reading folsom and professor hayles’ comments on overwhelming quantities of information in database systems (“the tempestuous relationship of narrative and data”), i immediately thought of scott fraser’s (a professor at caltech and director of their brain and the biological imaging centers) regular comment that “scientists today collect more data than humans can perceive.”
while technological developments are making it possible to generate and store large datasets of information, researchers are also looking for new ways to interpret them so that they can be understood on human terms. an inspiring example for such a narrative is the painting-motivated diffusion tensor representation technique developed in professor fraser’s lab in 1998.
i briefly mentioned this in class a couple of weeks ago, but i thought i’d post a link to the full paper visualizing diffusion tensor images of the mouse spinal cord (laidlaw et al.). as outlined in the paper, the nine dimensional mri diffusion tensor data is typically shown like this…
but using the laidlaw technique developed in collaboration with the caltech conceptual artist, davidkremers, the group created visualization that looked like this…
“Our second method applies concepts from oil painting to display diffusion tensor images. We used multiple layers of brush strokes to represent the tensor image and the associated anatomical scalar image. The brush strokes reflect the geometric nature of values derived from the tensors and of the relationships among the values. Also, the use of underpainting and saturated complementary colors evokes a sense of depth. Together, these painting concepts help create a visual representation for the data that encodes all of the data in a manner that allows us to explore the data for a more holistic understanding.”
diffusion data = visualization
anatomical image = underpainting lightness
voxel size = checkerboard spacing
ratio of largest to smallest eigenvalue = stroke length/width ratio and transparency
principal direction (1) = stroke direction
principal direction (2) = stroke red saturation
magnitude of diffusion rate = stroke texture frequency
this visualization presents the information in an intuitive way that doesn’t require years of training to understand some of the basic information of the dataset. for instance, deterioration of the spinal cord on the right (patchiness) is obvious as compared to the healthy organism on the left.
the brilliance (in my opinion) of this work is that it combines a nine-dimensional data set into a single 2D image, and in doing so allows the viewer to intuitively “look into the future”… the organism on the right has no physical symptoms of the disease, but based off the visualization, it’s possible to predict where and when it will develop spinal cord damage.