Skip to main content
SpringerLink
Log in

Machine learning for the automatic assessment of aortic rotational flow and wall shear stress from 4D flow cardiac magnetic resonance imaging

  • Magnetic Resonance
  • Published:
European Radiology Aims and scope Submit manuscript

Abstract

Objective

Three-dimensional (3D) time-resolved phase-contrast cardiac magnetic resonance (4D flow CMR) allows for unparalleled quantification of blood velocity. Despite established potential in aortic diseases, the analysis is time-consuming and requires expert knowledge, hindering clinical application. The present research aimed to develop and test a fully automatic machine learning-based pipeline for aortic 4D flow CMR analysis.

Methods

Four hundred and four subjects were prospectively included. Ground-truth to train the algorithms was generated by experts. The cohort was divided into training (323 patients) and testing (81) sets and used to train and test a 3D nnU-Net for segmentation and a Deep Q-Network algorithm for landmark detection. In-plane (IRF) and through-plane (SFRR) rotational flow descriptors and axial and circumferential wall shear stress (WSS) were computed at ten planes covering the ascending aorta and arch.

Results

Automatic aortic segmentation resulted in a median Dice score (DS) of 0.949 and average symmetric surface distance of 0.839 (0.632–1.071) mm, comparable with the state of the art. Aortic landmarks were located with a precision comparable with experts in the sinotubular junction and first and third supra-aortic vessels (p = 0.513, 0.592 and 0.905, respectively) but with lower precision in the pulmonary bifurcation (p = 0.028), resulting in precise localisation of analysis planes. Automatic flow assessment showed excellent (ICC > 0.9) agreement with manual quantification of SFRR and good-to-excellent agreement (ICC > 0.75) in the measurement of IRF and axial and circumferential WSS.

Conclusion

Fully automatic analysis of complex aortic flow dynamics from 4D flow CMR is feasible. Its implementation could foster the clinical use of 4D flow CMR.

Key Points

• 4D flow CMR allows for unparalleled aortic blood flow analysis but requires aortic segmentation and anatomical landmark identification, which are time-consuming, limiting 4D flow CMR widespread use.

• A fully automatic machine learning pipeline for aortic 4D flow CMR analysis was trained with data of 323 patients and tested in 81 patients, ensuring a balanced distribution of aneurysm aetiologies.

• Automatic assessment of complex flow characteristics such as rotational flow and wall shear stress showed good-to-excellent agreement with manual quantification.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Similar content being viewed by others

Notes

  1. https://guiesbibtic.upf.edu/recerca/hpc/system-overview

Abbreviations

4D flow CMR:

Three-dimensional (3D) time-resolved phase-contrast cardiac magnetic resonance imaging

BAV:

Bicuspid aortic valve

CMR:

Cardiac magnetic resonance

DQN:

Deep Q-Network

HV:

Healthy volunteers

IRF:

In-plane rotational flow

ML:

Machine learning

PCMRA:

Phase-contrast-enhanced magnetic resonance angiogram

RL:

Reinforcement learning

SFRR:

Systolic flow reversal ratio

STJ:

Sinotubular junction

TAV:

Tricuspid aortic valve

WSS:

Wall shear stress

References

  1. Guala A, Teixido-Tura G, Dux-Santoy L et al (2019) Decreased rotational flow and circumferential wall shear stress as early markers of descending aorta dilation in Marfan syndrome: a 4D flow CMR study. J Cardiovasc Magn Reson 1:1–11. https://doi.org/10.1186/s12968-019-0572-1

    Article  Google Scholar 

  2. Dux-Santoy L, Guala A, Teixido-Tura G et al (2019) Increased rotational flow in the proximal aortic arch is associated with its dilation in bicuspid aortic valve disease. Eur Heart J Cardiovasc Imaging 20(12):1407–1417

    Article  Google Scholar 

  3. Dux-Santoy L, Guala A, Sotelo J et al (2020) Low and oscillatory wall shear stress is not related to aortic dilation in patients with bicuspid aortic valve. A time-resolved phase-contrast magnetic resonance imaging. Arterioscler Thromb Vasc Biol 40(1):1–11

