Watson — IBM's new supercomputer can interpret patient's symptoms in an instant (The Hindu Businessline - October 17, 2011)

Sherlock Holmes solved his cases by throwing weird questions at his loyal doctor-narrator-friend, Watson. Soon, doctors too may have their own smart-thinking Watson to crack tough medical puzzles. IBM has named its new and about-to-debut wiz kid after its founder Thomas J.Watson. Here, it says, is a supercomputer that 'speaks' the same complex language that we do. Feed it a difficult medical case report where you cannot figure out the illness and Watson can do what perhaps no other computing system has come close to: in three seconds, it can tell you exactly what ails the person.

The people's supercomputer

In situations where a diagnosis eludes and confounds the doctor, the supercomputer narrows down the options to may be two or three diseases in almost no time, according to Dr Manish Gupta, Director IBM Research-India and Chief Technologies, IBM India/South Asia. It would have 'read' data equal to a million books or about 200 million pages of medical literature that is unavailable on the web or Google, and compared the patient's record with similar cases before coming up with the likely disease or recommending a treatment.

In effect, it would help the doctor to diagnose and treat the patient efficiently, eliminating wrong guesses or the need for second and third opinions. It would also alert the doctor about side-effects of a drug or two clashing prescriptions. Point to note - "Except that in this case we are not at all talking about replacing the doctor."

Arguably Watson looks to be the first "people's supercomputer". The others are mostly used by scientists, in institutional labs for the military, weather modelling, drug discovery and such high-end purposes. Watson's rare talent to understand and respond to the nuanced human language was showcased in the US in February when the quick-thinking machine defeated two champions of a tough quiz show, 'Jeopardy'; it gave its answer well ahead of them, and went on to bag over $1 million in prize money.

Sample a clincher: "In 1898, Portugal celebrated the 400th anniversary of this explorer's arrival in India." In three seconds, Watson was bang on with the answer Vasco da Gama.

According to Dr Gupta, "Clearly one industry that stood out as one that could benefit from the Watson system was the healthcare industry. A doctor is essentially doing something analogous to what a person does on a quiz show: questions that are asked on the quiz show are parallel to a patient coming in and describing his problem. There may be medical test reports, descriptions of the problem and the current state. In confusing or ambiguous situations, doctors rely on literature they had read in school or current medical developments and their own experience before coming out with one or more hypotheses (of what the disease could be)."

Though Watson won the game, Dr Gupta admitted it goofed up sometimes. "Very consciously, very humbly, we are trying to set this up more as an aid, a complementary tool of the doctor. We recognise fully (that it) is a very serious matter; when you deal with patients' lives you cannot rely on the machine to make decisions. The ultimate decision has to rest with the doctor."

Doctors encounter complicated, ambiguous conditions in 5-10 per cent of the cases they get. Today, he said, medical literature had reached a tipping point, with new research taking place all the time and making it hard for even well-meaning doctors to stay up-to-date in their field. This is where Watson came in, with the ability to understand and respond to natural language, 'react' to patients' symptoms or records and throw up relevant options from its resources. "We are hoping to leverage the Watson system's remarkable capabilities of identifying the relevant information from all available literature to the patients' condition, or find similar cases, and come up with recommendations."

Healthy benefits

IBM has an SBU (strategic business unit) to commercialise Watson; the process has apparently just begun. India, with its record growth in the healthcare industry, is very much on the radar. Dr Gupta said, "It is too early to say. In the next 12-24 months we should begin to see some systems with these capabilities." Discussions are also on with "a few significant players in the eco-system", and for business models and innovations that can go into it. The next ones to be plugging into Watson could be help desks, customer service people, the financial world including banks and stock markets and even the legal field.

Meanwhile, the system is getting enhanced into an interactive one which comes back with its own doubts or seeks more inputs. Speech-based solutions are being added through a tie-up with Nuance Communications. On September 12, IBM named WellPoint, Inc. as a partner who would create the first commercial applications of Watson by early 2012. Big Blue is also working with experts from Columbia University Medical Center and the University of Maryland School of Medicine.

IBM, Dr Gupta said, envisages Watson to be in the back-end, in the cloud. The hospital or clinic subscribes to it. In difficult cases, the doctor discreetly 'consults' Watson via a mobile device like a smart phone or laptop, feeding in the patient's ID or RFID and getting disease options and recommendations. Doctors in the US have validated it. As for the cost borne by the patient, Dr Gupta said it would come down with the application of information technology. You wouldn't do redundant tests; the doctor gets a single view of patient records, in his view.

The supercomp specs

It took 25 IBM scientists four years to ready Watson. For the fact freaks, Watson is powered by 10 racks of IBM POWER7 servers running Linux. It uses 15 terabytes of RAM, 2880 processor cores and can work at 80 teraflops or 80 trillion operations a second. It is not connected to the Internet and goes beyond a routine question-answer machine even as it combines IBM's Deep Question Answering (QA), Natural Language Processing and Machine Learning technologies, according to IBM.

"We eventually see medical care evolving into a system where doctors increasingly use analytics. Already there is a plethora of tools, imaging or ultrasound machines that doctors use today." However, the very first step, Dr Gupta conceded, was for hospitals to digitise patient records to be fed to the computing system. Many hospitals are not ready for this.

"Clients in India are very discerning; they adopt technology very rapidly, as we have seen happen in the telecom industry," Mr Gupta said. "We leapfrogged the West in many aspects of how we leverage technology, more by the sheer scale. I would not be surprised to see some leapfrogging with this too."