How Watson for Oncology is advancing cancer care

IBM Watson project manager Leanne LeBlanc views analytics of healthcare data. Each person generates more than 1,100 terabytes of health-related data across his or her lifetime, the equivalent of more than 300 million books. Photo credit: Jon Simon/Feature Photo Service for IBM

What alarms Dr. Craig Thompson, president and CEO at Memorial Sloan-Kettering Cancer Center, is that 20% of cancer patients in the U.S. are misdiagnosed.

Thompson made those remarks during an IBM investor briefing in February as he discussed the cancer hospital's partnership with IBM Watson.

Three years ago, Memorial Sloan-Kettering and IBM Watson formed a partnership to develop Watson for Oncology, a cognitive computing system that can analyze large volumes of data including medical literature, patient health records, and clinical trials, to offer personalized, evidence-based treatment recommendations for cancer patients. Over the course of about 15,000 hours logged during this period, a team of physicians and researchers loaded thousands of patient cases, nearly 500 medical journals and textbooks, 12 million pages of medical literature, as well as MSK-curated research, as they trained the system to be a “learned colleague.”

And healthcare professionals continue to feed new data to the system regularly, making it smarter and more intuitive every day. Currently, the system offers recommendations for lung, breast, and colorectal cancers, and it is expanding to gastric-related cancers as well.

“We talk about how to have tighter integration between research and medicine,” said Rob Merkel, IBM Watson's healthcare and life sciences leader. “In research, there are over 700,000 articles published per year. The average researcher reads 200 articles per year. When you factor in all the data sources including clinical and exogenous factors, we believe in one life, an individual generates over 1,100 terabytes of information. It's far beyond human cognition; there's no way for the brain to process it.”


With the acquisition of Merge Healthcare, IBM Watson is able to read radiological data and medical images. Photo credit: IBM

Imagine how this changes the patient-physician relationship and experience. Once Watson for Oncology has the patient's information, it can instantly sort through medical literature from all over the world, find the literature that is most relevant to that patient's specific cancer, and prioritize potential treatment options based on the evidence and the patient's health record.

With the use of its different APIs, the system can also read radiological data, hand-written notes, identify images in specific terms (i.e. being able to identify a particular patient's hand by recognizing distinguishing features), and has voice recognition ability.

It also understands context. “If Watson is ingesting a breast cancer patient and it reads in the notes that the patient's sister had a mastectomy, Watson knows that's an indication of family history even though the word ‘family' may not appear,” said Merkel.

More from Watson Week: What is cognitive computing, and what does it look like in healthcare?

And when there is conflicting data, Watson for Oncology alerts the user. If a nurse notes a patient's tumor as one size and the lab report has another size, Watson considers which is more recent, makes a recommendation, while also making note that there is inconsistency, explained Merkel. Incorrect diagnoses often lead to higher costs.

In 2016, there will be approximately 1.7 million new cases of cancer and nearly 600,000 people will die from the disease in the U.S., according to the National Cancer Institute. About 40% of Americans will be diagnosed with cancer during their lifetime and it is one of the leading causes of death worldwide.

In about fifteen minutes, Watson presents an analysis that would typically take months to come up with, said Merkel. It presents the evidence for each of its recommendations for the physician and patient to discuss.

“This is key,” Thompson said at the investor briefing. “It doesn't replace the doctor-patient relationship.”


IBM is collaborating with 16 cancer institutes, including the University of North Carolina, to apply Watson's cognitive computing power to quickly translate DNA insights into personalized treatment options for patients. Photo credit: Jared Lazarus/Feature Photo Service for IBM

Currently, Watson for Oncology is in use at a number of hospitals worldwide, including Bumrungrad International Hospital in Bangkok; Manipal Hospitals in seven cities in India; and Memorial Sloan Kettering in New York.

Additional Watson oncology-related solutions are being used at MD Anderson Cancer Center and the Mayo Clinic, said Merkel. The  American Cancer Society also recently partnered with IBM to develop a virtual health advisor for cancer patients that will combine Watson's cognitive computing with ACS resources.

At the February IBM investor briefing, Dr. James Miser, chief medical information officer at Bumrungrad International Hospital, said the 580-bed hospital provides consultation to more than a million patients from 190 countries a year and has used Watson for Oncology since December 2015.

More from Watson Week: Second-generation IBM star pushes for partnerships

“One of the things that I have been very worried about from my own practice is keeping up with all the information that's going to be increasingly available,” said Miser. “Watson for Oncology allows oncologists to keep up, to keep current, and give current best practice recommendations to their patients.”

It takes the knowledge of doctors from the world's most prominent cancer institutions and the data they've collected over the years, and incorporates them into a single system that then proposes personalized treatment recommendations, he said.

“We don't have a vast experience but our doctors do like it,” said Miser. “And our patients appreciate the opportunity of having an online, immediate second opinion from a pretty good source of data.”


So what does the future of Watson for Oncology look like? Merkel said IBM Watson is working with partners to tackle the challenges of cancer from almost every angle. The company is training the system in additional types of cancer, in genomic analytics (Watson for Genomics), clinical trial matching (Watson for Clinical Trial Matching), research and hypothesis generation (Watson Discovery Advisor), and ways to engage patients to manage their own treatment.

More from Watson Week: A look at Watson Health in New York

“Our hope is that Watson will be an essential tool to oncologists everywhere,” said Merkel. “Our goal is to assist reducing the cognitive burden of physicians in keeping up with medical literature by providing clinically actionable insights to assist them in treating patients.”