Magnetic Resonance Imaging (MRI) is a widely used, non-invasive imaging technique that provides a means to reveal structural and functional information of different body tissues in detail. Susceptibility Weighted Imaging (SWI) is a field in MRI that utilizes the information from the magnetic susceptibility property of different tissues using the gradient echo phase information. Although longer echo times (TEs) have been widely used in applications involving SWI, there are a few problems related with the long TE data, such as the strong blooming effect and phase aliasing even at macroscopic levels. In this thesis, the use of very short TEs is proposed to study susceptibility mapping. The short TEs can be used to study structures with susceptibilities an order of magnitude larger (such as air and bones in and around the brain sinuses, skull and teeth) than those within soft tissue. Using a new iterative susceptibility mapping technique that we recently developed, it becomes possible to map the geometry of such structures, which to date has proven difficult due to the lack of water content (for sinuses) or due to very short T2* (for bones). The method of phase replacement inside the sinuses proposed in this thesis provides more accurate phase information for the inversion than assuming zero or some arbitrary constant inside these structures. The first and second iterations were responsible for most of the changes in mapping out the susceptibility values. The mean susceptibility value in the sphenoid sinus is calculated as +9.3 ± 1.1ppm, close to the expected value of +9.4ppm for air. The reconstruction of the teeth in the in-vivo data provides a mean Δχ(teeth-tissue)=–3.3ppm, thanks to the preserved phase inside the jaw. The mean susceptibility inside a relatively homogeneous region of the skull bone was measured to be Δχ(bone-tissue)=–2.1ppm. Finally, these susceptibilities can be used to help remove the unwanted background fields prior to applying either SHARP or HPF. In addition, the effects of the background field gradient on flow compensation are studied. Due to the presence of these background gradients, an unwanted phase term is induced by the blood flow inside the vessels. Using a 3D numerical model and in vivo data, the background gradients were estimated to be as large as 1.5mT/m close to the air-tissue interfaces and 0.7mT/m inside the brain (leading to a potential signal loss of up to 15%). The quantitative susceptibility mapping (QSM) results were improved in the entire image after removing the confounding arterial phase thanks to the reduced ringing artifacts. Lastly, a novel approach to improve the susceptibility mapping results was introduced and utilized to monitor the changes in venous oxygen saturation levels as well as the changes in oxygen extraction fraction instigated by the vasodynamic agents, caffeine and acetazolamide. The internal streaking artifacts in the susceptibility maps were reduced by giving an initial susceptibility value uniformly to the structure-of-interest, based on the a priori information. For veins, the iterative results, when the initial value of 0.45 ppm was used, were the best in terms of the highest accuracy in the mean susceptibility value (0.453 ppm) and the lowest standard deviation (0.013 ppm). Using this technique, the venous oxygen saturation levels (inside the internal cerebral veins (ICVs)) for normal physiological conditions, post-caffeine and post-Diamox for the first volunteer were calculated as (mean ± standard deviation): Y_Normal = 69.1 ± 3.3 %, Y_Caffeine = 60.5 ± 2.8 % and Y_Diamox = 79.1 ± 4.0%. For the caffeine challenge, the percentage change in oxygen extraction fraction (OEF) for pre and post caffeine results was calculated as +27.0 ± 3.8%; and for the Diamox challenge, the percentage change in OEF was calculated as −32.6 ± 2.1 % for the ICVs. These vascular effects of Diamox and caffeine were large enough to be easily measured with susceptibility mapping and can serve as a sensitive biomarker for measuring reductions in cerebro-vascular reserve through abnormal vascular response, an increase in oxygen consumption during reperfusion hyperoxia or locally varying oxygen saturation levels in regions surrounding damaged tissue. In conclusion, our new approach to QSM offers a means to monitor venous oxygen saturation reasonably accurately and may provide a new means to study neurovascular diseases where there are changes in perfusion that affect the oxygen extraction fraction. / Thesis / Doctor of Philosophy (PhD) / Magnetic Resonance Imaging (MRI) is a widely used, non-invasive imaging technique that provides a means to reveal structural and functional information of different body tissues in detail. Susceptibility Weighted Imaging (SWI) is a field in MRI that utilizes the information from the magnetic susceptibility property of different tissues using the gradient echo phase information. Firstly, we demonstrate that using our phase replacement technique, it becomes possible to map the geometry of structures with almost no MR signal, which to date has proven difficult due to the lack of water content (for sinuses) or due to very short T2* (for bones). Secondly, the effects of the background field gradient on flow compensation were studied. Due to the presence of these background gradients, an unwanted phase term is induced by the blood flow inside the vessels. And, lastly, we present our new approach utilizing SWI data, offering a means to monitor venous oxygen saturation reasonably accurately and, potentially, a new means to study neurovascular diseases where there are changes in perfusion that affect the oxygen extraction fraction.

