Estimation and visualisation of brain functional and effective connectivity

Functional and effective connectivity are two important concepts in the field of neuroscience that describe how different regions of the brain communicate and work together to support various cognitive and behavioural functions. Despite the many advances in functional and effective connectivity research, there are still several important research gaps that need to be addressed. This thesis explores the novel estimation and visualisation of brain functional and effective connectivity using electroencephalography recordings, with a particular focus on its potential impact on the diagnosis and monitoring of neurological disorders. This thesis proposes two novel methods for estimating brain functional connectivity and effective connectivity. The first method, Revised Hilbert-Huang Transformation, outperforms wavelet-based methods in terms of promising features and time-frequency resolution, providing a potential biomarker and diagnostic tool for Alzheimer's disease. The second method, causality detection attention-based convolutional neural networks, effectively estimates effective connectivity networks and identifies disrupted connectivity in Alzheimer’s disease patients. These methods contribute to the growing literature on connectivity estimation and offer valuable insights into the neural mechanisms underlying cognitive processes and neurodegenerative diseases, providing potential diagnostic and monitoring tools for healthcare professionals. This thesis also introduces a novel directed structure learning GNN (DSL-GNN) to leverage several EBC estimations to extract discriminative biomarkers for dementia classification. In studies of Alzheimer's disease, epilepsy, Parkinson's disease, and workload classification, the thesis demonstrates that the proposed brain connectivity methods have better performance compared with traditional methods based on individual channel. It suggests that functional and effective connectivity may track more changes from healthy people to patients to a certain extent, providing the possibility for earlier and more accurate diagnoses. Specifically, the thesis finds that specific regions of the brain can contribute to the diagnosis of epilepsy and dementia disease as well as workload classification based on brain connectivity. By advising the appropriate placement of electroencephalography sensors based on these identified regions, doctors and researchers can more efficiently and accurately diagnose and classify these neurological disorders, reducing the burden on healthcare systems.