discussion by evan golub - cs.umd.eduegolub/hdcc106/astronomicalmedicine-exampleted.pdfin “the...
TRANSCRIPT
Discussion by Evan Golub
� Was a graduate student in Harvard's School of
Engineering and Applied Sciences at the time.
� Degree in physics with interests in visualization
and perception.
� Currently a Postdoctoral Research Fellow in the
Computer Science Department at the University of
British Columbia & Associate in Computer Science
in the School of Engineering & Applied Sciences at
Harvard University
� An astronomer was searching for way to look at astronomical data and make sense of it.� Borkin suggested applying techniques from biomedical
imaging to astronomical data.
� The astronomer was able to realize new things as a result of the visualization.
� Heart disease diagnostics CT scan data.� Borkin decided to use tree diagrams from her
astronomy work and shades of red from her infovizcollaborators.
� Increased identification of risk regions from 39% to 91%.
� I would have liked to hear (briefly) how the doctors
and astronomers do at the game. Since I am not a
specialist in either area, I don’t know if it’s an easy
or hard game.
� Is there something in computing that CS practitioners
could easily tell the difference between but non-CS
people wouldn’t?
▪ Complexity of code?
▪ Real or fake programming languages?
� An astronomer was searching for way to look at astronomical data and make sense of it.� Borkin suggested applying techniques from biomedical
imaging to astronomical data.
� The astronomer was able to realize new things as a result of the visualization.
� Heart disease diagnostics CT scan data.� Borkin decided to use tree diagrams from her
astronomy work and shades of red from her infovizcollaborators.
� Increased identification of risk regions from 39% to 91%.
� Interesting to learn the (or at least an) early origin
of tree-based visualizations going back to Darwin.
� One of the visualizations might have been a
TreeMap (invented locally by Ben Shneiderman)
and used on sites like the MarketWatch “Map of
the Market” (http://www.marketwatch.com/tools/stockresearch/marketmap)
and Newsmap (http://newsmap.jp)
� They’ve even been used as art (see the 3rd floor of CSIC).
� The way MRI scanners and radio telescopes gather
data are similar in how they both go “slice by slice”
as they look through the subject of interest.
� The visualization techniques that the biomedical
world has been using can also be useful to the
astronomical world.
� 3D visualization toolkits that doctors use can also
be utilized by astronomers.
� By taking CT data about a heart and using a visualization
technique typically applied to nebula data, and not using a
rainbow color map, doctors who only find 39% of high-risk
regions with the first find 91% of the high-risk regions.
http://www.color-blindness.com/2012/10/22/ishiharas-test-for-colour-deficiency38-plates-edition/
� The human eye has approximately 120 million rods but
fewer than 10 million cones (between 6 and 8 million).
� Brightness and color are detected respectively by the rods
and cones on the retina.
� There are three types of cones, each with a sensitivity to
light of a different wavelength (short, medium, long).
They are essentially red, green, and blue.
� We have around twice as many red cones as green, and
around eight times as many green as blue.
http://hyperphysics.phy-astr.gsu.edu/hbase/vision/rodcone.htmlhttp://askabiologist.asu.edu/rods-and-cones
� It wasn’t clear whether the 3D image was the type of visualization that was commonly used for this application, since it was stated that it originally was used for studying DNA.
� It wasn’t clear how they know the 100% level of high-risk regions to compare against.
� I suggest explaining how the “100%” answer is found and that the end of the example showing the improvements offered by the red-scale 2D tree image should have compared results to a red-scale version of the 3D image for a more complete end-to-end comparison.
� In “The Structure of Scientific Revolutions” (1962) Thomas Kuhn discusses examples of outsiders to a discipline and paradigm shifts.� A classic historical example is how Watson and Crick
came from other sciences to present the idea of a double-helix for DNA’s structure.
� In “The Medici Effect: What Elephants and Epidemics Can Teach Us About Innovation” (2004) Frans Johansson talks about the “intersection” as a place where diverse ideas collide and lead to innovation.
� What area insiders from outside CS might be able
to contribute to our field in the future and what
outside fields could benefit from a CS insider?