The combination of functional magnetic resonance imaging (fMRI) and multivariate pattern analysis (MVPA) is a powerful method for investigating brain function, with multiple MVPA methods being applied to the task including Logistic Regression, Support Vector Machines, Neural Networks, and Gaussian Naive Bayes classifiers. Careful review of application of these methods revealed a common process used in most studies; the majority of variations occurring in the implementation choices in key sections such as feature selection or classification algorithms being employed. Thus, it is possible to develop modularised tools for performing MVPA of fMRI data which can be applied in a variety of ways through selection of appropriate components. Development of such a toolbox for use by the University of Birmingham Cognitive Neuroimaging Laboratory is described. The modular design allows for flexible application and provides a basis for development of novel methods, which is explored through implementation of a novel cross-validation method and development of a method for investigating the effects of learning on tuning of neural populations. The development process has resulted in an efficient, robust and reliable toolbox, capable of performing a pre-implemented set of standard multi-variate pattern analyses and provides a basis for further development of novel methods.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:678903 |
| Date | January 2016 |
| Creators | Meeson, Alan Charles |
| Publisher | University of Birmingham |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://etheses.bham.ac.uk//id/eprint/6468/ |
Page generated in 0.0014 seconds