Ekanayake, J;
(2016)
Applications of realtime fMRI for non-invasive brain computer interface-decoding and neurofeedback.
Doctoral thesis , UCL (University College London).
Preview |
Text
Thesis_Corrections_Final_070416a.pdf - Submitted Version Download (24MB) | Preview |
Abstract
Non-invasive brain-computer interfaces (BCIs) seek to enable or restore brain function by using neuroimaging e.g. functional magnetic resonance imaging (fMRI), to engage brain activations without the need for explicit behavioural output or surgical implants. Brain activations are converted into output signals, for use in communication interfaces, motor prosthetics, or to directly shape brain function via a feedback loop. The aim of this thesis was to develop cognitive BCIs using realtime fMRI (rt-fMRI), with the potential for use as a communication interface, or for initiating neural plasticity to facilitate neurorehabilitation. Rt-fMRI enables brain activation to be manipulated directly to produce changes in function, such as perception. Univariate and multivariate classification approaches were used to decode brain activations produced by the deployment of covert spatial attention to simple visual stimuli. Primary and higher order visual areas were examined, as well as potential control regions. The classification platform was then developed to include the use of real-world visual stimuli, exploiting the use of category-specific visual areas, and demonstrating real-world applicability as a communications interface. Online univariate classification of spatial attention was successfully achieved, with individual classification accuracies for 4-quadrant spatial attention reaching 70%. Further, a novel implementation of m-sequences enabled the use of the timing of stimuli presentation to enhance signal characterisation. An established rt-fMRI analysis loop was then used for neurofeedback-led manipulation of category-specific visual brain regions, modulating their functioning, and, as a result, biasing visual perception during binocular rivalry. These changes were linked with functional and effective connectivity changes in trained regions, as well as in a putative top-down control region. The work presented provides proof-of-principle for non-invasive BCIs using rt-fMRI, with the potential for translation into the clinical environment. Decoding and 4 neurofeedback applied to non-invasive and implantable BCIs form an evolving continuum of options for enabling and restoring brain function.
Archive Staff Only
 |
View Item |
