Repository logo

Multilayer network methodologies for brain data analysis and modelling



Change log


Dimitri, Giovanna Maria 


The term neuroscience includes in itself a plethora of research areas devoted to undercover the most fascinating complex organ of our body: the brain. A common denominator of neuroscience areas, is the need for the application of methodologies to integrate different features. In this thesis, we focused on the analysis of two types of brain data: brain data coming from Traumatic Brain Injury (TBI) patients and data collected for the study of neurocognitive healthy ageing. In both cases there was the need of applying computational techniques able to integrate different features. To do so we used multilayer networks. For two groups of TBI patients (adults and paediatrics), time series data were collected from the observations of IntraCranial Pressure (ICP) and Heart Rate (HR). We first detected events of simultaneous increase of HR and ICP, which we called brain-heart crosstalks. Subsequently time series were translated into graphs, and network measures, during brain-heart crosstalks, were obtained. These were then included as predictors in a mortality outcome model, with crosstalks. Causality measures were also investigated, using a Granger causality approach, to understand the dynamics of signals during these events. We further applied multilayer networks to study neurocognitive ageing. To do so, we implemented a pipeline for community detection, which we called NetRank, applying it to the Cam-CAN, a large cross-sectional cohort for the study of healthy neurocognitive ageing. Using multilayer networks modelling, we identified subgroups of individuals, with similar lifestyles, and we related them to structural and functional brain features. We believe that multilayer networks and their extensions represent a powerful tool to be used in integrative and cross modal neuroscience datasets. New insights on cognitive neuroscience and time series analysis, can in fact be gained trough multilayer network, possibly improving patients managements and allowing to develop new predictive tools.





Lio', Pietro


Multilayer networks, Neuroscience, Machine Learning


Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge