Summary

Cogito ergo sum :)

Made from MRI data, which was actually rather coarse (take a look at the picture of the surface reconstruction).
Still. Works. I think :)

Instructions

One DOES aquire knowledge over time :)
This is how I did it:
I am using MRI data with a slice thickness of 7mm. Less would be better but that was all I got at home. The data used is a "dark fluid" sequence, called "FLAIR" by our manufacturer. The advantage of that sequence type is that - as it says in the nick name "dark fluid" - water appears dark. Structures like bone or skin are of a dark grey whereas brain matter is of a lighter grey.
So segmentation of the brain surface is made rather easy. In this case this is done by using InVesalius. The STL file I get from that is cleaned up in Meshlab and there's some smoothing, too (I think I took "Taubin Smooth"). Finally the result of that was repaired using Netfabb.
The apex region has come out the best although some piece of the brain was missing in the scan ;) .
The bottom slices of the scan show some artifacts and contain structures like the skull basis and vessels that show up in the 3D data and have to be removed manually. Still, nice I think.
I must try to get data with a better resolution :)
CT data would offer that but there the segmentation of the brain is a lot of work - with messy results ;)

I Think