Myelin facilitates efficient neural signal transmission along axons. While predominantly present in white matter, myelin is also found in smaller amounts in the outermost layer of gray matter, known as the cortex, as well as deep subcortical structures. Quantitative magnetic resonance imaging (qMRI) metrics, including longitudinal relaxation rate (R1), can report myelin levels in vivo, although these measures are often estimated and can vary depending on pulse sequences and acquisition parameters used to collect the data. Multi-site approaches can increase the statistical power and ecological validity of studies by recruiting participants from larger and more diverse samples, but considerable variability has previously been reported in qMRI estimates of myelin computed using data from different MRI scanners. Here, we assessed intra- and inter-vendor variability in R1 maps acquired at three sites using scanners from two manufacturers, General Electric (GE) and Siemens. Two ‘traveling heads’ scans, where a participant visits and completes a scan at each site, were acquired to directly assess site effects independent of biological variability. Histogram-based site-specific scaling factors determined from the ‘traveling heads’ scans were then applied to harmonize the main dataset. Before site correction, comparable inter- and intra-vendor differences were observed in both the main dataset and the ‘traveling heads’ scans. Our method of harmonization reduced the inter-site variability considerably, while remaining sensitive to age effects in R1. We then characterized age trajectories of R1 across the whole brain using linear regressions. In the age range of 16-36, age was significantly associated with mean R1s of bilateral putamen, left pallidum, and left amygdala, in line with past research. Cortical analyses also replicated earlier findings, with higher age associations in regions of the motor and lateral frontal cortices. The findings contribute to the growing research in multi-site harmonization, while providing normative estimates and trajectories of whole-brain R1s that can be referred to in future clinical studies. / Thesis / Master of Science (MSc) / Myelin is crucial for efficient signal transmission in nerve cells, and advancements in magnetic resonance imaging (MRI) have allowed us to indirectly estimate myelin levels in the brain. Collaborative multi-site investigations have become increasingly popular due to their ability to recruit larger and more diverse samples, making their findings more generalizable to the public. However, differences in scanner hardware and configurations across sites can introduce site-dependent biases to MRI scans. In this thesis, we found considerable inter-site variability in our whole-brain MRI-based estimate of myelin based on longitudinal relaxation rate (R1). We thus developed a method that can remove unwanted site-dependent differences from images by comparing repeated scans across sites. Using the corrected data, we showed how myelin amounts in different brain structures changes with age during early adulthood. Overall, we highlight the importance of carefully considering site-specific differences in multi-site MRI data and characterized how myelin changes with age in young adults.

No description available.

The extension from Quantitative Magnetic Resonance Imaging to synthetic imaging has the clear advantage of being able to continually image the patient after the exam. MR techniques such as DWI are commonly used but have some clear disadvantages resulting from the use of echoplanar imaging. It should then be asked whether one imaging technique is objectively better. If one technique is better, the incorporation in clinical settings could produce better diagnostic rates, and save valuable time. In order to quantitatively assess the quality of these techniques, the SNR and CNR values of similar tissues were compared. The pre-analysis discussion concentrating on the spatial resolution and artifacts, supports that synthetic images have an advantage over DWI due to higher resolution and absence of artifacts. The SNR and CNR values were calculated for each patient and image type for the comparison, initially assuming that the synthetic images would have a higher mean SNR and CNR. In most cases the differences between scan types was found to not be statistically significant. In conclusion, this analysis could not support the initial theory that the synthetic images had a higher SNR or CNR. The research shows that they are more likely to be comparable. An investigation of the diagnostic power of the synthetic in comparison to standard DWI would give clinical relevance to these results.