MR signal modelling approaches to characterise tissue microstructure in in-vivo human brain

Abstract

The non-invasive medical imaging technique of magnetic resonance imaging (MRI) is routinely used in diagnosing and monitoring of diseases and disorders. MRI methods within the research setting are currently being developed for the assessment of myelination, demyelination and tissue microstructure variations in the brain. These specific markers have been studied for characterising brain development, and the assessment of neurodegenerative diseases and disorders, such as schizophrenia, autism, Alzheimer’s disease and traumatic brain injury or multiple sclerosis.Previously, various MRI contrast mechanisms associated with spatial variations in proton density, perfusion, diffusion, functional imaging and relaxation have been used to study neurodegenerative diseases and disorders affecting the central nervous system. However, with increasing availability of ultra-high field MRI instruments, the MRI signal magnitude and phase have been shown to provide complimentary information and new insights into soft tissue structure and composition. In particular, signal phase is highly sensitive to changes in tissue magnetic properties, leading to new contrast mechanisms depicting distinct spatial variations in the magnetic properties of tissues. Therefore, modelling of complex signals, i.e. considering signal magnitude and phase, enables development of methods sensitivity to tissue water fraction, relaxation time and tissue magnetic properties observed as an induced change in the scanner magnetic field. Tissue parametric maps can provide succinct information on myelination, demyelination, micro-bleeds, and changes in tissue microstructure.The overall aim of my research was to parameterise the complex valued multiple echo time gradient recalled echo MRI signal using signal compartment modelling within the ultra-high field MRI setting (i.e. 7T MRI). My specific focus was on the estimation of water fraction, relaxation time, and frequency shift (induced field change mapped in Hz) arising from different white matter signal compartments in the human brain.I made three significant advances. First, by applying a three signal compartment white matter model to parameterise the multiple echo time MRI signal, I found myelin water fraction, relaxation time and g-ratio to be similar across corpus callosum sub-regions which project white matter fibre into different cortical regions of the brain. Axonal water fraction was found to be larger around the mid-body of the corpus callosum. Some variations in myelin and axonal frequency shift were demonstrated as well across the corpus callosum of healthy participants. My approach is potentially useful for studies interested in mapping tissue microstructure variations within white matter regions.Second, I analysed a range of multi-compartment models and assessed how tissue parametric maps are affected by the choice of the model and its complexity. I further investigated the number of compartments and parameters necessary to obtain robust tissue parameter and g-ratio measures. I found the choice of model and associated model complexity to influence tissue parameter and g-ratio maps. I also showed that at least a three signal compartment with seven free parameter GRE-MRI signal compartment model is requiredThird, I ported the signal compartmentalisation approach to grey matter regions of the brain. I tested whether tissue parameters (water fraction, relaxation time, and frequency shift) estimated from a three signal compartment model can be used to differentiate normal from dysplastic tissue, which occurs in patients diagnosed with focal cortical dysplasia in focal epilepsy. I found that the frequency shift parameter was most sensitive to tissue microstructure changes due to focal cortical dysplasia. This approach of assessing tissue microstructure differences could be useful in the demarcation of abnormal tissues within the human cerebral cortex.