissues, focusing on speed of sound (SOS) and density. In the in-silico experiments using a ring array transducer, conventional FWI methods and BUIPIGAN were employed for multi-parameter imaging of two musculoskeletal tissue models. The results were evaluated based on visual appearance, structural similarity index measure (SSIM), signal-to-noise ratio (SNR), and relative error (RE). For SOS imaging of the tibia-fibula model, the proposed BUIPIGAN achieved accurate SOS imaging with best performance. The specific quantitative metrics for SOS imaging were SSIM 0.9573, SNR 28.70 dB, and RE 5.78%. For the multi-parameter imaging of the tibia-fibula and human forearm, the BUIPIGAN successfully reconstructed SOS and density distributions with SSIM above 94%, SNR above 21 dB, and RE below 10%. The BUIPIGAN also showed robustness across various noise levels (i.e., 30 dB, 10 dB). The results demonstrated that the proposed BUIPIGAN can achieve high-accuracy SOS and density imaging, proving its potential for applications in musculoskeletal ultrasound imaging.
Fund: Project supported by the National Natural Science Foundation of China (Grant Nos. 12122403 and 12327807).
Corresponding Authors:
Chengcheng Liu
E-mail: chengchengliu@fudan.edu.cn
Cite this article:
Pengxin Wang(王鹏鑫), Heyu Ma(马贺雨), Tianyu Liu(刘天宇), Chengcheng Liu(刘成成), Dan Li(李旦), and Dean Ta(他得安) Multi-parameter ultrasound imaging for musculoskeletal tissues based on a physics informed generative adversarial network 2025 Chin. Phys. B 34 044301
[1] Su N, Yang J, Xie Y, Du X, Chen H, Zhou H and Chen L 2019 International Journal of Biological Sciences 15 776 [2] Mirza F and Canalis E 2015 European Journal of Endocrinology 173 R131 [3] Leslie W D and Morin S N 2014 Current Opinion in Rheumatology 26 440 [4] Compston J, McClung M and Leslie W 2019 The Lancet 393 364 [5] Cauley J 2013 The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 68 1243 [6] Si L,Winzenberg T, Jiang Q, ChenMand Palmer A 2015 Osteoporosis International 26 1929 [7] Organization W H 1997 The world health report 1997: conquering suffering; enriching humanity pp. 162-162 [8] Guglielmi G, Grimston S, Fischer K and Pacifici R 1994 Radiology 192 845 [9] Liu C, Li B, Diwu Q, Li Y, Zhang R, Ta D and Wang W 2018 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 65 2311 [10] Wang F, Zheng L, Theopold J, Schleifenbaum S, Heyde C E and Osterhoff G 2022 Journal of Orthopaedic Surgery and Research 17 174 [11] Huddleston A,Wishko D, Dunn G, Metcalf D, Craven J, Stern R, Friedman R and GreenfieldMA 1974 Physics in Medicine & Biology 19 251 [12] Li B, Liu C, Liu X, Tran T, Li Y, Li D, Bi D, Liu D and Ta D 2022 Chin. Phys. B 31 114303 [13] Garcia D, Le Tarnec L, Muth S, Montagnon E, Poree J and Cloutier G 2013 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 60 1853 [14] Jensen J, Nikolov S, Gammelmark K and Pedersen M 2006 Ultrasonics 44 e5 [15] Rao J, RatasseppMand Fan Z 2016 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 63 737 [16] Zheng R, Le L, Sacchi M and Lou E 2015 Ultrasound in Medicine & Biology 41 2955 [17] Guillermin R, Lasaygues P, Rabau G and Lefebvre J P 2013 The Journal of the Acoustical Society of America 134 1001 [18] Pérez-Liva M, Herraiz J, Udías J, Miller E, Cox B and Treeby B 2017 The Journal of the Acoustical Society of America 141 1595 [19] Suo M, Zhang D, Yang H and Yang Y 2023 Computer Methods and Programs in