Interactive 3D-Visualisation of Brain Spatial Networks
Motivation
With the advancement of data acquisition techniques, a huge number of datasets describing brain functions are generated today building the basis for computational approaches to neuroscience. This involves an increasing need for comprehensive and intuitive tools allowing both, exploration and visualisation, of such data. Proper representation of the brain connectivity information could accelerate the rate of neurobiological research in behavioral neuroscience. 3D representations of brain networks in their natural anatomical context can provide neuroscientists with complementary information on the spatial arrangement of a network.
Existing software for 3D graph visualisation have difficulties obfuscating edges and nodes due to many connections, which creates a chaotic image and makes interrelations too complex to interpret.
Our fundamental aim is to visualize better neuroscientific data and support analysis of them. We propose a visualisation approach for brain connectivity exploration tool that will correspond to all needs of neuroscientists and will decrease the level of graph’s visual clutter. It is a view-dependent algorithm that considers the camera position and observation direction.
Intro
For modern behavior neuroscience, it is highly important to find a relation between behavior, genes, structure, and neural functioning. However, this factor of multimodality increases the complexity of their analysis. One of the purposes of the current project is the development of visualizations suitable for the printed media and digital reading, which require well understandable 2D renders of a neurological network.
Method
We have developed an approach that employs a “relative to the camera” but not global position in space which creates an image of a network optimized for a viewer position.
Use of curves allows arranging all the connections in a way that the peripheral to the brain center and distal to each other nodes’ connections will skirt more internal and closely-grouped ones. To achieve this kind of appearance, we have calculated control points of a Bezier spline. Our calculation is based on the projected values, i.e., the device coordinates in a screen 2D space. Definition of the curves consists of four main parameters: vector that possesses a peak point, direction of a bend, radius and curve width.
A vector of curvature belongs to a line which is perpendicular to a hypothetical straight line between nodes and to camera view vector. Bend is directed outwards from the brain volume center in device coordinates.
• Curvature extent is quadratically proportional to the distance between the nodes in 2D.
• Curve width is directly proportional to the distance between the nodes in projected coordinates.
• Normalization of the distance between nodes regarding the screen resolution and camera distance.
Tools used at current project: JavaScript, React.js, WebGL rendering, Three.js.
Also, we used the custom shader for brain volume rendering that supports an anatomical concept and does not overlap and cover a network itself.
Results
Our prototype is computationally cheaper than other methods for the edge clutter reduction. Our prototype for a visualisation was applied for different networks with a range of sizes from 10 to 300 connections. Therefore, it is possible to perform a real-time interaction with an anatomical 3D model with smooth unnoticeable changes in bows’ shapes, keeping a clean structure of a network. The web-based nature of the initial application requires fast rendering in order to be. These results allow a possible implementation in any other tool for brain connectivity exploration, or even for any network exploration in different fields.
These results allow a possible usage of the developed algorithm for a vast range of neuroscientific data. The visualisation component itself could be implemented in any other tool for brain connectivity exploration, or even for any network exploration in different fields.
Comparison between the classic node-link diagram approach and new visualisation method is presented below.
Future work
Through this method for a visualisation we aim to improve the effectiveness of the biomedical research related to big networks and to simplify a way of sharing results.
Through this method for a visualisation we aim to improve the effectiveness of the biomedical research related to big networks and to simplify a way of sharing results.
Further development includes:
- option for close connectivity exploration by adding a gradient fill of the curves (Figure 6);
- multimodality through different line-types (Figure 6);
- color-blind mode;
- switch to voxel (more precise) level;
- export into *.obj;
- GUI improvements.
Work on the whole project was done by Daria Kruzhinskaia, Florian Ganglberger, Katja Bühler (from VRVis Research Center, Biomedical Visualization, Vienna, Austria) and Joanna Kaczanowska, Wulf Haubensak (from group of behavior neuroscientists at Research Institute of Molecular Pathology IMP, Vienna Biocenter VBC, Vienna, Austria)
Thanks for watching!
