Combined analysis of Myelin Water Imaging and advanced Diffusion MRI in health and pathology

The ability to visualize the brain in vivo using magnetic resonance imaging (MRI) has been of great importance both in clinical practice and research. Yet imaging is often limited to describing the macrostructure, at spatial resolutions of 1 mm. To unravel the anatomy at a sub-voxel level we are bound to inferring information in an indirect manner. Diffusion MRI (dMRI) quantifies the random motion of water molecules, which contains information on the integrity of cell membranes and axonal organization. In recent years, the link between the diffusion signal and underlying tissue (in particular white matter) has been made by a variety of novel dMRI models. However, through the fractional anisotropy (FA) the basic model diffusion tensor imaging (DTI) only yields very general information on 'tissue integrity'. Furthermore, no diffusion model is able to specifically quantify the myelin content, because diffusion times of water near myelin are simply too short. A different family of MRI modalities, called Myelin Water Imaging (MWI), allows to quantify the amount of water residing in between the myelin bilayers, based on the distinctive short T2 times it exhibits. The myelin water fraction (MWF) obtained through multi-exponential T2 relaxation (MET2) currently is the gold standard for in vivo quantification of myelin content. By combined analysis of MWI and dMRI data, results from both modalities can be better interpreted. The original contribution and main goal of this doctoral research project was to determine if combining MWI and advanced multi-shell dMRI models for use in basic and clinical research was feasible and had added-value. Because MWI was previously not used at KU/UZ Leuven, the first part of this thesis investigates practicalities on the implementation of MWI. Through simulations it was made clear that MWI data requires adequate compensation for stimulated echoes. It was furthermore shown that MWF is underestimated due to necessary regularization of the fitting procedure and that its precision is limited due to noise. Reproducibility and biological variability were assessed and resulted in minimal sample size estimations that meet current ongoing research projects. In the second part of this thesis, we assessed the potential added value of combining MWI and dMRI in health and pathology. Across the lifespan, the white matter undergoes normal changes. By analyzing MWI and dMRI data from 59 healthy volunteers between 17 and 70 years of age, we could confirm that with advancing age diffusion increasingly occurs in an isotropic way. Estimates of MWF, axonal-like diffusion (NDI) and mean kurtosis (MK, 'restrictedness of diffusion') reached an optimum around the age of 50. The theory of retrogenesis, postulating that late-developing white matter degenerates first, was partly confirmed, yet we showed that the early FA decrease in aging could be related to axonal dispersion rather than demyelination. Furthermore, associations between MWI and advanced dMRI metrics were made, clarifying that throughout white matter MWI has more common ground with advanced models such as DKI and NODDI than with basic models such as DTI. To test whether MWI and/or advanced dMRI allow a better characterization of pathological white matter, we applied both modalities in a spectrum of presumed white matter pathologies given different degrees of myelin change: demyelinating lesions, dysmyelinating lesions and suspected white matter damage. With aging, white matter hyperintensities (WMH) are occasionally detected on T2 weighted MRI scans without initial cognitive complaints. This is called leukoaraiosis. As progression of these lesions carries an increased risk of stroke, dementia and depression, follow-up may become clinically important. Using clinical MRI the tissue microstructure of WMH seems uniform, although from histology it is known that periventricular lesions (PVWMH) and deep WM (DWMH) are differently impacted. From scans of 25 older adults aged between 65 - 87 years, we found similar appearance of PVWMH and DWMH in terms of DTI metrics, while MWF detected demyelination in PVWMH and not in DWMH compared to normal appearing white matter (NAWM). This demonstrates that even though DTI metrics are known to be very sensitive for white matter microstructural changes, they are unable to distinguish subtypes that differ on myelination and use of MWF bears a clear added value in assessment of demyelinating lesions. This study also confirmed the decreasing trend of MWF in aging NAWM. Neurofibromatosis type 1 (NF1) is a genetic disorder in which children often present T2 weighted hyperintensities frequently called "unidentified bright objects" (UBOs). Their presence correlates with cognitive dysfunction. Although little evidence from human histology is available, the pathologic substrate of UBOs is believed to be dysmyelination, more specifically intramyelinic edema in absence of demyelination. We compared 30 UBOs and contralateral NAWM (cNAWM) of 17 children. Advanced dMRI showed reduced neurite-like diffusion in UBOs which is possible under conditions of edema, yet also leaves the option for axonal loss. However because water fractions from MWI were not different, demyelination and therefore axonal loss could be considered unlikely, leaving edema as possible hypothesis. Because the amount of isotropic diffusion was similar in UBOs and cNAWM, regular vasogenic edema would however be unlikely, suggesting intramyelinic edema is a plausible hypothesis. This study demonstrated that through combined interpretation of MWI and dMRI metrics, also the in vivo characterization of dysmyelinating lesions can be improved. In a previous study, chemotherapy treated breast cancer patients (C+) presented longitudinal changes in DTI metrics compared to healthy controls (HC) and non-chemotherapy treated patients (C-), suggesting white matter damage. Three to four years later we investigated these same groups using advanced dMRI and MWI in order to elucidate on the type of damage. Both in a voxel-based analysis (VBA) and in previously defined ROI, metric values were compared between 25 C+, 14 C- and 13 HC. No differences were detected in DTI metrics, confirmed by longitudinal normalization. Because of this apparent recovery, it was not possible to use MWI and advanced dMRI for defining the type of WM injury related to chemotherapy treatment. Further longitudinal assessment using VBA before and after chemotherapy treatment is necessary to determine whether MWI and/or advanced dMRI can elucidate on this. This doctoral research project has demonstrated that MWI metrics are most useful in cases where other, more sensitive, metrics have identified white matter changes. In those cases (e.g. demyelination or dysmyelination), combining MWI and dMRI allows hypothesis testing not possible using standard DTI alone. More research is necessary to assess the added value in pathologies involving presumed white matter changes.status: publishe