Generating synthetic high-resolution spinal STIR and T1w images from T2w FSE and low-resolution axial Dixon.

Databases Deep learning Factual Magnetic resonance imaging Spine

Journal

European radiology
ISSN: 1432-1084
Titre abrégé: Eur Radiol
Pays: Germany
ID NLM: 9114774

Informations de publication

Date de publication:
04 Sep 2024
Historique:
received: 10 05 2024
accepted: 19 08 2024
revised: 28 06 2024
medline: 5 9 2024
pubmed: 5 9 2024
entrez: 4 9 2024
Statut: aheadofprint

Résumé

To generate sagittal T1-weighted fast spin echo (T1w FSE) and short tau inversion recovery (STIR) images from sagittal T2-weighted (T2w) FSE and axial T1w gradient echo Dixon technique (T1w-Dixon) sequences. This retrospective study used three existing datasets: "Study of Health in Pomerania" (SHIP, 3142 subjects, 1.5 Tesla), "German National Cohort" (NAKO, 2000 subjects, 3 Tesla), and an internal dataset (157 patients 1.5/3 Tesla). We generated synthetic sagittal T1w FSE and STIR images from sagittal T2w FSE and low-resolution axial T1w-Dixon sequences based on two successively applied 3D Pix2Pix deep learning models. "Peak signal-to-noise ratio" (PSNR) and "structural similarity index metric" (SSIM) were used to evaluate the generated image quality on an ablations test. A Turing test, where seven radiologists rated 240 images as either natively acquired or generated, was evaluated using misclassification rate and Fleiss kappa interrater agreement. Including axial T1w-Dixon or T1w FSE images resulted in higher image quality in generated T1w FSE (PSNR = 26.942, SSIM = 0.965) and STIR (PSNR = 28.86, SSIM = 0.948) images compared to using only single T2w images as input (PSNR = 23.076/24.677 SSIM = 0.952/0.928). Radiologists had difficulty identifying generated images (misclassification rate: 0.39 ± 0.09 for T1w FSE, 0.42 ± 0.18 for STIR) and showed low interrater agreement on suspicious images (Fleiss kappa: 0.09 for T1w/STIR). Axial T1w-Dixon and sagittal T2w FSE images contain sufficient information to generate sagittal T1w FSE and STIR images. T1w fast spin echo and short tau inversion recovery can be retroactively added to existing datasets, saving MRI time and enabling retrospective analysis, such as evaluating bone marrow pathologies. Sagittal T2-weighted images alone were insufficient for differentiating fat and water and to generate T1-weighted images. Axial T1w Dixon technique, together with a T2-weighted sequence, produced realistic sagittal T1-weighted images. Our approach can be used to retrospectively generate STIR and T1-weighted fast spin echo sequences.

Identifiants

pubmed: 39231829
doi: 10.1007/s00330-024-11047-1
pii: 10.1007/s00330-024-11047-1
doi:

Types de publication

Journal Article

Langues

eng

Sous-ensembles de citation

IM

Subventions

Organisme : HORIZON EUROPE European Research Council
ID : 101045128-iBack-epic-ERC2021-COG

Informations de copyright

© 2024. The Author(s).

