MRI quantification of myocardial perfusion data

Abstract

Most, if not all, physical systems can be interpreted qualitatively and quantitatively. While a qualitative assessment can produce enough meaningful information about the system, observer bias can sometimes distort the truth and can result in an inaccurate assessment. Typically, myocardial perfusion is assessed clinically in a qualitative manner but through quantitative methods, perfusion can be assessed in a way that is free from observer bias. Myocardial perfusion assessments are typically done with functional imaging modalities such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) and these results are mostly used qualitatively. Quantitative results using magnetic resonance imaging (MRI), have emerged recently as a viable option for myocardial perfusion assessment due to the superior image quality. At present, quantification of myocardial perfusion from MRI is not used extensively outside of the clinical research setting. For routine clinical use, quantification is problematic due to the difficulty and length taken to do the analysis. This thesis focuses on the mathematical methods used in the quantification of myocardial perfusion, specifically model-independent B-spline deconvolution and Fermi-based model-dependent deconvolution. Full automation of the mathematical analysis was also a focus of the study. These methods were applied to a test data set that was obtained from images acquired as a part of a large clinical trial investigating the residual effects of foetal anaemia in healthy adults. Myocardial blood flow quantification showed little difference between the two approaches. However, the investigation into automating these techniques revealed that a model-independent B-spline analysis proved to be more robust and compared well with a manual, model-dependent Fermi analysis, although the former was more susceptible to noise in the image data. While both approaches have their advantages and disadvantages, it is the model-independent B-spline approach that is the more appropriate candidate for future implementations of automated perfusion analysis.