    Article  Google Scholar 

  4. Guala A, Rodriguez-Palomares J, Galian-Gay L et al (2019) Partial aortic valve leaflet fusion is related to deleterious alteration of proximal aorta hemodynamics. Circulation. 139(23):2707–2709

    Article  Google Scholar 

  5. Bissell MM, Hess AT, Biasiolli L et al (2013 Jul) Aortic dilation in bicuspid aortic valve disease: flow pattern is a major contributor and differs with valve fusion type. Circ Cardiovasc Imaging 6(4):499–507

    Article  Google Scholar 

  6. Guzzardi DG, Barker AJ, Van Ooij P et al (2015) Valve-related hemodynamics mediate human bicuspid aortopathy: insights from wall shear stress mapping. J Am Coll Cardiol 66(8):892–900

    Article  Google Scholar 

  7. Guala A, Rodriguez-Palomares JF, Dux-Santoy L et al (2019) Influence of aortic dilation on the regional aortic stiffness of bicuspid aortic valve assessed by 4-dimensional flow cardiac magnetic resonance. JACC Cardiovasc Imaging 12(6):1020–1029

    Article  Google Scholar 

  8. Ruiz-Muñoz A, Guala A, Rodriguez-Palomares J et al (2022) Aortic flow dynamics and stiffness in Loeys–Dietz syndrome patients: a comparison with healthy volunteers and Marfan syndrome patients. Eur Heart J Cardiovasc Imaging 23(5):641–649

    Article  Google Scholar 

  9. Van Ooij P, Markl M, Collins JD et al (2017) Aortic valve stenosis alters expression of regional aortic wall shear imaging study of 571 subjects. J Am Heart Assoc 6(9):e005959

    Article  Google Scholar 

  10. Soulat G, Scott MB, Allen BD et al (2021) Association of regional wall shear stress and progressive ascending aorta dilation in bicuspid aortic valve. J Am Coll Cardiol Img S1936-878X(21):00510-6

    Google Scholar 

  11. Guala A, Dux-Santoy L, Teixido-Tura G et al (2022) Wall shear stress predicts aortic dilation in patients with bicuspid aortic valve. J Am Coll Cardiol Img 15(1):46–56. https://doi.org/10.1016/j.jcmg.2021.09.023

    Article  Google Scholar 

  12. Loncaric F, Camara O, Piella G, Bijnens B (2021) Integration of artificial intelligence into clinical patient management: focus on cardiac imaging. Rev Esp Cardiol (Engl Ed) 74(1):72–80

  13. Campello M, Gkontra P, Izquierdo C et al (2021) Multi-centre, multi-vendor and multi-disease cardiac segmentation: the M&Ms challenge. IEEE Trans Med Imag 40(12):3543–3554. https://doi.org/10.1109/TMI.2021.3090082

  14. Berhane H, Scott M, Elbaz M et al (2019) Fully automated 3D aortic segmentation of 4D flow MRI for hemodynamic analysis using deep learning. Magn Reson Med 2020:1–15

    Google Scholar 

  15. Ronneberger O, Fischer P, Brox T (2015) U-Net: convolutional networks for biomedical image segmentation. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol 9351, pp 12–20. https://doi.org/10.1007/978-3-319-24574-4_28

  16. Dux-Santoy L, Rodríguez-Palomares JF, Teixidó-Turà G et al (2022) Registration-based semi-automatic assessment of aortic diameter growth rate from contrast-enhanced computed tomography outperforms manual quantification. Eur Radiol 32(3):1997–2009. https://doi.org/10.1007/s00330-021-08273-2

  17. Rodriguez-Palomares J, Dux-Santoy L, Guala A et al (2018) Aortic flow patterns and wall shear stress maps by 4D-flow cardiovascular magnetic resonance in the assessment of aortic dilatation in bicuspid aortic valve disease. J Cardiovasc Magn Reson 20(1):28. https://doi.org/10.1186/s12968-018-0451-1

