Aligning Large-scale Medical and Biological Data Sets: Exploring a Monkey Brain

This project addresses the issue of developing interactive rendering methods for datasets which cannot be stored on a single hard drive or in main memory anymore. Our dataset is a set of 1400 slices (single cross-sections) of a monkey brain, which has been sliced more than 15 years ago at the Center for Neuroscience at UC Davis, and recently has been scanned at a very high resolution (more than 10MB per image in compressed format). The enormous resolution allows us to zoom from a global view down to the cell level, all in one image. This exciting range of rendering options requires scalable, multiresolution rendering techniques. The challenges we encounter with this data set is an extreme misalignment of the slices due to manual placement onto glass object carriers and manual insertion in the film scanner. We present a semi-automated method which compensates for most of these artifacts and identifies those slices that cannot be handled and aligned automatically. The algorithm reduces the number of slices that need to be treated manually enormously.