|
Abstract:
|
When imaging data is collected from different subjects for comparison or analysis across the population, there is almost always some preprocessing done before analysis. One of the most common steps is some form of registration (also known as spatial normalization). Here data is (non-)linearly warped geometrically from one domain to another, often taking with it some additional, functional information. However, it is routinely the case, that once the registration takes place, this warping information is discarded. Here, we will discuss approaches where this information, along with the registered functional data both form integral parts of the modelling. We will investigate how shape might be related to function via these models, in situations as diverse as musculoskeletal imaging through to neuroimaging. It will be shown that by including both components in the model, a much fuller description of the data can be achieved.
|
