Six ways IBM is putting Watson to work in hospitals

By Jonah Comstock
Share

IBM WatsonSince IBM launched its Watson Health business unit at HIMSS last year, the group has pushed out plenty of news -- mostly partnerships, acquisitions, and the opening of Watson Health's Cambridge office. But at the HIMSS Connected Health Conference, Shahram Ebadollahi, vice president of innovations at IBM and chief science officer of the IBM Watson Health Group, spoke about how those partnerships and acquisitions are manifesting in actual hospitals.

Here are six examples of ways IBM's cognitive computing engine either will help health care providers or is helping them right now.

1. Analyzing medical images: Building on IBM's recent acquisition of Merge Healthcare, Medical Sieve is a system that analyzes medical images and helps doctors spot details and irregularities.

"What we’re doing there is a system called Medical Sieve that actually looks at the content of images of different modalities and different organs to extract the minute details, to extract anomalies," Ebadollahi said. "Not only images but multi-modal data -- what I can extract from text and combine with what can be seen in the images, all combined and put in front of the oncologist and the radiologist who need to have the comprehensive view of the patient when they’re making decisions."

2. Watson for Oncology: Helping out cancer docs at Memorial Sloan-Kettering was one of Watson's very first healthcare endeavors, not too long after a version of the platform beat Jeopardy champion Ken Jennings. Now the evidence-based treatment advisor has actually been exported -- complete with all the accumulated knowledge of the Memorial Sloan-Kettering staff -- to Bumrungrad Hospital in Thailand.

"It’s as if you have encapsulated the knowledge of Memorial Sloan Kettering oncologists in this tool: how they make decisions about the cancer patient and the route to the source of that evidence," Ebadollahi said.

3. Gene-based treatment recommendation: Similarly, at a number of hospitals, including Cleveland Clinic, Yale, Columbia, Duke, the New York Genome Center, and, most recently, Boston Children's Hospital, Watson is helping doctors assess the role of genetics in cancer and rare diseases and apply that knowledge to their patients.

"Clinicians lack the time and the tools to bring DNA-based treatment options to patients," Ebadollahi said. "And in order to do that they need to correlate all this genomics data with the electronic medical record of the patient, the available knowledge and literature which is out there with respect to treatment issues and bring all of that together. The amount of literature, the rate of generation of knowledge, is doubling every five years. ... What Watson is trying to do here is to help oncologists sift through and look for the needle in the haystack of all the available literature, and bring that to the attention of the oncologists as they’re trying to make a decision about the treatment options."

4. Melanoma screening: A proof-of-concept study with Memorial Sloan Kettering also looked at using Watson in a dermatological context.

"We are looking at melanoma and skin lesions, and mobile apps that could help people better screen themselves and better take care of themselves," Ebadollahi said. "The system is a proof of concept on images that were taken with dermatoscope. The system achieves 94 percent accuracy, where a human can only get in the 75 percent accuracy range."

5. Improving EMR usability: A project with the Cleveland Clinic is looking at ways to use Watson to summarize the electronic medical record to make it easier for clinicians to find the information they need.

"Working with Cleveland Clinic specialists and physicians, we looked at the cognitive load and cognitive requirements of the physicians that are trying to understand this EMR data," Ebadollahi said. "We came up with a tool based on Watson which we call EMRA – electronic medical record advisor. This sifts through and reads all the clinical and physician notes, extracts concepts from them, generates a problem list, and then attaches elements of that list to anything that can be seen and read in the electronic medical record of that patient."

6. Chronic condition management: Finally, drawing on the technology from its Phytel acquistion, IBM is working on a tool that would help extend the care of patients with chronic conditions outside of the hospital and into their daily lives.

"For chronic condition management, there are 5,000 hours per year people aren’t seeing a physician or a nurse," Ebadollahi said. "How to help them to help themselves during those 5,000 hours to take better care of themselves? So we’re actually creating apps as an extension of our Phytel technology, which is about care management and care coordination, to provide these kinds of insights to the patient. Things as simple as, if they are passing by a retail pharmacy, should you pop up something that says they should go take their flu shot? Or something as complicated as tackling adherence issues with respect to vital medications they need to take."