Coupled parallel-coordinates view in nDive: Our collaborators using MR Spectroscopy, confocal microscopy, TEM tilt series, and simulation produce 3D volumetric scalar fields, the relationships between which they want to understand. The nDive project began in response to a request from our colleagues studying mucus composition who are working to provide multivariate data sets with different mucins labeled with different fluorophores.
Resource student David Feng has added an abstract Parallel Coordinates view of the multivariate 3D data sets to nDive (see figure). Parallel Coordinates is an abstract information-visualization technique that discards spatial information but clearly shows data-value clustering; for each voxel from the original data set, a line traveling through the different metabolite concentrations contained within that voxel is drawn. Clusters and patterns in values can be more apparent in this view. Our implementation adds two new features to standard parallel coordinates:
- Smoothly-curved lines make it easier for the viewer to follow connectivity and making the slope flat at the coordinate intersections preserves the ability to see clustering.
- To incorporate variance into this visualization, lines become diffuse and blurry as they become uncertain; indicating the range of uncertainty in their values.
Selecting groups of lines in the parallel coordinates visualization causes them to be colored uniquely and their corresponding voxels to be highlighted in the spatial visualization.