When IBM's Watson supercomputer took down Jeopardy champ Ken Jennings just about two years ago, IBM let slip that Watson's next gig would be as a voice-enabled physician's assistant. Now IBM is giving us a little more information, releasing a video demo that shows how Watson might help an oncologist diagnose and treat a cancer patient.
The demo, which warns that it's "not necessarily a direct reflection of systems under development," was created in partnership with Memorial Sloan-Kettering Cancer Center in New York City.
It showcases the IBM Watson Cancer Diagnosis and Treatment Adviser as an aid for the physician, never interacting directly with the patient. In the scenario in the video, a patient has just been referred to an oncologist after a biopsy and CT scan tested positive for lung cancer. Just before the first consultation, the doctor feeds Watson the patient's EMR, and Watson runs that data -- her basic information plus her recent test results -- through an extensive database of tens of thousands of medical journals, industry association guidelines, and specific hospital best practices. It tells the doctor, in a few minutes, a tentative list of treatment options (accompanied by confidence levels) and a suggested list of tests and areas where it needs more information. It also provides a list of ongoing clinical trials for which the patient might be eligible.
If the doctor orders the tests Watson suggests, Watson will use the results to present a better list of treatment options, with higher confidence levels, and to narrow down the list of clinical trials. Watson can also take into account patient preferences (the example in the demo is a patient who would prefer to avoid losing her hair) and new symptoms as they come up. Watson's process is transparent -- if the doctor wants to know why it chose a certain course of action, he or she can always click on a button for more information.
Finally, in partnership with Nuance (presumably, since IBM named them as a partner on the project last year) Watson's voice recognition plays a part in the process too, allowing the physician to add in parameters like patient preferences and new symptoms by dictating them. When the doctor and patient decide on a treatment option (provided they choose one of Watson's suggestions), the system automatically sends the plan to the patients insurance company which, the demo's narrator assures us, will instantly approve the treatment plan because it's evidence-based.
IBM and Memorial Sloan-Kettering made clear that Watson isn't taking on the role of a doctor. It's just that the volume of medical literature is too large for any human to be 100 percent familiar with, and its possible for a key piece of information to simply be overlooked. What Watson brings to the table is intelligently processing the huge amount of data available and presenting the physician with the relevant, important results. And it will only get better at what it does -- Watson will save the cases fed into it, presumably de-identified, and include that information in future evaluations.
“Watson’s capability to analyze huge volumes of data and reduce it down to critical decision points is absolutely essential to improve our ability to deliver effective therapies and disseminate them to the world,” Dr. Craig Thompson, president and CEO of Memorial Sloan-Kettering Cancer Center said in the statement.
IBM's demo didn't include any information about when the system might be rolled out, or what the specific form factor might be, but we'll be watching the story as it develops. Systems like IBM's Watson could be the solution to the much-lamented gap between the amount of data digital health technologies create and making that data relevant and actionable in a clinical setting.
UPDATE: The Associated Press is reporting that the system can be accessed by tablet or computer and will be adopted by the Maine Center for Cancer Medicine and WestMed in New York's Westchester County by the end of next month, according to IBM general manager Manoj Saxena.