The BOLD contrast employed in functional MRI studies is an ambiguous signal composed of changes in blood flow, blood volume and oxidative metabolism. In situations where the vasculature and metabolism may have been affected, such as in aging and in certain diseases, the dissociation of the more physiologically-specific components from the BOLD signal becomes crucial. The latest generation of calibrated functional MRI methods allows the estimation of both resting blood flow and absolute oxygen metabolism. The work presented here is based on one such proof-of-concept approach, dubbed QUO2, whereby taking into account, within a generalized model, both arbitrary changes in blood flow and blood O2 content during a combination of hypercapnia and hyperoxia breathing manipulations, yields voxel-wise estimates of resting oxygen extraction fraction and oxidative metabolism. In the first part of this thesis, the QUO2 acquisition protocol and data analysis were revisited in order to enhance the temporal stability of individual blood flow and BOLD responses, consequently improving reliability of the model-derived estimates. Thereafter, an assessment of the within and between-subject variability of the optimized QUO2 measurements was performed on a group of healthy volunteers. In parallel, an analysis was performed of the sensitivity of the model to different sources of random and systematic errors, respectively due to errors in measurements and choice of assumed parameters values. Moreover, the various impacts of the oxygen concentration administered during the hyperoxia manipulation were evaluated through a simulation and experimentally, indicating that a mild hyperoxia was beneficial. Finally, the influence of Alzheimer’s disease in vascular and metabolic changes was explored for the first time by applying the QUO2 approach in a cohort of probable Alzheimer’s disease patients and age-matched control group. Voxel-wise and region-wise differences in resting blood flow, oxygen extraction fraction, oxidative metabolism, transverse relaxation rate constant R2* and R2* changes during hypercapnia were identified. A series of limitations along with recommended solutions was given with regards to the delayed transit time, the susceptibility artifacts and the challenge of performing a hypercapnia manipulation in cohorts of elderly and Alzheimer’s patients. / Le contraste BOLD employé dans les études d’imagerie par résonance magnétique fonctionnelle (IRMf) provient d’une combinaison ambigüe de changements du flux sanguin cérébral, du volume sanguin ainsi que du métabolisme oxydatif. Dans un contexte où les fonctions vasculaires ou métaboliques du cerveau ont pu être affectées, tel qu’avec l’âge ou certaines maladies, il est crucial d’effectuer une décomposition du signal BOLD en composantes physiologiquement plus spécifiques. La dernière génération de méthodes d’IRMf calibrée permet d’estimer à la fois le flux sanguin cérébral et le métabolisme oxydatif au repos. Le présent travail est basé sur une telle technique, appelée QUantitative O2 (QUO2), qui, via un model généralisé, prend en considération les changements du flux sanguin ainsi que ceux en concentrations sanguine d’O2 durant des périodes d’hypercapnie et d’hyperoxie, afin d’estimer, à chaque voxel, la fraction d’extraction d’oxygène et le métabolisme oxydatif au repos. Dans la première partie de cette thèse, le protocole d’acquisition ainsi que la stratégie d’analyse de l’approche QUO2 ont été revus afin d’améliorer la stabilité temporelle des réponses BOLD et du flux sanguin, conséquemment, afin d’accroître la fiabilité des paramètres estimés. Par la suite, une évaluation de la variabilité intra- et inter-sujet des différentes mesures QUO2 a été effectuée auprès d’un groupe de participants sains. En parallèle, une analyse de la sensibilité du model à différentes sources d’erreurs aléatoires (issues des mesures acquises) et systématiques (dues aux assomptions du model) a été réalisée. De plus, les impacts du niveau d’oxygène administré durant les périodes d’hyperoxie ont été évalués via une simulation puis expérimentalement, indiquant qu’une hyperoxie moyenne était bénéfique. Finalement, l’influence de la maladie d’Alzheimer sur les changements vasculaires et métaboliques a été explorée pour la première fois en appliquant le protocole QUO2 à une cohorte de patients Alzheimer et à un groupe témoin du même âge. Des différences en terme de flux sanguin, fraction d’oxygène extraite, métabolisme oxydatif, et taux de relaxation transverse R2* au repos comme en réponse à l’hypercapnie, ont été identifiées au niveau du voxel, ainsi qu’au niveau de régions cérébrales vulnérables à la maladie d’Alzheimer. Une liste de limitations accompagnées de recommandations a été dressée en ce qui a trait au temps de transit différé, aux artéfacts de susceptibilité magnétique, de même qu’au défi que représente l’hypercapnie chez les personnes âgées ou atteintes de la maladie d’Alzheimer.