Dokument: Brain atlas variation in dynamical whole-brain modeling: How the definition of brain region shapes the simulated functional connectivity of individuals
| Titel: | Brain atlas variation in dynamical whole-brain modeling: How the definition of brain region shapes the simulated functional connectivity of individuals | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=61294 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20221201-145641-3 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Domhof, Justin Wilmer Martijn [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | PD Dr. Popovych, Oleksandr [Gutachter] Kollmann, Markus [Gutachter] | |||||||
| Stichwörter: | Brain connectome; Graph theory; Parcellations; Reliability; Resting-state brain dynamics; Structure-function relationship; Subject specificity; Whole-brain model | |||||||
| Dewey Dezimal-Klassifikation: | 500 Naturwissenschaften und Mathematik » 570 Biowissenschaften; Biologie | |||||||
| Beschreibung: | How the various parts of the brain are mutually connected can be expressed through the concepts of the structural connectivity (SC) and the functional connectivity (FC). Here, the SC describes how different brain areas are interlinked by anatomical connections that facilitate the propagation of electrical signals. Alternatively, the FC reflects which distinct brain regions are consistently and synchronously co-activated. The structure-function relationship as determined by the correlation coefficient between the SC and FC is moderate at best. Hence, many neuroimaging studies have investigated how these two types of connectivity can be better associated to one another.
Some of these studies employ dynamical whole-brain models. These models aim to replicate the FC as best as possible given the information stored in the SC. Indeed, dynamical whole-brain models have been shown to explain an amount of variance that exceeds straightforwardly correlating the SC and FC. Furthermore, studies suggest that these models are promising candidates for future clinical applications. However, the high computational loads associated with dynamical whole-brain models require the modeled system to be low-dimensional, while the SC and the FC are typically derived from high-dimensional magnetic resonance imaging (MRI) data. Hence, the dimensionality of the MRI images must be reduced. A so-called brain atlas or parcellation dividing the brain into a (low) number of brain regions may be used for this purpose. Many brain atlases have been constructed on a variety of methods and neurobiological data reflecting brain organization. It has been shown that a change of parcellation may considerably alter the results of analyses involving only empirical data. Nevertheless, a systematic assessment of the effect of the brain atlas on dynamical whole-brain modeling results is lacking. This thesis contains such an investigation. The first study of this thesis shows that a change of brain parcellation can considerably alter the accuracy with which the dynamical whole-brain models are able to replicate the FC. It also shows that this parcellation-induced variance in the validity of the models can be explained by group-averaged deviations in the network properties of the empirical connectomes, i.e. the SC and FC that are derived from MRI data. In contrast, the within-parcellation, between-subject variations in the quality of model fit could not be explained by the inter-individual differences in those network properties. In short, the study shows that the dynamical whole-brain modeling results are susceptible to the technique used to construct a particular parcellation, and identifies deviations in the network properties of the empirical SC and FC as the cause for this sensitivity. The second study additionally shows that the parcellation influences the reliability and the subject specificity of the modeling results to a higher degree than what is observed for the empirical FC. In addition, it shows that the FC generated by a dynamical whole-brain model can share subject-specific connectivity patterns with both the empirical SC and FC after model fitting. Moreover, it is shown that the acquired results critically depend on the exact implementation of the modeling paradigm. Hence, the study shows that not only the parcellation but also the model implementation can affect the reliability and subject specificity of the modeling results. The results comprising this thesis are highly relevant given the current focus on the personalization and the clinical application of dynamical whole-brain models. They provide a possible explanation for the personalized fits of the models to the empirical data. More importantly, they show that the choice of the brain parcellation could be more important for findings involving these models than for straightforward analyses of the empirical SC and FC. Finally, the thesis presents information that could be used by future dynamical whole-brain modeling studies for the appropriate, well-informed selection of the brain atlas. | |||||||
| Lizenz: |  Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung 4.0 International Lizenz | |||||||
| Fachbereich / Einrichtung: | Sonstige Einrichtungen/Externe » Institute in Zusammenarbeit mit der Heinrich-Heine-Universität Düsseldorf » Institut für Medizin, Forschungszentrum Jülich GmbH Mathematisch- Naturwissenschaftliche Fakultät | |||||||
| Dokument erstellt am: | 01.12.2022 | |||||||
| Dateien geändert am: | 01.12.2022 | |||||||
| Promotionsantrag am: | 31.05.2022 | |||||||
| Datum der Promotion: | 04.11.2022 |
