Precision measurement of cardiac structure and function in cardiovascular magnetic resonance using machine learning

Rhodri H Davies; João B Augusto; Anish Bhuva; Hui Xue; Thomas A Treibel; Yang Ye; Rebecca K Hughes; Wenjia Bai; Clement Lau; Hunain Shiwani; Marianna Fontana; Rebecca Kozor; Anna Herrey; Luis R Lopes; Viviana Maestrini; Stefania Rosmini; Steffen E Petersen; Peter Kellman; Daniel Rueckert; John P Greenwood; Gabriella Captur; Charlotte Manisty; Erik Schelbert; James C Moon

doi:10.1186/s12968-022-00846-4

Precision measurement of cardiac structure and function in cardiovascular magnetic resonance using machine learning

J Cardiovasc Magn Reson. 2022 Mar 10;24(1):16. doi: 10.1186/s12968-022-00846-4.

Authors

Rhodri H Davies^{1

2

3}, João B Augusto^{1

2}, Anish Bhuva^{1

2}, Hui Xue⁴, Thomas A Treibel^{1

2}, Yang Ye², Rebecca K Hughes^{1

2}, Wenjia Bai⁵, Clement Lau^{2

6}, Hunain Shiwani^{1

2}, Marianna Fontana^{1

7}, Rebecca Kozor⁸, Anna Herrey², Luis R Lopes^{1

2}, Viviana Maestrini⁹, Stefania Rosmini^{1

2}, Steffen E Petersen^{2

6}, Peter Kellman⁴, Daniel Rueckert¹⁰, John P Greenwood¹¹, Gabriella Captur^{1

3}, Charlotte Manisty^{1

2}, Erik Schelbert^{12

13}, James C Moon^{14

15}

Affiliations

¹ Institute of Cardiovascular Science, University College London, London, UK.
² Bart's Heart Centre, St Bartholomew's Hospital, West Smithfield, London, EC1A 7BE, UK.
³ MRC Unit for Lifelong Health and Ageing, University College London, London, UK.
⁴ National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, USA.
⁵ Data Science Institute, Imperial College London, London, UK.
⁶ William Harvey Research Institute, Queen Mary University of London, London, UK.
⁷ National Amyloidosis Centre, University College London, London, UK.
⁸ Sydney Medical School, University of Sydney, Sydney, Australia.
⁹ Department of Clinical, Internal, Anesthesiology and Cardiovascular Sciences, Sapienza University of Rome, Rome, Italy.
¹⁰ Biomedical Image Analysis Group, Department of Computing, Imperial College London, London, UK.
¹¹ Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, UK.
¹² Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, USA.
¹³ Minneapolis Heart Institute East, Saint Paul, MN, USA.
¹⁴ Institute of Cardiovascular Science, University College London, London, UK. j.moon@ucl.ac.uk.
¹⁵ Bart's Heart Centre, St Bartholomew's Hospital, West Smithfield, London, EC1A 7BE, UK. j.moon@ucl.ac.uk.

Abstract

Background: Measurement of cardiac structure and function from images (e.g. volumes, mass and derived parameters such as left ventricular (LV) ejection fraction [LVEF]) guides care for millions. This is best assessed using cardiovascular magnetic resonance (CMR), but image analysis is currently performed by individual clinicians, which introduces error. We sought to develop a machine learning algorithm for volumetric analysis of CMR images with demonstrably better precision than human analysis.

Methods: A fully automated machine learning algorithm was trained on 1923 scans (10 scanner models, 13 institutions, 9 clinical conditions, 60,000 contours) and used to segment the LV blood volume and myocardium. Performance was quantified by measuring precision on an independent multi-site validation dataset with multiple pathologies with n = 109 patients, scanned twice. This dataset was augmented with a further 1277 patients scanned as part of routine clinical care to allow qualitative assessment of generalization ability by identifying mis-segmentations. Machine learning algorithm ('machine') performance was compared to three clinicians ('human') and a commercial tool (cvi42, Circle Cardiovascular Imaging).

Findings: Machine analysis was quicker (20 s per patient) than human (13 min). Overall machine mis-segmentation rate was 1 in 479 images for the combined dataset, occurring mostly in rare pathologies not encountered in training. Without correcting these mis-segmentations, machine analysis had superior precision to three clinicians (e.g. scan-rescan coefficients of variation of human vs machine: LVEF 6.0% vs 4.2%, LV mass 4.8% vs. 3.6%; both P < 0.05), translating to a 46% reduction in required trial sample size using an LVEF endpoint.

Conclusion: We present a fully automated algorithm for measuring LV structure and global systolic function that betters human performance for speed and precision.

Keywords: Cardiac magnetic resonance; Cardiovascular imaging; Image processing; Machine learning; Ventricular function.

Publication types

Research Support, N.I.H., Intramural
Research Support, Non-U.S. Gov't

MeSH terms

Humans
Machine Learning*
Magnetic Resonance Imaging*
Magnetic Resonance Imaging, Cine / methods
Magnetic Resonance Spectroscopy
Predictive Value of Tests
Reproducibility of Results
Stroke Volume
Ventricular Function, Left

Abstract

Publication types

MeSH terms

Grants and funding