Serving three-dimensional models of brain structures online

One of the challenges of modern neuroscience is to integrate voluminous data of different modalities coming from different specimens. This task requires registration of the data to a common spatial framework, a stereotaxic brain atlas space. Many atlases have been published as collections of 2D sections; only in recent days MRI-based, volumetric atlases started to appear.

To facilitate creation of 3D brain atlases we developed the 3D Brain Atlas Reconstructor (3dBAR,   http://3dbar.org), software dedicated to transforming two dimensional atlas data of different quality, complexity and origin into three dimensional reconstructions of brain structures. We also propose a Common Atlas Format (CAF) based on SVG to hold any input data and simplify further processing. The format allows for complete, self-contained storage of processed 2D atlas data and is designed to be  interoperable with other software and to be stored in databases.

The reconstruction process leading to a three dimensional model proceeds in two steps. First the source atlas is transformed into a CAF dataset. Then the CAF is processed resulting in complete 3D models of desired structures in the form of a volumetric dataset or polygonal mesh. The workflow leads to automatic, customizable and reproducible reconstructions. Moreover, it enables tracking and reviewing of the whole reconstruction process as well as locating and eliminating reconstruction errors or data inconsistencies. The workflow was designed to simplify interoperability with other neuroinformatics tools by using open file formats (SVG, VRML, X3D, NIfTi) with the possibility to exchange the content at any stage of processing.

We have used this infrastructure to process several widely used atlases, including atlases published in PDF (“The Rat Brain in Stereotaxic Coordinates” by Paxinos and Watson and “The Mouse Brain in Stereotaxic Coordinates” by Paxinos and Franklin) or prepared from volumetric data (The Waxholm Space Atlas; Johnson et al, 2010). Four other datasets were derived from the Scalable Brain Atlas (SBA, http://scalablebrainatlas.incf.org).

To facilitate interoperability with other tools we created a generally accessible web service providing access to reconstruction features included in the standalone version of 3dBAR. The service allows a variety of operations on available CAF datasets, e.g. customizable reconstructions, downloads in various formats, viewing thumbnails of structures, customized viewing of slides from provided CAF datasets and access to structure ontology. Live view of 3D structures is also possible in WebGL enabled browsers where one can manipulate (rotate, zoom, etc.) lightweight versions of the reconstructions. Currently, the 3dBAR service is available directly or through a plugin in the SBA. allowing three dimensional visualization of brain regions. It is also planned to share functionality with WaxML services developed by the INCF DAI.

The final version of the service will contain an extended set of reconstruction options and a browser interface allowing users to log in, manage their reconstructions, submit requests for bigger, non-standard reconstructions, apply their own pipelines and perform calculations using a back-end cluster supporting the service.