Polish INCF Node turns brain atlases into 3D
In 2006, Daniel Wojcik and his group at the then soon-to-be-formed INCF Polish Node at the Nencki Institute for Experimental Biology started to work with some electrophysiological data from Andrzej Wrobel and his lab. The data were local field potentials from the forebrains of six rats, and Wojcik wanted to integrate them across animals. Preferably in a three-dimensional context, since the data was recorded on a three-dimensional grid. But he found out there were no three-dimensional atlases readily available.
I realized it would be very nice to have a tool which could work kind of like an overlay, that it would allow the user to buy a proprietary atlas and then locally generate three-dimensional structures explains Wojcik.
Due to the lack of resources, he had to leave the realization at the idea stage for quite some time. But he couldn’t refrain from spreading the idea at talks and seminars, and about two years later he met a masters student in computer science, Grzegorz Furga, who was willing to take the challenge on for his thesis. The proof of concept was enough to get the group a good grant and to hire people – among them PhD student Piotr Majka, who today is the chief developer on the atlas project. From the start of the project their goal was to develop general services encompassing many different atlases.
Ironically, the technical challenge of parsing different atlas formats initially turned out to be easier to solve than the problem of getting access to atlases to parse. Large established atlases are often copyrighted by their publishers. The solution was a collaboration with the Scalable Brain Atlas, a web service for visualizing brain structures in several different atlases. It is developed by Rembrandt Bakker from the INCF Netherlands Node, located at the Donders Institute for Brain, Cognition and Behaviour, and supported by INCF. Since the atlases available via the service are all free to use, the Polish group suddenly had access to plenty of use cases and material. The Scalable Brain Atlas, on the other hand, could now benefit from 3dBAR’s more extensive 3D visualization capabilities.
They make 3D surfaces that look nicer and more realistic. You can now use their 3dBAR website to combine areas and to make them transparent and to create complex visualizations. And on my end it is also very useful, because I can use their services to show what the quasi-3D graphics on the Scalable Brain Atlas website look like in true 3D. So we really both benefited from this says Rembrandt Bakker.
We give the reconstructed data back to him. He has a plugin on his website that connects to our service and requests representations of the structures, so users who want more precise reconstructions can do so. It is also nice that we are compatible – whenever we tackle a dataset, Rembrandt can access itPiotr Majka adds.
A travel grant from INCF kick started the collaboration, and enabled Piotr to spend a week with Rembrandt working on parsers to convert the Scalable Brain Atlas data to the 3dBAR service format. The group calls it the Common Atlas Format (CAF). A paper describing the CAF and the 3dBAR infrastructure was published in Neuroinformatics earlier this year.
The 3dBAR currently consists of standalone, offline software that can translate two-dimensional data into three dimensions, and a web service that offers hands-on opportunities to view and manipulate whole atlases as well as single slides, in real time. An API offers programmatic access to all the operations that can be performed in the browser, and if the open service is not flexible enough, users can log in to access custom modifications.
We took our offline tool and brought the most important features online. You can ask for specific nondefault formats, select nondefault pipelines and ask for a custom reconstruction says Piotr Majka.
The 3dBAR service was demonstrated to the Nodes at the recent INCF Nodes Workshop, and will be presented at the 5th INCF Congress of Neuroinformatics in Munich in September. Interested users do not have to wait till autumn, however – they can already access the web service at 3dbar.org, or download the offline tool from the group’s website as well as GitHub, the INCF Software Center or NITRC.
Everything the group develops is and will be open to the scientific community, including the 3D Monodelphis domestica opossum brain atlas the group has now started to develop. The plan is to integrate data from magnetic resonance studies on opossum with histology data in different stainings. Once finished, the atlas will also be offered within the 3dBAR framework and related services currently under development.
There are similar issues that people tackle in the digital atlasing infrastructure workgroup, and I think they are of common interest to people working with different animals. We are not trying this as a one-time thing, we hope that we will be able to do such things for different species and animals says Daniel Wojcik.
With the immense amount of work that goes into building a brain atlas, it is clear that leaving the atlas data open and accessible to the research community gives a better scientific return on the resources invested – not only in enabling useful services like 3dBAR and Scalable Brain Atlas to be developed, but also because it allows the community to share tools and improvements intimately tied to the data.