A Computational Architecture of Emotion Processing in Younger-Onset Dementia

Abstract

Breakdown in cognition and behaviour are key features in Alzheimer’s disease (AD) and frontotemporal dementia (FTD) in the early stages of the disease. These changes contribute to the emotion processing disturbance observed in these dementias. This thesis applies computational modelling to understand the underlying architecture of emotion processing disturbance and their neural correlates in these two common younger-onset dementias. The approach implemented in this thesis demonstrates the utility of studying clinical populations to understand the mechanisms associated with human cognition. Together, this thesis highlights the complex and multifaceted nature of emotion processing. Four empirical studies were undertaken to address these objectives using a combination of neuropsychological assessment, cognitive and behavioural measures, and structural neuroimaging techniques. Study 1 identified that both AD and FTD exhibited similar emotion processing impairments using hierarchical clustering analysis. This result was further confirmed by similar cognitive processes associated with emotion processing, as shown by a state-of-the-art network analysis (based on graph theory). Study 2 extended the emotion model by incorporating the contributions of behaviour, namely mood and motivation, using similar network analysis used in Study 1. Study 3, using a novel statistical analysis maximal information coefficient, revealed novel brain regions were associated with emotion processing in these two populations, which provide biological support to the computational models in Study 1 and Study 2. Finally, Study 4 modelled the evolution of emotion processing disturbances with disease severity in AD and FTD based on dynamical system theory, using the mathematical model cusp catastrophe.