Electronic Thesis and Dissertation Repository


Master of Science


Medical Biophysics


Dr Keith St Lawrence


The ability of the perfusion MRI technique, arterial spin labeling (ASL), to quantify cerebral blood flow (CBF) makes it attractive for longitudinal studies of changes in brain function, such as those related to chronic pain. However, ASL's poor spatial resolution makes image alignment between sessions difficult, leading to increased variance and greater Type-I errors. In addition, variability due to differences in basal blood flow between sessions and confounding effects such as the arterial transit time (ATT) have the potential to reduce reproducibility over time. The focus of this thesis is to investigate the ability of ASL to detect long-term changes in regional CBF within an individual on a voxel-wise level. It is hypothesized that ASL has the sensitivity to detect activation-induced CBF changes over periods as long as a month if the sources of variance that degrade between-session comparisons are minimized. To test this hypothesis rest and activation (motor task) CBF images were acquired from healthy subjects on three separate imaging sessions. Registration errors were minimized by using individual head molds to replicate the head position in successive sessions. Variations in resting CBF were controlled for by performing the imaging during the same time of day, and subjects were asked to refrain from using common substances, such as caffeine, that are known to affect CBF. Finally, ATT maps were generated on each session to investigate its stability. From these data sets, the within- and between-session variability in CBF was determined and motor-related activation maps were generated from rest and activation data acquired on from the same session and from sessions separated by a week and a month. The results demonstrated excellent reliability (intraclass correlation coefficients greater than 0.75) both within- (0.89 ± 0.2) and between-session (0.84 ± 0.15), and high reproducibility (within subject coefficient of variation, wsCV, greater than 20%) within- (wsCV = 4.7 ± 4.5%) and between-session (wsCV = 5.7 ± 4.4%). Between-session reproducibility of the ATT was high (wsCV = 5.0 ± 2.7%), suggesting that the confounding effect of ATT over a month was minimal. The similarity in within- and between-session variability and their activation maps indicated that registration errors between sessions were minimal. Measures of precision of activation demonstrated that less than ~20% of between-session activation were false positives. These results demonstrate the feasibility of conducting voxel-wise analysis of CBF images acquired on different days and highlight the potential of this technique for longitudinal studies.