In the current era of Big Data, the challenge of consuming and analyzing an unprecedented amount of medical information regularly confronts physicians.
“The amount of medical knowledge is doubling every two years,” says Deborah DiSanzo, general manager of Watson Health. “There is simply no way that a doctor today, or tomorrow, or in five years is going to be able to keep up with this onslaught of information.”
Watson Health, IBM’s cognitive computing platform, is here to help. Watson Health is a clinical decision support system that can read and analyze personal, clinical, research and social data collected from a variety of platforms like medical literature, electronic health records, wearables and apps.
Many people are familiar with Watson’s 2011 debut on Jeopardy!, where it beat the show’s record-holding champion Ken Jennings. But what most people don’t know is that Watson’s strategy for answering Jeopardy! questions— tapping into a vast reservoir of exogenous data and curating a list of potential answers ranked in order of confidence—is strikingly similar to the strategy physicians use when they construct a differential diagnosis or treatment plan.
Watson Health’s vision is to help physicians make clinical decisions—such a generating a differential diagnosis or suggesting a treatment plan— by sifting through and analyzing an often siloed sea of available medical data. To do this, Watson Health constructs a cloud-based data-sharing hub. “We need to get that data in a democratized, anonymized data lake in the cloud that Watson can then do it’s cognitive computing on,” says DiSanzo. Since its launch in April of 2015, Watson Health has been working to establish partnerships with companies such as Explorys and Phytel, both of which will contribute to Watson’s cloud-based data.
The form Watson takes by the bedside will also be largely dependent on its partners. “What Watson does today is read. It reads and is able to read vast amount of data very quickly, and can make connections in that data” explains DiSanzo. “Whether that’s embedded in an Epic electronic record, or whether that’s embedded in Medtronic’s disease management app for diabetes, or whether that’s included in the CVS population health app…it doesn’t matter. This is something our partners can build into their systems.”
In addition to reading, Watson is learning how to “see.” In August of this year, IBM acquired Merge Healthcare Inc.’s medical imaging management platform, which contains over 30 billion medical images, for Watson to read and learn from. In an age where medical images inundate doctors—IBM estimates that radiologists in some ERs see as many as 10,000 images a day—Watson could potentially triage images or read them against other data sets.
Currently Watson Health is partnered with 16 hospitals around the country. Last week, Boston Children’s Hospital (BCH) announced its intention to implement Watson’s cognitive computing capabilities to help physicians diagnose and treat children suffering from rare diseases. To start, Watson will focus on steroid-resistant nephrotic syndrome (SRNS), a genetic form of kidney disease.
“There are about 250 clinical studies and papers written on this particular rare disease,” explains DiSanzo. “A nephrologist would be able to read all those clinical trials, but maybe it takes them a year to get through them all, with their busy schedules.”
Watson can not only read the trials and papers much more quickly, it will hopefully also recognize patterns across the studies and perhaps read them in conjunction with retrospective EHR data and gene sequencing data collected by BCH.
“Our hope is that we will help the clinicians at Boston Children’s find better pathways—better treatments— for children who suffer from this rare disease,” says DiSanzo.
Abby Ballou is the managing editor of MedTech Boston. She has a B.A. and M.Phil in English literature from NYU and the CUNY Graduate Center, respectively. When she isn't writing and editing for MedTech Boston, Abby enjoys reading, rock climbing, watching classic movies and listening to opera.
Send this to a friend