Parashkev Nachev 'The use and misuse of in silico lesion mapping' скачать в хорошем качестве

Parashkev Nachev 'The use and misuse of in silico lesion mapping'

Once thought to require nothing more sophisticated than mass-univariate statistics, lesion-deficit mapping is increasingly recognized to be amongst the most complex problems in neuroscience. Its distinctive strength--inference to neural necessity--is paradoxically the source of its greatest vulnerability: dependence on the parameterisation of the lesioned brain as a whole, at least whereas nearly everywhere--the lesions are large in proportion to the mapped substrate. Such dependence violates the foundational assumptions of mass-univariate analysis, voiding its conclusions of all force. First demonstrated by combining simulated functional ground truths with real lesion data, it is a vulnerability some have tried to rectify through the same in silico approach, retaining the mass-univariate framework it has revealed to be critically deficient. I discuss what simulations can and cannot licitly do here, and sketch out a path to evaluating the high-dimensional multivariate models lesion-deficit mapping will always require, whether we like it or not. Moderated by Stephanie Forkel (@StephForkel) and Lauren Thiebaut (@LaurThiebaut)