Improving and validating methods in lesion behaviour mapping

Abstract

The investigation of diseased brain is one of the major methods in cognitive neuroscience. This approach allows numerous insights both into human cognition and brain architecture. Most prominent is the method of lesion behaviour mapping, where inferences about functional brain architecture are drawn from focally lesioned brains. In the last 15 years, the state-of-the-art implementation of lesion behaviour mapping has been voxel-based lesion behaviour mapping, which is based on the framework of statistical parametric mapping. Recently, the validity of this method has been criticised and multivariate methods have been proposed to complement or even replace it. In my thesis, I aim to evaluate these different methodological approaches to lesion behaviour mapping and to provide guidelines on how lesion-brain inference should be drawn. In my first empirical work, I investigate the validity of voxel-based lesion behaviour mapping. It shows that previous studies overestimated biases inherent to the method, and that validity can be improved by the use of correction factors. The second empirical work deals with a recently developed method of multivariate lesion behaviour mapping. On the one hand, I clarify how this method can be used to obtain valid lesion-brain inference. On the other hand, I show that the method is not able to overcome all limitations of voxel-based lesion behaviour mapping. In my last work, I apply multivariate lesion behaviour mapping to investigate the neural correlates of higher motor cognition. This analysis is the first to identify a brain network to underlie apraxia, a disorder of higher motor cognition, which underlines the benefits of the new multivariate approach in brain networks.