Impact of data processing on diffusion MR image exploration and understanding

Abstract

The aim of this thesis was to investigate the influence of data processing steps on the understanding and interpretation of diffusion MRI-based studies, both in basic and applied research. Diffusion MRI is the preferred imaging technique to map the structural connections of the brain, thus can help us understanding the biological basis of neurodegenerative diseases like in Alzheimer's Disease or unrevealing the vast network of white matter fiber bundles. Diffusion MRI processing methods could be divided into two classes of data preparation steps: “mandatory” or “optional”. Mandatory or non-optional steps are necessary to apply due to the nature of the error present in the data. For example, image realignment due to subject motion is unavoidable in order to achieve sufficient physical overlap of the same tissue from image to image; therefore, motion correction is a mandatory processing step. On the other hand, optional processing steps give the researchers large degrees of freedom during data analysis. During modeling of the diffusion MR signal there are numerous available methods in which case the exact choice will make a significant difference in both practical and statistical terms. Most notably, we propose a new fiber bundle pathway, the Superoanterior Fasciculus (SAF). By taking advantage of advanced diffusion MRI methodology, a consistent bilateral pathway was identified via fiber tractography (FT) in the frontal lobe, which has not been previously described. Reproducibility across multiple participants, different data samples and acquisition settings boosted the confidence that this finding is not based on imaging artifacts.