On Tuesday in this blog I expressed my frustration with VisWeek’s information visualization research awards process. I don’t want to leave you with the impression, however, that the state of information visualization research is bleak. Each year at VisWeek I find a few gems produced by thoughtful, well-trained information visualization researchers. They identified potentially worthy pursuits and did well-designed research that produced useful results. While puzzling over the criteria that the judges must have used when selecting this year’s best paper, I spent a few minutes considering the criteria that I would use were I a judge, and came up with the following list with points totaling to 100:
Effectiveness (It does what it’s supposed to do and does it well.) — 30 points
Usefulness (What it does addresses real needs in the world.) — 30 points
Breadth (Many people will find it useful.) — 10 points
Applicability (It applies to a wide range of uses.) — 10 points
Innovativeness (It does what it does in a new way.) — 10 points
Technicality (It exhibits technical excellence.) — 10 points
Given more thought, I’m sure I would revise this to some degree, but this gives you an idea of the qualities that I believe should be awarded and the importance of each.
One research paper, among a few, that thrilled me by its elegance and exceptional usefulness was presented yesterday by Michelle Borkin of Harvard University’s School of Engineering and Applied Sciences titled “Evaluations of Artery Visualizations for Heart Disease Diagnosis.”
I’ll allow the abstract that opens the paper to summarize the work:
Heart disease is the number one killer in the United States, and finding indicators of the disease at an early stage is critical for treatment and prevention. In this paper we evaluate visualization techniques that enable the diagnosis of coronary artery disease. A key physical quantity of medical interest is endothelial shear stress (ESS). Low ESS has been associated with sites of lesion formation and rapid progression of disease in the coronary arteries. Having effective visualizations of a patient’s ESS data is vital for the quick and thorough non-invasive evaluation by a cardiologist. We present a task taxonomy for hemodynamics based on a formative user study with domain experts. Based on the results of this study we developed HemoVis, an interactive visualization application for heart disease diagnosis that uses a novel 2D tree diagram representation of coronary artery trees. We present the results of a formal quantitative user study with domain experts that evaluate the effect of 2D versus 3D artery representations and of color maps on identifying regions of low ESS. We show statistically significant results demonstrating that our 2D visualizations are more accurate and efficient than 3D representations, and that a perceptually appropriate color map leads to fewer diagnostic mistakes than a rainbow color map.
What you see on the left side of the image above is the conventional rainbow-colored 3D visualization that most cardiologist’s rely on today. Although this visualization accurately represents the physical structure of coronary arteries, this is not necessarily the best view for identifying areas of low ESS. One of the problems is occlusion, which forces cardiologists to rotate the view so it can be seen from all perspectives, which is time-consuming and prone to error. The use of rainbow colors as a heat map to represent a range of quantitative values from low to high ESS is another significant problem that leads to inefficiency and error.
Michelle Borkin eloquently told the story of how she and her colleagues, an interdisciplinary team that included medical experts, worked through the process of designing, testing, and improving HemoVis. Part of the story that fascinated me was the way that they dealt with the expectations and biases of cardiologists, formed by their training and experience with existing systems. People often develop strong preferences for visualizations that perform poorly, merely because they are familiar or superficially attractive. Opening them to other possibilities can be challenging. Because the team worked so well with the cardiologists and because they did fine work with benefits that could be demonstrated, they managed to loosen the cardiologists’ hold on the familiar and open them to a solution that worked much better. When adoption of a new system results in lives being saved, this is a great success.
I don’t want to describe HemoVis in detail, because I want you to read the paper and fully appreciate the beauty of this work. I do want to mention a few features, however, that illustrate the design’s excellence. As you can see in the HemoVis screen below, the coronary arteries are arranged as a simple tree structure rather than according to their actual physical layout. There are three major branches and sub-branches off of them. To make the interior walls of the arteries entirely visible at a glance, they have been opened up and flattened, much as cardiologists sometimes “butterfly” an artery. The width of the artery representation varies in relation to the circumference of artery. A diverging color scale with gray for the low range of risk and red for the high range of risk worked dramatically better than the rainbow scale of conventional images. Just as maps of London’s metro system are easier to use by commuters when the lines are arranged differently than actual geography, this rearrangement of the coronary arteries and simplified color scale perfectly supports the task of spotting the locations of low EST. Form supporting function this effectively is a thing of beauty. Tests involving cardiologists demonstrated that HemoVis required little training and resulted in a significant increase in the number of risk areas that were identified, the elimination of false positives, and a dramatic reduction in the amount of time that was needed to complete the task. In other words, HemoVis has the potential of saving lives.
Follow this link to download the paper to see an exemplar of fine information visualization research. Several other information visualization papers that have been presented at VisWeek this year exhibit fine work as well (including all of the infovis papers submitted by Stanford), but this one in particular touched my heart. (Yes, the pun was intended.)