CFE 2020: Start Registration
View Submission - CMStatistics
Title: How to assess the neuroanatomical correspondence between brain imaging, gene expression and histological data Authors:  Aaron Alexander-Bloch - University of Pennsylvania (United States) [presenting]
Abstract: Comparing the spatial or neuroanatomical pattern of different brain maps is, increasingly, a fundamental part of the experimental logic used in neuroimaging research. Three example comparisons help illustrate this neuroanatomical correspondence problem: 1) cortical folding versus cortical thickness; 2) myelination versus post-mortem gene expression; 3) neuronal cell density versus sensory multimodality. Each of these three examples of neuroanatomical correspondence has been the basis of strong biological inference. But in many studies, the neuroanatomical correspondence problem has been addressed simply by visual comparison or by spatial statistics whose assumptions are clearly violated. We recently proposed a so-called spin test, which generates a null distribution of correspondence by applying random rotations to spherical representations of cerebral the cortex. Dozens of studies have now adopted (and adapted) the spin test. The spin test has also been subject to legitimate statistical criticism, and alternative approaches to the neuroanatomical correspondence problem have been proposed. Unsurprisingly, the various approaches have different strengths and weaknesses when applied to different kinds of datasets -- including the three illustrative cases considered above -- leading to the conclusion that the best approach to the neuroanatomical correspondence problem depends on the details of the experimental context.