Luke Tierney

I will provide a set of simulated brain images obtained from the BrainWeb simulated brain database at One issue to examine is how to identify major brain tissue types from an image or set of images. Simple approaches can be based on mixture models and classification methods. More sophisticated approaches might take into account spatial similarity of adjacent regions. The size of the images creates some interesting computational challenges. An advantage of using simulated data is that the ground truth is known. The simulator also allows some image quality parameters to be varied, so the effects of these parameters can be examined. The image files are about 6M compressed each and I'll probably have around 20 of them. I can put them on a web page, upload someplace, or bring on a USB drive -- whatever works. I'll get some R scripts ready for some basic analyses. One package that may be useful for visualization is misc3d.