Biomedicine 231 107404 [20] Wu Y, Yan W, Liu Z, Zhang Q, Zhou L, Song J, Qiu W, Ding M and Yuchi M 2024 Physics in Medicine & Biology 69 105024 [21] Li Y, Wang J, Su C, Lin W, Wang X and Luo Y 2023 Chin. Phys. B 32 014303 [22] Da Silva S, Karsou A, de Souza A, Capuzzo F, Costa F, Moreira R and Cetale M 2022 Geophysical Journal International 231 1363 [23] Zhang W, Luo J and Teng J 2015 Chinese Journal of Geophysics 58 216 [24] Dong L, Chi B, Tao J and Liu Y 2013 Chinese Journal of Geophysics 56 685 [25] Yang J, Liu Y and Dong L 2014 Chinese Journal of Geophysics 57 628 [26] Liu T, Han S, Xie L, Xing W, Liu C, Li B and Ta D 2024 Physics in Medicine & Biology 69 125026 [27] Shi Q, Zhou T, Li Y, Liu C and Ta D 2022 IEEE Transactions on Computational Imaging 8 1063 [28] Suo M, Zhang D, Yang H and Yang Y 2023 Neural Computing and Applications 35 25027 [29] Paszke A, Gross S, Massa F, et al. 2019 Advances in Neural Information Processing Systems 2019 32 [30] Pang B, Nijkamp E and Wu Y 2020 Journal of Educational and Behavioral Statistics 45 227 [31] Karniadakis G, Kevrekidis I, Lu L, Perdikaris P, Wang S and, Yang L 2021 Nat. Rev. Phys. 3 422 [32] Raissi M, Perdikaris P and Karniadakis G 2019 Journal of Computational Physics 378 686 [33] Guo J, Wang H, Gu S and Hou C 2024 Chin. Phys. B 33 050701 [34] Moseley B, Markham A and Nissen-Meyer T 2020 arXiv: 2006.11894 [35] Shukla K, Di Leoni P, Blackshire J, Sparkman D and Karniadakis G 2020 Journal of Nondestructive Evaluation 39 61 [36] Gupta H, McCann M, Donati L and Unser M 2021 IEEE Transactions on Computational Imaging 7 759 [37] Rasht-Behesht M, Huber C, Shukla K and Karniadakis G 2022 Journal of Geophysical Research: Solid Earth 127 e2021JB023120 [38] Yang F and Ma J 2023 Journal of Geophysical Research: Solid Earth 128 e2022JB025493 [39] Hu W, Chen J, Liu J and Abubakar A 2018 IEEE Signal Processing Magazine 35 132 [40] Gui J, Sun Z, Wen Y, Tao D and Ye J 2021 IEEE Transactions on Knowledge and Data Engineering 35 3313 [41] Salimans T, Goodfellow I, ZarembaW, Cheung V, Radford A and Chen X 2016 Advances in Neural Information Processing Systems 29 [42] Arjovsky M, Chintala S and Bottou L 2017 Proceedings of the 34th International Conference on Machine Learning, 2017, Sydney, Australia, p. 214 [43] Gulrajani I, Ahmed F, Arjovsky M, Dumoulin V and Courville A 2017 Advances in Neural Information Processing Systems 30 [44] Guitton A and Alkhalifah T 2017 Geophysics 82 R299 [45] Alexey D 2020 arXiv: 2010.11929 [46] Chollet F 2017 Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1251-1258 [47] He K, Zhang X, Ren S and Sun J 2016 Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778 [48] Culjat M, Goldenberg D, Tewari P and Singh R 2010 Ultrasound in medicine & biology 36 861 [49] Li Y, Shi Q, Li Y, Song X, Liu C, Ta D and Wang W 2021 Chin. Phys. B 30 014302
Altmetric calculates a score based on the online attention an article receives. Each coloured thread in the circle represents a different type of online attention. The number in the centre is the Altmetric score. Social media and mainstream news media are the main sources that calculate the score. Reference managers such as Mendeley are also tracked but do not contribute to the score. Older articles often score higher because they have had more time to get noticed. To account for this, Altmetric has included the context data for other articles of a similar age.
