New software tools help take the pain out of working with huge three-dimensional image datasets and aid in mapping neuronal networks.
Computer-generated three-dimensional (3D) images are finding ever wider use in entertainment, and even scientists are increasingly using 3D images generated from image stacks acquired during serial section imaging of biological samples. Although the surface-rendering methods used in the entertainment industry are fast and mature, “biologists have a lot of image data, which is very different from surface data, and they don't have a very good renderer,” says Hanchuan Peng of the Howard Hughes Medical Institute's Janelia Farm Research Campus.
![](http://media.springernature.com/relative-r300-703_m1050/springer-static/image/art%3A10.1038%2Fnmeth0510-344a/MediaObjects/41592_2010_Article_BFnmeth0510344a_Figa_HTML.jpg)
Researchers at Janelia Farm have established an ambitious goal of discovering how the brain works, and an important part of this endeavor is creating a 3D brain atlas. For the fruit fly brain, the first target of Janelia Farm researchers, this is a realistic goal. Individual neurons can be labeled with GFP, and by acquiring fluorescence image stacks through the brain, it should be possible to map all the neurons and their branches.
But once investigators have their stacks of images, how do they interact with the data in a convenient way and make quantitative measurements? Existing tools either require the user to step through the individual two-dimensional images and select regions of interest—which is cumbersome and tedious—or use a virtual reality environment requiring specialized equipment to view and interact with a 3D version of the data.
Peng found that when he was trying to construct a 3D atlas of the fruit fly brain derived from tens of thousands of images, the existing tools were inadequate or difficult to use, so he decided to create a new set of tools. The first results of this effort are V3D and V3D-Neuron.
V3D is designed expressly for working with 3D volumetric data and is built on an efficient 3D renderer that allows real-time visualization and manipulation of multi-gigabyte–sized data on a standard computer. Peng and colleagues worked hard to make it user friendly. “With V3D you can drag and drop an image stack and see it in [three dimensions]. If you want to go to anywhere and see more detail, it is just one or two clicks,” he says.
The most innovative aspect of the software is how easily users can select visual features without stereo viewing or virtual reality hardware. There are two options. First, while viewing the 3D image, a user can mouse-click on a region of interest, rotate the image and click on the same region from another angle. The two clicks define rays that intersect at the desired 3D location and create a marker position there. Alternatively, the software will determine the most likely 3D position the user intended to mark using only a single mouse-click by examining the intensity information along the single ray. Peng says that although it is possible to create image data that fools the latter method, with real data the method is surprisingly accurate. And because the user can always rotate the image and quickly adjust the location, it is very fast.
“The most important part of the 3D pinpointing is that once you have the 3D marker information, you can directly measure it or use it as prior knowledge in computer algorithms,” says Peng. His team used this approach in their neuron tracing tool V3D-Neuron. Instead of relying on manual tracing or automated tracing followed by manual correction, V3D-Neuron allows the user to quickly pinpoint markers only at the terminals of the neuronal branches in three dimensions, and then an algorithm finds the optimal connecting paths. “This produced much better performance,” says Peng.
Work on V3D is continuing. 3D curve drawing has been implemented, other people are designing plugins, and a V3D hackathon is scheduled for this summer at Janelia Farm. It may not be as fun as 3D gaming but V3D promises to make working with 3D image data in the lab much more enjoyable.
References
Peng, H. et al. V3D enables real-time 3D visualization and quantitative analysis of large-scale biological image data sets. Nat. Biotechnol. 28, 348–353 (2010).
Rights and permissions
About this article
Cite this article
Evanko, D. Connecting the dots in 3D. Nat Methods 7, 344 (2010). https://doi.org/10.1038/nmeth0510-344a
Issue Date:
DOI: https://doi.org/10.1038/nmeth0510-344a