Functional connectivity networks describe how regions of the brain interact. The timing, location, and frequency of these interactions inform about memory, decision making, motor movement, affective states, and more. However, while these interactions are well described as networks, these networks, like many others throughout nature, are constantly changing. Complex network evolution poses a highly dimensional problem but also contains much information about the system in question. In this thesis, a recent class of evolving complex network models was explored and extended to capture the functional connectivity dynamics observed in neuronal networks. Functional connectivity was investigated through data- and model-driven techniques at the cellular level, with cultures of cortical neurones on multi-electrode arrays, and at the whole-brain level, with electroencephalography. At the neuronal level, complex spatial dependencies were identified in bursts of excitation and two novel network models, the Starburst model and the Excitation Flow model, are used to capture the resulting functional connectivity. At the whole-brain level, functional connectivity dynamics were used to perform single-trial classification of intentional motor movement. Again, spatiotemporal dependencies were identified and used to present three novel techniques for modelling the network dynamics. The first two techniques decomposed networks into network templates (one model-based and one spectral-based) and modelled the dynamics with hidden Markov models. The final technique was a generalised evolving version of the Starburst model. The hidden Markov model of spectrally decomposed networks was shown to classify motor intentions with an accuracy around 80%. Firstly, this thesis shows that time plays an important role in the production of the complex network topologies observed in functional connectivity, both at the cellular and whole-brain leve1. Further, it is shown that evolving complex network models are very useful tools for modelling these topologies and that the network dynamics can be used to uncover features that are crucial to identifying functional states.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:590143 |
| Date | January 2012 |
| Creators | Spencer, Matthew |
| Publisher | University of Reading |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0109 seconds