AbstractFunctional connectivity (FC), the temporal co-occurrence of spatially separate neurophysiological events, remains a promising tool for finding neurological biomarkers of psychosis aetiology and risk. Though a FC measurement that is reliable enough to use in a clinical setting has not been discovered, the evolving field of FC analysis provides a wealth of new approaches for querying disease related biomarkers. The purpose of this PhD was to test a set of dynamic FC network analysis methods in a range of early psychosis populations, to provide new avenues for clinical neuroimaging. Across the studies of this thesis two broad approaches were used: 1) the use of novel dynamic FC measurements and 2) the extraction of alternative representations of FC, as opposed to traditional FC representations based on the average of entire functional time-series. This PhD used methods reliant on the modelling of time-resolved community structure, this forming the basis for the methodological approaches chosen.
Study 1 of this thesis compared ‘flexibility’ (a measure of dynamic modular organisation) between resting state fMRI scans from patients with schizophrenia (N=55) and healthy controls (HC) (N=72). In addition, novel methods of ‘between resting state network synchronisation’ and the probability of transition between time-resolved communities were used to further describe group differences in dynamic modular organisation. This study suggested higher flexibility of nodes in the thalamus and cerebellum to be a feature of schizophrenia, as well as higher flexibility of the fronto-parietal, subcortical, and cerebellar networks. Additional analyses showed a complex pattern of altered between RSN modular configuration in schizophrenia, and suggested higher flexibility of the thalamus to involve increased switching between brain modules covering the default mode and somatosensory-motor systems. Such findings suggested more chaotic dynamic modular configuration to be a feature of schizophrenia and localised this finding to brain areas and systems previously suggested to have abnormal FC in psychosis.
Study 2 used a framework for extracting periods of high network integration
(‘cartographic profiling’) in order to provide integrated FC profiles of task fMRI scans from a sample of Clinical High Risk (CHR) (N=57) and HC (N=24) participants. Integrated FC ‘metastates’ were then used to search for sub-networks (Network Based Statistics) related to CHR status, as well as to longitudinal changes in functioning and psychosis symptom scores. This study tested the hypothesis that FC networks derived from integrated metastates would be altered in CHR subjects relative to controls, and that integrated FC features would associate with clinical and functional outcomes. A key finding of this study was the existence of a sub-network, involving brain areas suggestive of bottom-up sensory processing, which was associated with longitudinal changes in positive psychotic symptomatology in the CHR group. This sub-network was not found using whole time-series ‘static’ FC, suggesting this sub-network to be most present during periods of high network integration.
Study 3 included resting state fMRI scans of those with First Episode Psychosis
(FEP) (N=37), 22q11 deletion syndrome (22q11DS) (N=34), and HCs (N=79). Analyses were performed using a 2x2 factorial design (FEP / 22q11DS status) in order to isolate results related to psychosis in the 22q11DS cohort. In a static FC analysis, no sub-networks were associated with a 22q11DS * psychosis interaction, and so interaction terms were removed to interpret main effects of 22q11DS and psychosis status. Distinct patterns of dysconnectivity were found to be associated with 22q11DS and psychosis status. Dynamic modular configuration analyses suggested lower within RSN dynamic modular configuration of two nodes in the left paracingulate gyrus (default mode and fronto-parietal RSNs) in the 22q11DS cohort. An exploratory analysis showed patterns of schizophrenia vs HC differences of between- and within-RSN modular coupling from a previous study to be similar to those seen in the FEP (R = 0.23, p = 0.042) but not 22q11DS groups (with and without psychosis). This study suggests dynamic modular configuration features related to 22q11DS, and shows patterns of dynamic modular configuration to be different between populations with 22q11DS and psychosis, regardless of psychosis status in those with 22q11DS.
|Date of Award||1 Aug 2021|
|Supervisor||Philip McGuire (Supervisor), Matthew Kempton (Supervisor), Nicolas A. Crossley (Supervisor) & Paola Dazzan (Supervisor)|