Watson better learn to read crabby handwriting and minds. The fundamental issue in applying AI to medical data is the low information dataset. We are working on a couple of smaller AI projects in my hospital and finding that even when electronic, most entries in the record are narrative not discrete.

Applying ranges and logic to data pulled out of narrative records is tricky and leads to unusual responses. Even when discrete, the data set may be difficult to use. My favorite example seen nationally is blood transfusion data. It can be in units of 250ML, 500ML or just ML. Users don't look at the unit label for a field. If we ask for the most precise measurement in ML, we get amounts of 1 or 2. We know nobody got a thimble of blood.

AI interventions have to be quick. It's no good telling the provider later he made a mistake. The goal is to steer them in the right direction before they act without presenting them with so much info they throw in the towel. Not easy and in infancy.

One of my buds is IT director for a company that resupplies generators. The logisitics for it are crazy as you route trucks on available streets, deal with priority of the customers (hospitals front of the line) and optimize resupply into mostly empty tanks before they actually empty, etc... And you have >10 days of this 24/7 after a storm.

After Hurricane Allison here, some companies in this sector went out of business before the power came back on. We were better prepared for Ike, but I think that's because facilities are more willing to sign contracts with the pricey but reputable businesses.

This is the most expensive way to get fuel: you need a massive amount for a short time, and you need it consistently during that time.

Joined Slashdot just to post on this thread. No longer a lurker! 20 years IT, working with health systems for 6. Healthcare desperately needs math and physics gurus, and you will get respect in this industry. The data sets are finally large enough to need your expertise. I've worked on projects with meteorologists looking at boundary detection and then anatomy matching in 3D imaging data sets, We've done Gaussian curvature in mitral valve deformation. I've got another project looking at quality outcomes using mathemateticians we hired away from Oxford and Trondheim universities. Some of these are startups, some are nationally funded research projects. Some blur that line. All are ethical, in sense of first do no harm. In the tougher issue of weighing intentional harm vs capability, it's good the be surrounded by people who consider that their life's work. As those projects come up, each is evaluated and risk/harm is openly discussed. I love this environment, and love that occasionaly I get to run stats to show interventions prompted by new data modelling techniques saved lives over the traditional standard of care. You can make a difference, please do so.