Synopsis

Keywords: Cardiovascular: Cardiac, Image acquisition: Machine learning, Image acquisition: Image processing

Cardiac motion artefacts affect further downstream analysis, or even render images unusable for clinical diagnosis. This affects clinical workflow in hospitals and may require patient recall, delaying timely diagnosis and treatment starts. Identifying motion corruption, preferably at time of scanning, or applying retrospective motion correction, would help to alleviate these problems. Cardiac motion artefacts can in principle be detected and corrected for at the time of scanning. This paves the way for online quality control as well as active scanning.

AI-enabled CMR imaging - from artefacts to analysis

Cardiac motion artefacts affect further downstream analysis, or even render images unusable for clinical diagnosis. This affects clinical workflow in hospitals and may require patient recall, delaying timely diagnosis and treatment starts. Identifying motion corruption, preferably at time of scanning, or applying retrospective motion correction, would help to alleviate these problems.
Artificial intelligence, in particular from the class of machine learning and deep learning, has shown great promise for application in medical imaging, including cardiac MRI. However, the success of such techniques is limited by the availability and quality of the training data. A common approach is to train methods on well annotated and curated databases of high-quality image acquisitions, which then may fail on real patient cases in a hospital setting where image acquisitions are prone to motion (and other) artefacts.
By using realistic cardiac motion artefact simulations for data augmentation, we can train AI models that can detect and remove cardiac motion artefacts during image reconstruction, which can be improve and even be combined with further downstream tasks. As an outlook, we can apply these principles towards online quality control as well as active scanning preparations.

Acknowledgements

This work was supported by the Wellcome EPSRC Centre for Medical Engineering at Kings College London (WT 203148/Z/16/Z), the EPSRC (EP/P001009/1 and EP/R005516/1) and by the NIHR Cardiovascular MedTech Co-operative. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, EPSRC, or the Department of Health. This research has been conducted using the UK Biobank Resource (application 17806) on a GPU generously donated by NVIDIA Corporation. The UK Biobank data are available for approved projects from https://www.ukbiobank.ac.uk/.