    Article  PubMed  PubMed Central  Google Scholar 

  18. Gil-Sala D, Guala A, Garcia Reyes ME et al (2021) Geometric, biomechanic and haemodynamic aortic abnormalities assessed by 4D flow cardiovascular magnetic resonance in patients treated by TEVAR following blunt traumatic thoracic aortic injury. Eur J Vasc Endovasc Surg 62(5):797–807. https://doi.org/10.1016/j.ejvs.2021.07.016

  19. Alansary A, Oktay O, Li Y et al (2019) Evaluating reinforcement learning agents for anatomical landmark detection. Med Image Anal 53(Midl):156–164

    Article  Google Scholar 

  20. Mnih V, Kavukcuoglu K, Silver D et al (2015) Human-level control through deep reinforcement learning. Nature. 518(7540):529–533

    Article  CAS  Google Scholar 

  21. Ghesu FC, Georgescu B, Zheng Y et al (2019) Multi-scale deep reinforcement learning for real-time 3D-landmark detection in CT scans. IEEE Trans Pattern Anal Mach Intell 41(1):176–189

    Article  Google Scholar 

  22. Ghesu FC, Georgescu B, Mansi T, Neumann D, Hornegger J, Comaniciu D (2016) An artificial agent for anatomical landmark detection in medical images BT - Medical Image Computing and Computer-Assisted Intervention. MICCAI 2016:229–237

    Google Scholar 

  23. Johnson KM, Lum DP, Turski PA, Block WF, Mistretta CA, Wieben O (2009) Improved 3D phase contrast MRI with off-resonance corrected dual echo VIPR. Magn Reson Med 60(6):1329–1336

    Article  Google Scholar 

  24. Yushkevich PA, Piven J, Hazlett HC et al (2006) User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 31(3):1116–1128

    Article  Google Scholar 

  25. Isensee F, Jaeger PF, Kohl SAA, Petersen J, Maier-Hein KH (2020) nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 18:203–2011

    Article  Google Scholar 

  26. Aviles J, Maso Talou GD, Camara O et al (2021) Domain adaptation for automatic aorta segmentation of 4D flow magnetic resonance imaging data from multiple. In: Functional Imaging and Modeling of the Heart FIMH 2021 Lecture Notes in Comp Sci, pp 112–121. https://doi.org/10.1007/978-3-030-78710-3_12

  27. Bensalah MZ, Bollache E, Kachenoura N et al (2014) Geometry is a major determinant of flow reversal in proximal aorta. Am J Physiol Heart Circ Physiol 306:1408–1416

    Article  Google Scholar 

  28. Hess AT, Bissell MM, Glaze SJ et al (2013) Evaluation of circulation, Γ, as a quantifying metric in 4D flow MRI. J Cardiovasc Magn Reson 15(Suppl 1):E36

    Article  Google Scholar 

  29. Stalder AF, Russe MF, Frydrychowicz A, Bock J, Hennig J, Markl M (2008) Quantitative 2D and 3D phase contrast MRI: optimized analysis of blood flow and vessel wall parameters. Magn Reson Med 60(5):1218–1231

    Article  CAS  Google Scholar 

  30. Morales X, Mill J, Delso G et al (2021) 4D flow magnetic resonance imaging for left atrial haemodynamic characterization and model calibration. Vol. 12592 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Springer International Publishing, vol 12592, pp 156–165. https://doi.org/10.1007/978-3-030-68107-4_16

  31. Guala A, Teixido-Tura G, Rodriguez-Palomares JF et al (2019) Proximal aorta longitudinal strain predicts aortic root dilation rate and aortic events in Marfan syndrome. Eur Heart J 40(25):2047–2055

    Article  Google Scholar 

Download references

Acknowledgements

We are grateful to Hannah Cowdrey for English revision.

Data and models availability statement

The data underlying this article will be shared on reasonable request to the corresponding author. The four trained algorithms for the identification of anatomical landmarks are available at https://github.com/CardiovascularImagingVallHebron/4D_flow_landmark_detection/tree/master/DQN/Models/AORTA.