Références

Schlaeger S, Drummer K, Husseini ME et al (2023) Implementation of GAN-based, synthetic T2-weighted fat saturated images in the routine radiological workflow improves spinal pathology detection. Diagnostics 13:974
doi: 10.3390/diagnostics13050974 pubmed: 36900118 pmcid: 10000723
Haubold J, Demircioglu A, Theysohn JM et al (2021) Generating virtual short tau inversion recovery (STIR) images from T1-and T2-weighted images using a conditional generative adversarial network in spine imaging. Diagnostics 11:1542
doi: 10.3390/diagnostics11091542 pubmed: 34573884 pmcid: 8467788
Li H, Paetzold JC, Sekuboyina A et al (2019) DiamondGAN: unified multi-modal generative adversarial networks for MRI sequences synthesis. In: Medical Image Computing and Computer Assisted Intervention (MICCAI 2019), 22nd International Conference, Shenzhen, China, 13–17 October 2019, part IV 22. Springer International Publishing, pp 795–803
Finck T, Li H, Grundl L et al (2020) Deep-learning generated synthetic double inversion recovery images improve multiple sclerosis lesion detection. Invest Radiol 55:318–323
doi: 10.1097/RLI.0000000000000640 pubmed: 31977602
Conte GM, Weston AD, Vogelsang DC et al (2021) Generative adversarial networks to synthesize missing T1 and FLAIR MRI sequences for use in a multisequence brain tumor segmentation model. Radiology 299:313–323
doi: 10.1148/radiol.2021203786 pubmed: 33687284
Thomas MF, Kofler F, Grundl L et al (2022) Improving automated glioma segmentation in routine clinical use through artificial intelligence-based replacement of missing sequences with synthetic magnetic resonance imaging scans. Invest Radiol 57:187–193
doi: 10.1097/RLI.0000000000000828 pubmed: 34652289
Lavdas I, Glocker B, Rueckert D et al (2019) Machine learning in whole-body MRI: experiences and challenges from an applied study using multicentre data. Clin Radiol 74:346–356
doi: 10.1016/j.crad.2019.01.012 pubmed: 30803815
Lv J, Zhu J, Yang G (2021) Which GAN? A comparative study of generative adversarial network-based fast MRI reconstruction. Philos Trans A Math Phys Eng Sci 379:20200203
pubmed: 33966462
Xie H, Lei Y, Wang T et al (2022) Synthesizing high-resolution magnetic resonance imaging using parallel cycle-consistent generative adversarial networks for fast magnetic resonance imaging. Med Phys 49:357–369
doi: 10.1002/mp.15380 pubmed: 34821395
Fayad LM, Parekh VS, de Castro Luna R et al (2021) A deep learning system for synthetic knee MRI: is artificial intelligence based fat suppressed imaging feasible? Invest Radiol 56:357
doi: 10.1097/RLI.0000000000000751 pubmed: 33350717 pmcid: 8087629
Kim S, Jang H, Jang J et al (2020) Deep-learned short tau inversion recovery imaging using multi-contrast MR images. Magn Reson Med 84:2994–3008
doi: 10.1002/mrm.28327 pubmed: 32479671
Kim S, Jang H, Hong S et al (2021) Fat-saturated image generation from multi-contrast MRIs using generative adversarial networks with Bloch equation-based autoencoder regularization. Med Image Anal 73:102198
doi: 10.1016/j.media.2021.102198 pubmed: 34403931 pmcid: 9127743
Schlaeger S, Drummer K, Husseini M et al (2021) [241] Synthetic T2-weighted fat sat delivers valuable information for pathology assessment in the spine: validation of a task-specific generative adversarial network. Clin Neuroradiol 31:S54–S54
Kawahara D, Nagata Y (2021) T1-weighted and T2-weighted MRI image synthesis with convolutional generative adversarial networks. Rep Pract Oncol Radiother 26:35–42
doi: 10.5603/RPOR.a2021.0005 pubmed: 33948300 pmcid: 8086713
Bamberg F, Kauczor H-U, Weckbach S et al (2015) Whole-body MR imaging in the German National Cohort: rationale, design, and technical background. Radiology 277:206–220
doi: 10.1148/radiol.2015142272 pubmed: 25989618
John U, Hensel E, Lüdemann J et al (2001) Study of Health In Pomerania (SHIP): a health examination survey in an east German region: objectives and design. Soz Präventivmed 46:186–194
doi: 10.1007/BF01324255 pubmed: 11565448
Völzke H, Alte D, Schmidt CO et al (2011) Cohort profile: the study of health in Pomerania. Int J Epidemiol 40:294–307
doi: 10.1093/ije/dyp394 pubmed: 20167617
Allen N, Sudlow C, Downey P et al (2012) UK Biobank: current status and what it means for epidemiology. Health Policy Technol 1:123–126
doi: 10.1016/j.hlpt.2012.07.003
Schlaeger S, Klupp E, Weidlich D et al (2018) T2-weighted Dixon turbo spin echo for accelerated simultaneous grading of whole-body skeletal muscle fat infiltration and edema in patients with neuromuscular diseases. J Comput Assist Tomogr 42:574–579
doi: 10.1097/RCT.0000000000000723 pubmed: 29613984
Brandão S, Seixas D, Ayres-Basto M et al (2013) Comparing T1-weighted and T2-weighted three-point Dixon technique with conventional T1-weighted fat-saturation and short-tau inversion recovery (STIR) techniques for the study of the lumbar spine in a short-bore MRI machine. Clin Radiol 68:e617–e623
doi: 10.1016/j.crad.2013.06.004 pubmed: 23932678
Delfaut EM, Beltran J, Johnson G et al (1999) Fat suppression in MR imaging: techniques and pitfalls. Radiographics 19:373–382
doi: 10.1148/radiographics.19.2.g99mr03373 pubmed: 10194785
Colosimo C, Gaudino S, Alexandre AM (2011) Imaging in degenerative spine pathology. Springer, Vienna, pp 9–15
Wang B, Fintelmann FJ, Kamath RS et al (2016) Limited magnetic resonance imaging of the lumbar spine has high sensitivity for detection of acute fractures, infection, and malignancy. Skelet Radiol 45:1687–1693
doi: 10.1007/s00256-016-2493-5
Alcaide-Leon P, Pauranik A, Alshafai L et al (2016) Comparison of sagittal FSE T2, STIR, and T1-weighted phase-sensitive inversion recovery in the detection of spinal cord lesions in MS at 3 T. AJNR Am J Neuroradiol 37:970–975
doi: 10.3174/ajnr.A4656 pubmed: 26797141 pmcid: 7960295
Dagestad MH, Toppe MK, Kristoffersen PM et al (2024) Dixon T2 imaging of vertebral bone edema: reliability and comparison with short tau inversion recovery. Acta Radiol 65:273–283
Kristoffersen PM, Vetti N, Storheim K et al (2020) Short tau inversion recovery MRI of Modic changes: a reliability study. Acta Radiol Open 9:2058460120902402
pubmed: 32064122 pmcid: 6990614
Vanel D, Dromain C, Tardivon A (2000) MRI of bone marrow disorders. Eur Radiol 10:224–229
doi: 10.1007/s003300050038 pubmed: 10663751
Brett M, Taylor J, Burns C et al (2009) NIPY: an open library and development framework for FMRI data analysis. NeuroImage 47:S196
doi: 10.1016/S1053-8119(09)72223-2
Isola P, Zhu J-Y, Zhou T, Efros AA (2017) Image-to-image translation with conditional adversarial networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, 2017. pp 1125–1134
Ronneberger O, Fischer P, Brox T (2015) U-net: convolutional networks for biomedical image segmentation. In: Medical Image Computing and Computer-Assisted Intervention (MICCAI 2015), 18th International Conference, Munich, Germany, Springer International Publishing, pp 234–241
Mao X, Li Q, Xie H et al (2017) Least squares generative adversarial networks. In: Proceedings of the IEEE International Conference on Computer Vision, Venice. pp 2794–2802
Bieder F, Wolleb J, Durrer A et al (2023) Memory-efficient 3D denoising diffusion models for medical image processing. In: Medical Imaging with Deep Learning (MIDL 2023), Proceedings of Machine Learning Research, Nashville, Tennessee, PMLR, pp 552–567
Zhu J-Y, Park T, Isola P, Efros AA (2017) Unpaired image-to-image translation using cycle-consistent adversarial networks. In: Proceedings of the IEEE International Conference on Computer Vision, Venice. pp 2223–2232
Cohen J (1960) A coefficient of agreement for nominal scales. Educ Psychol Meas 20:37–46
doi: 10.1177/001316446002000104
Ouyang H, Meng F, Liu J et al (2022) Evaluation of deep learning-based automated detection of primary spine tumors on MRI using the Turing test. Front Oncol 12:814667
doi: 10.3389/fonc.2022.814667 pubmed: 35359400 pmcid: 8962659

