Synopsis

Keywords: Contrast mechanisms: Flow, Cardiovascular: Vascular, Image acquisition: Machine learning

Introduction

• Overview of 4D flow MRI applications and their limitations.
• Motivation for machine learning (ML) techniques to improve the speed and accuracy of 4D flow MRI.

Image reconstruction of accelerated 4D flow MRI

• Challenges with reconstruction algorithms for accelerated 4D flow data and how ML reconstruction approaches can be more efficient.
• FlowVN and 3D U-Net: neural network models for fast, automatic image reconstruction of accelerated 4D flow MRI data.^1,2,3

Super-resolution 4D flow MRI

• Frameworks generating synthetic high-resolution 4D flow MRI images from computational fluid dynamics (CFD) simulations for training super-resolution networks.^4,5,6,7
• 4DFlowNet and SRflow: super-resolution networks used to recover high-resolution 4D flow MRI data.^4,7
• Physics informed neural networks for 4D flow MRI implementations.^8,9
• WSSNet: a machine learning network trained to estimate wall shear stress (WSS) using patient-specific CFD simulations and synthetic 4D flow MRI data.¹⁰

Phase correction methods

• Sources of phase offsets in 4D flow MRI and their impact on velocity maps.
• How machine learning techniques can apply corrections automatically.
• U-Net models used for generating phase error fields for phase correction and correcting aliasing in 4D flow datasets.^11,12

Automatic vessel segmentation in 4D flow MRI data

• The importance of accurate vessel segmentation for calculating mean velocities, flow, and wall shear stress.
• The challenges of low blood-tissue contrast and the need for fast, robust, and automatic delineation.
• 3D U-Net segmentation networks for labelling the aorta and cardiac chambers in a systolic or in multiple timeframes.^13-20
  

Outlook and summary

• ML classification of healthy and cardiovascular disease patients using 4D flow MRI data and haemodynamic features.^21,22
• Outlook and summary on ML-based 4D flow MRI data reconstruction and on automated, operator independent and robust 4D flow MRI data analysis.

Acknowledgements

References

1. Vishnevskiy, V., Walheim, J. & Kozerke, S. Deep variational network for rapid 4D flow MRI reconstruction. Nat Mach Intell 2, 228–235 (2020).

2. Haji-Valizadeh, H. et al. Highly accelerated free-breathing real-time phase contrast cardiovascular MRI via complex-difference deep learning. Magn Reson Med 86, 804–819 (2021).

3. Kim, D, Jen, M-L, Eisenmenger, LB, Johnson, KM. Accelerated 4D-flow MRI with 3-point encoding enabled by machine learning. Magn Reson Med. (2023)

4. Ferdian, E. et al. 4DFlowNet: Super-Resolution 4D Flow MRI Using Deep Learning and Computational Fluid Dynamics. Front Phys 8, (2020).

5. Rutkowski, D. R., Roldan-Alzate, A. & Johnson, K. Enhancement of cerebrovascular 4D flow MRI velocity fields using machine learning and computational fluid dynamics simulation data. Sci Rep (2021).

6. Dirix P., Buoso S., Peper E.S., Kozerke S. Synthesis of patient-specific multipoint 4D flow MRI data of turbulent aortic flow downstream of stenotic valves. Sci Rep. (2022)

7. Shit S, Zimmermann J, Ezhov I, Paetzold JC, Sanches AF, Pirkl C, Menze BH. SRflow: Deep learning based super-resolution of 4D-flow MRI data. Front Artif Intell. 2022

8. Kissas, G. et al. Machine learning in cardiovascular flows modeling: Predicting arterial blood pressure from non-invasive 4D flow MRI data using physics-informed neural networks. Comput Methods Appl Mech Eng 358, 112623 (2020).

9. Fathi, M. F. et al. Super-resolution and denoising of 4D-flow MRI using physics-informed deep neural nets. Comput Methods Programs Biomed 197, 105729 (2020).

10. Ferdian, E., Dubowitz, D. J., Mauger, C. A., Wang, A. & Young, A. A. WSSNet: Aortic Wall Shear Stress Estimation Using Deep Learning on 4D Flow MRI. Front Cardiovasc Med 8, (2022).

11. You, S. et al. Deep Learning Automated Background Phase Error Correction for Abdominopelvic 4D Flow MRI. Radiology 302, 584–592 (2022).

12. Berhane, H. et al. Deep learning–based velocity antialiasing of 4D-flow MRI. Magn Reson Med 88, 449–463 (2022).

13. Bratt, A. et al. Machine learning derived segmentation of phase velocity encoded cardiovascular magnetic resonance for fully automated aortic flow quantification. Journal of Cardiovascular Magnetic Resonance 21, 1–11 (2019).

14. Garcia, J., Beckie, K., Hassanabad, A. F., Sojoudi, A. & White, J. A. Aortic and mitral flow quantification using dynamic valve tracking and machine learning: Prospective study assessing static and dynamic plane repeatability, variability and agreement. JRSM Cardiovasc Dis 10, 2048004021999900 (2021).

15. Tsou, C.-H. et al. Using deep learning convolutional neural networks to automatically perform cerebral aqueduct CSF flow analysis. Journal of Clinical Neuroscience 90, 60–67 (2021).

16. Berhane, H. et al. Fully automated 3D aortic segmentation of 4D flow MRI for hemodynamic analysis using deep learning. Magn Reson Med 84, 2204–2218 (2020).

17. Garrido-Oliver, J. 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 1–11 (2022).

18. Bustamante, M., Viola, F., Engvall, J., Carlhäll, C.-J. & Ebbers, T. Automatic Time-Resolved Cardiovascular Segmentation of 4D Flow MRI Using Deep Learning. Journal of Magnetic Resonance Imaging (2022).

19. Corrado, P. A. et al. Fully automated intracardiac 4D flow MRI post-processing using deep learning for biventricular segmentation. Eur Radiol 1–10 (2022).

20. Corrado, P. A., Seiter, D. P. & Wieben, O. Automatic measurement plane placement for 4D Flow MRI of the great vessels using deep learning. Int J Comput Assist Radiol Surg 17, 199–210 (2022).

21. Niemann, U. et al. Cardiac cohort classification based on morphologic and hemodynamic parameters extracted from 4D PC-MRI data. arXiv preprint arXiv:2010.05612 (2020).

22. Franco, P. et al. Identification of hemodynamic biomarkers for bicuspid aortic valve induced aortic dilation using machine learning. Comput Biol Med 141, 105147 (2022).