|
Abstract:
|
In many application areas it is of interest to compare the shapes and sizes of high-dimensional surfaces, and to investigate symmetry. We focus on a particular application in neuroscience, investigating large scale cortical shape differences between control and schizophrenia patients. We introduce an automatic maximum likelihood method for brain registration, identifying the inter-hemispherical join, the anterior commissure, and the posterior commissure. Likelihood based inference is considered, and significant differences between the two groups are observed. The model is extended to account for curvature in the inter-hemispherical join. General practical issues in high-dimensional data analysis will be discussed.
|
