Edinburgh Imaging

PhD projects 2017 001

Multicentre evaluation of MRI variability in the quantification of infarct size in experimental focal cerebral ischaemia.


Ischaemic stroke is a leading cause of death and disability in the developed world. Despite that considerable advances in experimental research enabled understanding of the pathophysiology of the disease and identified hundreds of potential neuroprotective drugs for treatment, no such drug has shown efficacy in humans. The failure in the translation from bench to bedside has been partially attributed to the poor quality and rigour of animal studies. Recently, it has been suggested that multicentre animal studies imitating the design of randomised clinical trials could improve the translation of experimental research. Magnetic resonance imaging (MRI) could be pivotal in such studies due to its non-invasive nature and its high sensitivity to ischaemic lesions, but its accuracy and concordance across centres has not yet been evaluated.

This thesis focussed on the use of MRI for the assessment of late infarct size, the primary outcome used in stroke models. Initially, a systematic review revealed that a plethora of imaging protocols and data analysis methods are used for this purpose. Using meta-analysis techniques, it was determined that T2-weighted imaging (T2WI) was best correlated with gold standard histology for the measurement of infarct-based treatment effects. Then, geometric accuracy in six different preclinical MRI scanners was assessed using structural phantoms and automated data analysis tools developed in-house. It was found that geometric accuracy varies between scanners, particularly when centre-specific T2WI protocols are used instead of a standardised protocol, though longitudinal stability over six months is high. Finally, a simulation study suggested that the measured geometric errors and the different protocols are sufficient to render infarct volumes and related group comparisons across centres incomparable. The variability increases when both factors are taken into account and when infarct volume is expressed as a relative estimate. Data in this study were analysed using a custom-made semi-automated tool that was faster and more reliable in repeated analyses than manual analysis.

Findings of this thesis support the implementation of standardised methods for the assessment and optimisation of geometric accuracy in MRI scanners, as well as image acquisition and analysis of in vivo data for the measurement of infarct size in multicentre animal studies. Tools and techniques developed as part of the thesis show great promise in the analysis of phantom and in vivo data and could be a step towards this endeavour.

  • Xenios Milidonis
  • PhD
Link to University of Edinburgh holdings: