I’m involved with a research project here at Red Hat that is really very interesting. It’s an open-source platform for executing medical image processing codes and managing the results. Imagine you’re a radiologist and you want to quickly run some kind of algorithm — say, a brain volume scan — on the 100 brain MRI slices you just got. You can either go to the command line, assemble all the files, remember how to execute the image processing program, and hope it works, or you can use this nifty app — called ChRIS — we’re helping to build and do it all by pointing and clicking.
For the big Red Hat Summit event coming up in May, we want to add a new capability to this app which is even more interesting — the ability to compare the result of running one of these plugins with existing results held by hospitals doing the same kind of work. The catch is that these results will be protected by patient confidentiality (as they should be), so the hospitals can’t legally share them with other hospitals even for research. For this project, though, we are leveraging a thing called Secure Multi-Party Computing to allow us to share knowledge about the collection of results across hospitals, without actually sharing any of the private data. So a radiologist should be able to ask things like “What percentile does this particular amygdala fall in relative to others in the same age range” and get an answer that is informative but doesn’t compromise anyone’s privacy. This is really cool and potentially game changing for radiology, and medicine in general.
HOWEVER there is a little problem. The UI for the ChRIS app is in the middle of being reworked, and because this is a volunteer project, it is unlikely we will be able to finish it in time for the planned demo. So the chances are we’re going to fake the video to show what the app will look like when it’s actually finished, even though it isn’t.
Now, if anyone asks us, we’ll cheerfully tell them that there is work left to do and that the UI doesn’t really work yet. This isn’t a commercial product, after all. But even then, is it really wise — or ethical, even — for us to do this? We’re trying to generate enthusiasm around the project and gain more contributors; would we be better off to be completely honest about exactly what state the app is in right now?
I think the answer is actually “no.” In describing what a piece of software does, it is almost always better to give the sales description (“It will shine your shoes, too!”) than the engineering description (“It’s a pile of garbage that no one should ever touch.”) People expect the optimistic spin, so if you don’t spin, they think your app must in fact be really terrible, which probably isn’t fair.
What a strange industry…
Leave a Reply