Monday, May 25, 2015

Will Watson Compete For Radiology Jobs?

IBM recently released an ad from the Ogilvy Group agency currently airing on national TV that directly addresses how Watson, IBM's ongoing supercomputing research project popularized by its star turn on the TV show 'Jeopardy!, may one day play a role in diagnostic imaging:

Is this realistic, or just some far off pipe dream? Technological aspirations often tend to outstrip technical reality. For example, computer-aided diagnosis (CAD) in mammography has not become a panacea for screening mammograms, and ultimately does not appear to have had a significant impact on the labor demand for mammographers. Anecdotally, mammographers I have spoken with regard CAD as too sensitive, and a potential liability. Sure, the software will pick up the subtle lesion occasionally, but it will pick up many more that are merely normal tissue, but nonetheless force the radiologist to make an active decision to dismiss. What happens if the mammographer decides a positive finding on CAD is negative, but cancer develops later anyway? Unless the software can improve both sensitivity *and* specificity, it may create as many problems as it solves.

For argument's sake though, let's say the new technology can improve both sensitivity and specificity.  Several open questions come to mind: will the software aid radiologists? Does it seek to replace radiologists? If it does, who takes on liability if the software 'misses' a lesion, as it inevitably will? And who pays for it all? It is hard to see these questions being answered in the affirmative in the current healthcare system, but after seeing ACA pass, it is clear that the environment is not static. If costs start to rise again, and IBM and similar software vendors market themselves well, one could see hospitals giving such systems a try. The tl;dr - money talks and BS walks: if these systems can do the job better than radiologists, they will be used.

How can radiologists operate under such potential uncertainty? Instead of ducking the issue, radiologists should work towards simultaneously figuring out how best to employ these systems while also demonstrating the value-add of a physician diagnostic imager both to referring clinicians and to patients. If the face of the radiologist is merely the report they generate, very soon that face may take a very Big Blue complexion.