Auteurs

Robert Graf (R)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany. robert.graf@tum.de.
Institut für KI und Informatik in der Medizin, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany. robert.graf@tum.de.

Paul-Sören Platzek (PS)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Evamaria Olga Riedel (EO)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Su Hwan Kim (SH)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Nicolas Lenhart (N)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Constanze Ramschütz (C)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Karolin Johanna Paprottka (KJ)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Olivia Ruriko Kertels (OR)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Hendrik Kristian Möller (HK)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.
Institut für KI und Informatik in der Medizin, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany.

Matan Atad (M)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.
Institut für KI und Informatik in der Medizin, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany.

Robin Bülow (R)

Institute for Diagnostic Radiology and Neuroradiology, University Medicine Greifswald, Greifswald, Germany.

Nicole Werner (N)

Institut für Community Medicine, Abteilung SHIP-KEF, University Medicine Greifswald, Greifswald, Germany.

Henry Völzke (H)

Institut für Community Medicine, Abteilung SHIP-KEF, University Medicine Greifswald, Greifswald, Germany.

Carsten Oliver Schmidt (CO)

Institut für Community Medicine, Abteilung SHIP-KEF, University Medicine Greifswald, Greifswald, Germany.

Benedikt Wiestler (B)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Johannes C Paetzold (JC)

Professor of Visual Information Processing, Department of Computing, Imperial College London, London, United Kingdom.

Daniel Rueckert (D)

Institut für KI und Informatik in der Medizin, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany.
Professor of Visual Information Processing, Department of Computing, Imperial College London, London, United Kingdom.

Jan Stefan Kirschke (JS)

Department of Diagnostic and Interventional Neuroradiology, School of Medicine, Technical University of Munich, Munich, Germany.

Classifications MeSH