Abstract

The analysis of structure in three-dimensional images is increasingly valuable for biomedical research and computational science. At the same time, the computational burden of processing images is increasing as devices produce images of higher resolution (e.g., typical CT scans have gone from 128^{3 to roughly 512}3 resolutions). With the latest scanning technologies, it is also more common for the the values measured at each sample to be multi-dimensional rather than a single scalar, which further complicates implementing mathematically correct methods.

Diderot is a domain-specific language (DSL) for programming advanced 3D image visualization and analysis algorithms. These algorithms, such as volume rendering, fiber tractography, and particle systems, are naturally defined in terms of a continuous tensor field that is reconstructed from the discrete image data. There are three main goals for the Diderot project:

1) provide a high-level mathematical programming model that will enable rapid prototyping and exploration of the algorithmic design space.

2) enable high-performance implementations across a range of parallel
   platforms. by exploiting the domain-specific knowledge and the inherentparallelism in image analysis.

3) Enable scientists to develop their own image analyses and visualizations.

This talk will give an overview of the Diderot project and describe some of the research problems that we must solve to attain our goals.

Diderot is joint work with Gordon Kindlmann and Lamont Samuels.