Abstract
The brain is the most complex system in the human body and understanding its function and structure is of high scientific interest as reflected by several ongoing multinational scientific efforts. “Connectomics”, based on
diffusion MRI, assesses the structure of the brain as a network graph and provides new insights of the topological organisation of healthy and pathological brains. However, network reconstruction involves several steps, such as segmentation, parcellation, registration, fiber orientation estimation, and fiber tracing - all of them with a high degree of parametrization. This thesis investigated the impact of different combinations of state-of-the-art tractography algorithms, diffusion weighted acquisition and parcellation schemes on structural connectivity and derived network measures. Additionally, the intra- and inter-subject variability was tested and the meaningfulness of the
structural connectivity map validated by 8 major white matter bundles. This work showed that at least 100,000 fibers have to be generated to obtain a connected structural connectivity network which is reliable for further analysis.
A high intra-subject but a moderate inter-subject reproducibility was found and group-wise studies are only meaningful with parcellation scales of 83 and 129 brain regions. In general, the selection of the number of nodes had the most substantial impact on the network measures, but also the tracing algorithm. In conclusion, measures of structural connectivity are highly depended on acquisition and processing and therefore not comparable between
studies. High-resolution diffusion MRI in combination with a probabilistic multi fiber orientation model yields most connections but comes at the cost of increased scan as well as calculation time.
In contrast, conventional DTI acquisition with FACT fiber tracing is fast and has the least false positive connections, which is preferable for clinical structural connectivity studies.
diffusion MRI, assesses the structure of the brain as a network graph and provides new insights of the topological organisation of healthy and pathological brains. However, network reconstruction involves several steps, such as segmentation, parcellation, registration, fiber orientation estimation, and fiber tracing - all of them with a high degree of parametrization. This thesis investigated the impact of different combinations of state-of-the-art tractography algorithms, diffusion weighted acquisition and parcellation schemes on structural connectivity and derived network measures. Additionally, the intra- and inter-subject variability was tested and the meaningfulness of the
structural connectivity map validated by 8 major white matter bundles. This work showed that at least 100,000 fibers have to be generated to obtain a connected structural connectivity network which is reliable for further analysis.
A high intra-subject but a moderate inter-subject reproducibility was found and group-wise studies are only meaningful with parcellation scales of 83 and 129 brain regions. In general, the selection of the number of nodes had the most substantial impact on the network measures, but also the tracing algorithm. In conclusion, measures of structural connectivity are highly depended on acquisition and processing and therefore not comparable between
studies. High-resolution diffusion MRI in combination with a probabilistic multi fiber orientation model yields most connections but comes at the cost of increased scan as well as calculation time.
In contrast, conventional DTI acquisition with FACT fiber tracing is fast and has the least false positive connections, which is preferable for clinical structural connectivity studies.
| Original language | English |
|---|---|
| Qualification | Master of Science |
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Publication status | Published - 2015 |
Keywords
- Connectome
- structural connectivity
- dwMRI