Funding

This study has been supported by funding from the Instituto de Salud Carlos III (projects PI14/01062, PI17/00381 and PI20/01727), the Spanish Ministry of Science, Innovation and Universities (RTC2019-007280-1 and RTI2018-101193-B-I00), the Spanish Ministry of Economy and Competitiveness (PRE2018-084062), the Spanish Society of Cardiology (SEC/FEC-INV-CLI 20/015 and SEC/FEC-INV-CLI 21/030), the Agency for Management of University and Research Grants of the Generalitat de Catalunya (2020-FI-B-00690) and the Biomedical Research Networking Center on Cardiovascular Diseases (CIBERCV). Guala A. has received funding from Spanish Ministry of Science, Innovation and Universities (IJC2018-037349-I) and from “la Caixa” Foundation (LCF/BQ/PR22/11920008).

Author information

Author notes

  1. Juan Garrido-Oliver, Jordina Aviles, and Marcos Mejía Córdova contributed equally to this work.

Authors and Affiliations

  1. Vall d’Hebron Institute of Research, Barcelona, Spain

    Juan Garrido-Oliver, Lydia Dux-Santoy, Aroa Ruiz-Muñoz, Gisela Teixido-Tura, Arturo Evangelista, Ignacio Ferreira-González, Jose Rodriguez-Palomares & Andrea Guala

  2. Physense, BCN Medtech, Department of Information and Communications Technologies, Universitat Pompeu Fabra, Barcelona, Spain

    Jordina Aviles, Marcos Mejía Córdova, Xabier Morales Ferez, Guillermo Jiménez & Oscar Camara

  3. Department of Cardiology, Hospital Vall d’Hebron Universitat Autonoma de Barcelona, Barcelona, Spain

    Gisela Teixido-Tura, Arturo Evangelista, Ignacio Ferreira-González & Jose Rodriguez-Palomares

  4. CIBER-CV, Instituto de Salud Carlos III, Madrid, Spain

    Gisela Teixido-Tura, Arturo Evangelista, Jose Rodriguez-Palomares & Andrea Guala

  5. Auckland Bioengineering Institute, Auckland, New Zealand

    Gonzalo D. Maso Talou

  6. CIBER-ESP, Instituto de Salud Carlos III, Madrid, Spain

    Ignacio Ferreira-González

  7. Universitat Autonoma de Barcelona, Bellaterra, Spain

    Ignacio Ferreira-González & Jose Rodriguez-Palomares

Authors
  1. Juan Garrido-Oliver
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Jordina Aviles
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Marcos Mejía Córdova
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Lydia Dux-Santoy
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Aroa Ruiz-Muñoz
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Gisela Teixido-Tura
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Gonzalo D. Maso Talou
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Xabier Morales Ferez
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Guillermo Jiménez
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Arturo Evangelista
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Ignacio Ferreira-González
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Jose Rodriguez-Palomares
    View author publications

    You can also search for this author in PubMed Google Scholar

  13. Oscar Camara
    View author publications

    You can also search for this author in PubMed Google Scholar

  14. Andrea Guala
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding authors

Correspondence to Oscar Camara or Andrea Guala.

Ethics declarations

Guarantor

The scientific guarantor of this publication is Andrea Guala.

Conflict of interest

The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.

Statistics and biometry

Several authors have significant statistical expertise.

Informed consent

Written informed consent was obtained from all subjects (patients) in this study.

Ethical approval

Institutional Review Board approval was obtained.

Methodology

• prospective

• cross-sectional study/diagnostic or prognostic study

• performed at one institution

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

ESM 1

(PDF 15 kb)

Rights and permissions

Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Garrido-Oliver, J., Aviles, J., Córdova, M.M. et al. Machine learning for the automatic assessment of aortic rotational flow and wall shear stress from 4D flow cardiac magnetic resonance imaging. Eur Radiol 32, 7117–7127 (2022). https://doi.org/10.1007/s00330-022-09068-9

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00330-022-09068-9

Keywords

Search

Navigation