A Computer that Treats Cancer
November 7, 2013
|Mark G. Kris, M.D., FCRH '73, spoke about the use of IBM's artificial intelligence computer Watson to help treat cancer pateints.
Photo by Bruce Gilbert
You might let Watson, the supercomputer that defeated Jeopardy!
champion Ken Jennings, to be your proxy on a trivia game show. But would you allow it to treat your cancer?
For Mark G. Kris, M.D., FCRH ’73, the answer is a resounding “Yes.”
Dr. Kris, an attending physician in the thoracic oncology service at Memorial Sloan-Kettering (MSK) Cancer Center and the William and Joy Ruane Chair in Thoracic Oncology, was the featured speaker at Fordham University Science Council
’s fall lecture on Nov. 6.
In his talk, Dr. Kris described how oncologists at MSK are using Watson, the artificially intelligent (AI) computer designed by IBM.
“The amount of information that has become available in medicine is extraordinary. But there’s no way that one physician can access all of that information to help a patient,” said Dr. Kris, the lead physician in the MSK-IBM Watson Collaboration.
“Computers take on that problem by tapping into all of that knowledge and using it to assist doctors in decision-making.”
The equivalent of 6,000 high-end home computers, Watson is a data reservoir, containing entire encyclopedias, dictionaries, thesauri, news articles, sacred texts, literary works, and more. But what distinguishes Watson from other computers is its abilities to understand language and to learn—the “holy grail” of AI.
Watson’s software contains algorithms that work together to discern the complexity and nuances of human language, including double meanings of words, puns, metaphors, rhymes, and inferred hints. The computer can also employ evidence-based learning by noting when it solves a problem correctly and applying this understanding later.
Besides dethroning Ken Jennings, Watson computers are being put to work in a number of fields. In healthcare, physicians use the supercomputer to choose treatments for cancer patients.
Before even meeting a patient in person, an oncologist will enter the patient’s medical history and relevant information into Watson. The computer applies this information to its vast database of medical textbooks, scholarly journals, and treatment outcomes in previous cases to offer a list of possible of treatments, ranked according to Watson’s confidence in the answer.
“Information is not knowledge, and knowledge is not wisdom,” Dr. Kris said. “But what Watson does is give you a list of possible answers and then ranks them. That’s what doctors do, too—make a list of differential diagnoses.”
The physician presents these findings to the patient, and the pair decides on a preferred treatment. Their decision is then entered into Watson. Once the treatment outcome is known, Watson makes note of whether it was successful and takes that knowledge into account for the next case.
“We train Watson as we would train a doctor—by repetition, by apprenticeship, by showing what previous results were,” Dr. Kris said. “And unlike a medical student or a doctor, Watson doesn’t forget once it learns something.”
“This is not pie-in-the-sky thinking. It’s happening now, and I believe this is the future.”
The Fordham University Science Council’s mission is to promote and facilitate the continued advancement of science education at Fordham, to provide opportunities within the science, technology, engineering, and mathematics (STEM) disciplines for students and alumni, and to promote faculty research.
to read more about the Council. View a video of the full talk below.
Founded in 1841, Fordham is the Jesuit University of New York, offering exceptional education distinguished by the Jesuit tradition to more than 15,100 students in its four undergraduate colleges and its six graduate and professional schools. It has residential campuses in the Bronx and Manhattan, a campus in West Harrison, N.Y., the Louis Calder Center Biological Field Station in Armonk, N.Y., and the London Centre at Heythrop College, University of London, in the United Kingdom.