Understanding of Brain Function

Decoding patterns of neural activity onto cognitive states is one of the central goals of functional brain imaging. Standard univariate fMRI analysis methods, which correlate cognitive and perceptual function with the blood oxygenation-level dependent (BOLD) signal, have proven successful in identifying anatomical regions based on signal increases during cognitive and perceptual tasks. Recently, researchers have begun to explore new multivariate techniques that have proven to be more exible, more reliable, and more sensitive than standard univariate analysis. Drawing on the eld of statistical learning theory, these new multivariate pattern analysis (MVPA) techniques possess explanatory power that could provide new insights into the functional properties of the brain. However, unlike the wealth of software packages for univariate analyses, there are few packages that facilitate multivariate pattern classi cation analyses of fMRI data. This in turn prevents the adoption of these methods by a large number of research groups to fully assess their potential with respect to cognitive neuroscience research. Here, a novel, Python-based, cross-platform, and open-source software framework, called PyMVPA, for the application of multivariate pattern analysis techniques to fMRI datasets is introduced. PyMVPA makes use of Python's ability to access libraries written in a large variety of programming languages and computing environments to interface with the wealth of existing machine-learning packages. The framework is presented in this thesis, and illustrative examples on its usage, features, and programmability are provided. In addition, this thesis provides an overview of promising strategies for the application of MVPA to neuroimaging datasets. While various possibilities are reviewed based on previously published studies, the primary focus lies on the sensitivity analysis technique that is shown to provide interesting additional information which are readily available as part of any typical MVPA-based study. Moreover, this technique is eminently suited for modality-independent data analysis, a feature that is demonstrated by an example of a uniform analysis of datasets from four di erent brain imaging domains. The thesis concludes with a discussion about the challenges that have to be faced to establish MVPA as a standard analysis procedure, including the statistical evaluation of results, as well as potential pitfalls in their interpretation.