Repeatability of Cardiac Magnetic Resonance Radiomics: A Multi-Centre Multi-Vendor Test-Retest Study
3 December 2020
A new collaborative study involving Professor Steffen Petersen, our member and Queen Mary Turing Fellow, and researchers at Queen Mary University of London investigates a new image analysis toolkit for MRI scans.
In this Q&A, Dr Zahra Raisi-Estabragh explains how Radiomics analysis can help us to better diagnose heart disease.
What is new about the study?
The way we currently image the heart in hospitals does not fully reflect the complexity of the heart's architecture. To truly appreciate how the heart changes with disease, we need to be able to see the heart in much greater detail. At the William Harvey Research Institute, Queen Mary University of London, and in close collaboration with colleagues from the University of Barcelona and the MRC Lifecourse Epidemiology Unit, University of Southampton, we are investigating a new image analysis toolkit, called radiomics, which can be applied to heart magnetic resonance imaging (MRI) scans. Radiomics analysis can extract and express highly detailed information about the shape and texture of the heart. In some early studies, we can see that this level of detail can help us diagnose heart conditions more quickly and more accurately. However, if we are to develop this methodology for use in hospitals, we need to be sure that the technique itself is stable and precise- that is, when radiomics analysis is applied to scans of the same person, the shape and texture measures generated will be the same (or very similar) unless there is a change in the health status of that person. This is a key requirement if the method is to be useful as a tool for patients.
How did you carry out the study?
We made use of the VOLUMES resource coordinated by colleagues from University College London, to access images from patients who had two heart MRI scans in short succession. We applied our radiomics toolkit to these images and compared the shape and texture metrics between paired scans. This is the first study to systematically examine the technical performance of radiomics analysis in this way and a key step in developing the methodology as a new image analysis tool.
Is there anything surprising about the results?
Our findings demonstrate that whilst most radiomics metrics are stable, a minority demonstrate poor precision. This is important information, because it can guide how much importance we give to different metrics generated by the toolkit. For example, a radiomics measure that we know is very imprecise is less reliable than one that has demonstrable stability. In this study, we document for the first time, these properties for all radiomics metrics from heart MRI scans.
Why is the study important?
It is very important that new technologies undergo proper scientific scrutiny before they are introduced in hospitals. This study is a key step for the development of radiomics and provides essential information about the technical performance of the methodology.
What are the wider implications?
This study will be a valuable resource in guiding future researchers to select the most robust metrics from the toolkit for more detailed study. This work is part of a wider research strategy to develop heart MRI radiomics as a tool to improve patient care. Our ultimate goal is to use our imaging 'toolkit' to allow faster and more accurate diagnosis of heart disease, improve our estimations of future risk of heart conditions, and better understand the processes underlying cardiovascular disease.
More information can be found here: Raisi-Estabragh Zahra, Gkontra Polyxeni, Jaggi Akshay, Cooper Jackie, Augusto João, Bhuva Anish N., Davies Rhodri H., Manisty Charlotte H., Moon James C., Munroe Patricia B., Harvey Nicholas C., Lekadir Karim, Petersen Steffen E. Repeatability of Cardiac Magnetic Resonance Radiomics: A Multi-Centre Multi-Vendor Test-Retest Study. J Front. Cardiovascular. Med 7, 289 (2020). https://doi.org/10.3389/fcvm.2020.586236
Updated by: Michal Filus