Onicas, Adrian (2023) Applications of Network Science in Neuroimaging. Advisor: Ricciardi, Prof. Emiliano. Coadvisor: Cecchetti, Dr. Luca . pp. 158. [IMT PhD Thesis]
![[img]](http://e-theses.imtlucca.it/style/images/fileicons/text.png) |
Text (Doctoral thesis)
CCSN33_Adrian_Onicas_Applications-of-Network-Science-in-Neuroimaging (1).pdf - Published Version Available under License Creative Commons Attribution Non-commercial Share Alike. Download (60MB) |
Abstract
At the intersection between neuroimaging and network science, network neuroscience has brought remarkable opportunities to advance the understanding of the human brain. At macroscale, the brain can be seen as a complex system relying on communication between its regions. Advanced neuroimaging techniques can map functional and structural brain communication, enabling the study of network-level alterations in neurological disorders during development. To improve reproducibility and provide a robust characterization of neurological disorders, collaborative initiatives involving neuroimaging data collection across multiple sites have started to emerge. However, multisite data acquisition poses significant challenges for managing increasingly larger and more complex datasets, especially for data analysis pipelines required for whole-brain network analysis. To this end, the current work aims to (1) assess different data harmonization techniques and (2) characterize structural and functional network alterations in mild traumatic brain injury. In addition to the typical applications for whole-brain network analysis, network science can be utilized for advanced time series analysis and address challenges for signal processing in functional neuroimaging. Visibility graphs can map time series into networks where nodes represent time points, and have rapidly found applications across areas of science, including resting state functional neuroimaging. However, to validate their use, the current work aims to test if task activity can be identified based on the local network centrality (node degree) in synthetically generated data and event-related task fMRI time series. The Advancing Concussion in Pediatrics (A-CAP) study is the largest study of mild traumatic brain injury to date, and was used to address the first two aims of this work. To understand network-level alterations following mild traumatic brain injury in the pediatric population, the current work validates the use of ComBat harmonization for network analysis pipelines and tests for longitudinal alterations in network topology. ComBat harmonization had improved performance in removing site effects when applied on network parameters instead of edge-wise connectivity weights, and demonstrated excellent within-site consistency with the network parameters before harmonization for structural and functional networks. Network parameters based on structural and functional connectivity show no effects of injury before or after harmonization in the post-acute phase following mild TBI. However, further longitudinal analysis of global and nodal abnormalities in the functional connectome indicates that variability in time post-injury, post-concussive symptoms, biological sex, and age moderate the effect of injury in local and global functional network topology. To address the third aim of the current work, two datasets were used. First, synthetic data was generated to resemble well-controlled eventrelated task fMRI signals, by adding varying levels of noise. An accuracy score was defined to compare the identifiability of task events based on visibility graph transformation versus the raw fMRI time series across noise levels. The results were replicated using a slow, event-related picture presentation dataset, with extensive scanning of four participants. When applied to time series analysis, visibility graphs can accurately identify task events and are robust to gaussian noise in synthetic time series and to participant motion in real task fMRI data. The current work addresses substantial contributions in mapping the human brain using neuroimaging and network science
| Item Type: | IMT PhD Thesis |
|---|---|
| Subjects: | R Medicine > RC Internal medicine |
| PhD Course: | Cognitive, Computational and Social Neurosciences |
| Identification Number: | https://doi.org/10.13118/imtlucca/e-theses/375 |
| NBN Number: | urn:nbn:it:imtlucca-29082 |
| Date Deposited: | 20 Apr 2023 07:13 |
| URI: | http://e-theses.imtlucca.it/id/eprint/375 |
Actions (login required, only for staff repository)
 |
View Item |
