Multi-site imaging studies are critical in research into high-priority diseases such as dementia, cancer, heart disease, and paediatrics. Despite careful calibration and quality assurance, MRI (magnetic resonance imaging) scanners from different vendors and at different sites produce differing images, and this is a challenge for those studying and imaging tissue microstructure. Over the last 10 years, many methods have been developed for improving imaging, but there is little understanding of their reproducibility across different sites. This is major barrier to the translation of these advanced methods into clinical practice and multi-site studies.
The MRI multi-site reproducibility study assessed the variability of advanced diffusion MRI imaging of the brain in ten healthy volunteers performed on scanners of the same make and model at University College London, Cardiff University and the University of Cambridge. Analysis of the images obtained allows a calculation of the variations between the scanners. With this understanding, the differences between scanners of different makes and models can be investigated.
To support these studies, NPL acquired a comparable MRI system and an advanced test phantom from The University of Manchester, which was also scanned at the different sites. This is the first step toward a standard test object that will allow for advanced calibration of MR facilities and the ability to compare different types of scanner. NPL is working with academic partners at UCL to acquire and analyse the data and produce a detailed report on the reproducibility of advanced diffusion MRI.
This project has resulted in increased experience of MRI multi-site studies and strengthened strategic collaborations within the UK. This has led to a better understanding of how reproducible and reliable tissue images of tissue microstructure are when measured by different teams at different sites. It is also the first deployment of new advanced diffusion phantoms in a study of this kind, and the first to compare phantom and human data of these advanced methods in a multi-site context.
The work aims to provide quantitative estimates of the structure of brain tissue from image data only and have huge potential for diagnosing and monitoring diseases such as cancer and dementia. Translation of these techniques into clinical environments would provide significantly more information to clinicians and patients than is available from conventional scans, improving medical judgements and personalising treatment and prognosis.
Knowledge of reproducibility also impacts clinical trial design. Broader participation in advanced imaging means that existing scanning infrastructure can be used more effectively, and that researchers can be more confident in their conclusions. Obtaining and using more information from each study participant reduces study costs and allows applications in more diseases than is currently possible.