Watson: The Next Cancer Breakthrough?

Mark G. Kris, MD


May 29, 2012

This feature requires the newest version of Flash. You can download it here.

Hello. I am Mark Kris from Memorial Sloan-Kettering. I would like to spend a few moments talking about some new directions in oncology -- ways that we are seeking for oncologists to do a better job, by bringing together all of the advances that we have in informatics and care, and putting them into the hands of oncologists to enable better decisions for individual patients.

About 2 months ago, my institution announced a partnership with IBM. That partnership was centered on taking the Watson technology and applying it to oncology treatment. Many of you Jeopardy! fans saw how the Watson supercomputer was able to win Jeopardy! against the 2 top champions. It was an impressive display of how a supercomputer could be harnessed to address a specific need. It was a lot of fun, and it showed a lot of, although not its full, potential. Many individuals have asked whether that kind of technology could be brought to bear in making better decisions for patients with cancer.

Think of the situation we face every day. We have an exploding amount of medical information coming at us from all different directions. We have more and more information about individual patients coming to us: more blood tests, more scan results. We have experts, and guideline panels who have culled through the information (at least the information available to them at the time) and put together what they think is the best way to treat patients and have also, where necessary, "filled in the blanks." Where level-1 data didn't exist, expert panels would say, "This is what we think, as a consensus, you should do in these situations."

What if all of that could be brought together for the care of an individual patient? Watson technology, and technologies like it, actually have that potential. With this technology, Watson would extract a library of medical information -- what is known about consensus statements, guidelines, and the like -- along with information on the individual patient. Going into the electronic medical record, Watson would extract information from imaging studies, blood tests, physician notes, and so on, and also be guided by the experts.

In our case, these are experts from my institution, Memorial Sloan-Kettering, but others have proposed to do this kind of analysis as well. The American Society of Clinical Oncology (ASCO®) has proposed a similar kind of system. What if all that could be brought together and ultimately given to you in a handheld device, or on your desktop computer in your office? You could use it in preparing to see a patient. You could use it after your initial visit to make a decision about management.

The interesting aspect about Watson technology, and what I didn't fully understand when I watched that Jeopardy! match, is that Watson does not give an answer. Watson gives a list of answers, and it ranks them according to its confidence in the correctness of those answers -- which is, in many ways, exactly what the oncologist does. We see a patient, we see a clinical situation, we think of all the possibilities for that situation -- all the possible treatments for that patient with that stage of cancer -- and then, from among those possible choices, we choose the best one. Watson could help us do that.

The other cool thing about Watson, and this is a practical point of great importance to oncologists, is that the computer is an impartial observer. Let's say an insurance provider agreed that if the kind of answers that came out of Watson were the answers that you chose as well, it would immediately approve the treatment. So if you chose the treatment that Watson had on its list, it would be preapproved. That would save a tremendous amount of time. The effort that would go into establishing the correct order of treatments, and particularly the correct treatment for an individual patient (on the basis of their clinical information), would make the insurance company more confident that the best decisions are being made.

As we move forward, we are increasingly being asked to show that our decisions are correct, on the basis of the medical evidence. This is something we all do today, but it is not easy to prove. The Watson technology would allow us to do that automatically. By its nature, it would show that our decisions are evidence-based. It would answer all the quality questions from our own institutions and within our own practices, and it would answer questions from people who are looking at us to see how well we are using the resources given to oncologists.

This kind of technology is moving very quickly, and it is likely that we are going to see prototypes within a year or so. So I am alerting you to some new developments in oncology. I happen to be in the center of it in my work at Memorial Sloan-Kettering, partnering with IBM, and also at ASCO®. Many of you have seen the blueprint for cancer care and research that ASCO® has put forward, and I was involved in that. Many of us are working in this area. It's a very exciting future for oncology, and it is going to be very helpful. It is going to improve care and our confidence that we are making the best decisions for our patients. It will make things happen quicker and more efficiently.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.