3D Brain

I recently wanted to see where acetylcholinesterase (AChE), the enzyme that breaks down the neurotransmitter acetylcholine, is located in the brain. And specifically, I wanted to see the distribution of AChE in 3D. This data is available online (for the mouse brain) at both the Allen Brain Atlas and BrainMaps.org, so I went to both sites to see the AChE distribution in 3D. Here I was veritably shocked by the comparison. First, let's look at the Allen Brain Atlas 3D file for AChE, which is the big yellow amorphous blob shown on the right


Then I went to BrainMaps.org and obtained the 3D file for the AChE distribution in the mouse brain, and this is what it looks like in the figure to the right. Note that in both of these figures, the viewpoint is oblique lateral, with anterior pointing to the left and spinal cord on the right.

Understandably, I was shocked by the difference in quality between the two 3D distributions. First off, I have difficulty interpreting the first figure (from the Allen Brain Atlas) because it is just a big amorphous blob of yellow spots.

Secondly, and more troublesome, is that the first figure (from the Allen Brain Atlas) is just plain incorrect. The distribution of AChE, in 2D, looks like the figure at the right. Here we see that AChE is located primarily in the striatum (the dark red color). The striatum is clearly discernible in the second figure (from BrainMaps.org) but is not visible at all in the first figure (from the Allen Brain Atlas).

And here's the kicker: it's not just AChE where the Allen Brain Atlas data is completely wrong!   Nonetheless, it's not the objective of this article to be critical of the Allen Brain Atlas since I have done this elsewhere. I will only note that $40 million should have resulted in decent 3D reconstructions, and better quality in situ data. The fact that so much money was poured into this project and it just produced a pile of crap still astounds me, since it goes against the GIGO (garbage in, garbage out) rule, or at least requires a modification, courtesy of Paul Allen: "$40 million in, garbage out". Granted, it's not as euphonious as GIGO, but it works.

In any event, the objective of this article is to consider 3D brains and how to render 3D brain structures and distributions. We have polygonal modeling, which is currently employed at brainmaps.org, but this has the drawback that surfaces need to be defined, which may not be practical for continuous distributions. Surfaces may be considered "isosurfaces" of a continuous distribution, but what if we want to view the complete distribution in 3D? Then we're talking about volumetric visualization and not surface visualization (unless you're talking about isosurfaces of a volume).

Surfels are one possibility too, but suffer from poor rendering and performance issues. So what we are left with, apparently, is polygons as the best way to visualize the brain in 3D.

The 3D brains used in the figures above, and more, are located at the 3D Brain Objects Database at BrainMaps.org. Note that you can view the 3D brains directly in your browser!