"Activity patterns in the brain: breaking up the problem into pieces"
Human brain image data can provide overwhelmingly complex patterns related to neural activity. These patterns vary across individuals, as do the shapes of their brains. To draw inferences about common activity patterns across individuals, it would be prudent to first match corresponding structures within which those patterns are observed. We have recently developed a method and software package for automating anatomical labeling of human brain image data (http://www.arnoklein.net/mindboggle.html) . The program, called Mindboggle, breaks up gross brain anatomy into small pieces, and performs combinatoric matching between the pieces of different brains and applies a modified Self-Organizing Map algorithm  to label about these pieces. New extensions to Mindboggle include the ability to label activity data directly from the gross anatomy, as well as provide multiple anatomical labels for each data point based on the anatomy of multiple brains.  Klein, A., Hirsch, J. 2004. Mindboggle: a scatterbrained approach to automate brain labeling. (in press, NeuroImage)  Kohonen, T. 1997. Self-organizing maps, 2nd ed. Springer-Verlag, New York.