Abstract (eng)
The human brainstem is a very challenging part of the brain for fMRI scans, due to an increased level of physiological noise and its anatomical peculiarities. In this theses we evaluated, if valid and reliable BOLD signals can be detected from the human brainstem, with the risk map technique, a method with minimized methodological assumptions.
We first compared the BOLD signal strength in brainstem motor nuclei and motor cortex. Then we tested, if and how data quality may be improved, by using four different preprocessing conditions. Finally, we tested influence of different reference functions of the hemodynamic response on our data.
We could show an improvement of BOLD signal strength of our data by spatial smoothing and improved sensitivity and specificity with masking conditions. However, results indicate that voxel reliability in brainstem, detected with the risk map technique, is much below the performance in cortical areas. An fMRI analysis technique with minimized methodological assumptions would require more scan repetitions – a limiting factor for a patient investigation. The data also point out that – irrespective of the analysis technique used - in order to improve validity of the functional brainstem results, it is important to avoid as many artifacts as possible already during the data recording process.