I like trying out new methods of data analysis and applying them to neuroimaging data. For example, in a recent study I developed a simple multivariate connectivity approach that can incorporate a linear transformation function from one brain region to another. This can be interpreted as tracking the information flow between multiple brain regions when the transformation function is known. In another study, we reconstructed the patterns of activity underlying multivariate classification akin to classifier weight maps, but not confounded by noise covariance among voxels. This allowed us to visualize the relative contribution of each voxel to the final classification and observe how patchy the underlying response pattern is.

To make such methods of data analysis available to as broad an audience as possible, together with Kai Görgen I created a decoding toolbox for multivariate pattern analysis that can be used with minimal prior knowledge and easily extended to the needs of the user.