Preview

PDF, English Download (9MB) | Terms of use

Abstract

Arterial spin labelling (ASL) magnetic resonance imaging allows completely non-invasive quantification of perfusion and is valuable for the evaluation of tissue function, activity, and viability. However, it has not yet been established in the clinical routine world-wide partly due to a lack of standardisation. This thesis aims to contribute to the transition of ASL into the clinical routine by investigating sources of variability in ASL-based perfusion quantification in the brain and in the kidneys. Firstly, quantification results obtained with different processing options and corrections or with different acquisition parameters were compared using synthetic data, data from healthy volunteers, and patient data. Differences in acquisition parameters and processing options used for analysis of brain ASL data resulted in significant differences in perfusion quantification. Secondly, synthetic ASL data sets of the kidneys mimicking in vivo acquisitions were generated. A data analysis pipeline was developed and evaluated using the synthetic data sets. The registration performed well for both kidneys, with mean structural similarity index measures increasing by 25% on average. The quantification yielded cortical and medullary perfusion values that agreed with a mean percentage difference of 21% and 16% for cortex and medulla, respectively, to the perfusion assumed for the generation of the synthetic data sets. Segmentation results from the processing pipeline agreed well with original segmentation masks, with Dice indices ranging 0.80-0.93, 0.78-0.89, and 0.64-0.84 for whole kidney, cortex, and medulla, respectively. Thirdly, kidney ASL data were acquired in healthy volunteers and analysed with the developed processing pipeline. Four ASL measurements were performed for each subject varying between free breathing or synchronised breathing and with or without cardiac triggering. Registration performed best when considering the entire image, with a 87% success rate and a mean duration of 30 minutes. Percentage differences between literature values and mean perfusion values were equal to or below 32%, 61%, and 53% for whole kidney, cortex, and medulla, respectively. Across subjects, perfusion values obtained for the four different measurements were only significantly different between the free breathing and synchronised breathing measurement when considering the whole left kidney. Temporal signal-to-noise ratio was not found to differ significantly between the four measurements. Renal perfusion was found to depend on the trigger delay chosen for cardiac triggering. This study’s results suggest that an acquisition in free breathing without cardiac triggering is the best choice for clinical applications.