New perspectives in visualisation for neuroimaging

Scientific discovery depends critically on the visualisation of data. Visualisation has a special role in imaging experiments, specifically, the comprehension and analysis of data, and the communication of outcomes. Our ability to derive new knowledge from increasingly large and more detailed images depends on understanding, applying and advancing appropriate visualisation strategies to the data at hand. In neuroimaging, visualisation techniques are ordinarily applied to 3-dimensional images, or time-evolving (4-d) images, yet the communication and publication of study outcomes routinely depends on static, 2-d representations of 3-d or 4-d phenomena.

This predicament is not unique to the neuroimaging discipline. However the constraints of 2-d display media have influenced the development of data analysis and visualisation techniques for neuroimaging. For example: (1) techniques for flattening the cortical surface (e.g. Fischl et al., 1999, http://dx.doi.org/10.1006/nimg.1998.0396 ) are in significant part motivated by the difficulty of visualising a highly folded and warped sheet; (2) the ubiquitous red, green, blue shading of diffusion tensor images to show left-right, anteroposterior and superior-inferior white matter fibre direction indeed allows the encoding of 3-d information in a 2-d figure, but the result cannot be fully interpreted by 5-10% of the community (the red-green colour-blind).

We have developed tools that directly address the broad challenge of publishing 3-d and 4-d scientific data (including images) as fully-interactive figures within Adobe PDF documents (Barnes & Fluke, 2008, http://dx.doi.org/10.1016/j.newast.2008.03.008 ; Ruthensteiner et al., 2010, http://dx.doi.org/10.1016/j.micron.2010.03.010). No special viewing software is required, and 3-d PDF figures can now be generated using free software.

Here, we present our technique applied to standard neuroimaging data: 3-d MR images, 4-d fMRI images, cortical surfaces, diffusion tensor images and derived datasets. We describe how publishing and communicating using interactive, 3-d figures, allows us to begin addressing some of the shortcomings evident in discipline-specific visualisation and analysis: is cortical flattening necessary when we can directly publish and visualise complex, 3-d surface structures as figures in PDF articles? Are RGB fibre maps appropriate when we can directly publish a less-derived, more meaningful 3-d tensor image within an academic paper?