Interesting vis, though I'm not too sure how useful it is to us in the health sector because it's not clear how they determine the strength of relationships. Infovis-wise, I guess it could also be improved by using the position of linked elements (rotational angle and distance) to convey information too.

As the CEO of CureHunter and an architect on the system, I have a vested interest to declare up front--but Kieren's comment/post
raises a good question. If you will go to the main home page for the site: www.curehunter.com
you will see the physician's interface. The strengths of the graph relationships are determined by the frequency with which a successful clinical outcome is reported in the peer-reviewed literature and the strength of the specific clinical outcome on a scale of 1-10. For example, a drug that consistently achieves "complete remission" for a cancer might be rated 8 on the scale, and one
where patients showed only marginal improvement and tumors continued to metastasize might be 0-1, if some marginal gain was noted in clinical trial. The AI-based weighting system generally passes the Turing Test with real physicians, a fact which allows the underlying graph data to be directed and used in discovery models and real time evidence-based medicine Clinical Decision Support Systems--where the graphs help the most evidence-supported meds be selected. In theory, this will improve the rate of positive patient outcomes overall. I hope that helps a little. See also, the National Science Foundation Conference: http://biomedweb.info where I will be directing 2 panels on the modeling methods in CureHunter. Thanks for commenting. The Graph Theory in
CureHunter is somewhat the tip of an iceberg and I know it opens up a lot of scientific questions for deeper scrutiny and intelligent scepticism.

Judge