First Up: Diabetic Retinopathy
Ophthalmologist P. Lloyd Hildebrand, MD, FACS, CEO of ifa subsidiary Inoveon and the point person in overseeing the development work in ophthalmology, says that imaging analytics for diabetic retinopathy is the first project the Watson Health development team is working on.
“That involves basically gathering representative images with all the biomarkers of the disease, teaching the machines to do algorithmic analyses, and then being able to integrate all those analyses to determine the severity of the disease,” he says. Topcon acquired 50.1 percent of ifa Systems, a developer of health information technology for ophthalmology, last year.
 |
| The headquarters of Watson Health in Cambridge, Mass. |
During a presentation in the Technology Pavilion at the American Academy of Ophthalmology meeting last month in Chicago, Steve Tolle of IBM Watson Health said the idea is to bring information stored in tens of millions of patient records to the clinic. “What if I could present to you, as you treat a patient, data on thousands of patients with the same exact condition?” Tolle said. “Not from graphics, but based on the same exact lesion, the same exact disease. What would you do with that information?”
Watson’s ‘Voracious’ Appetite
To ramp up, Watson is digesting massive amounts of data. “We’ve run 10 million images already through Watson, so Watson can now automatically detect what organ it’s looking at and in some cases what condition the organ is in,” Tolle said at the AAO. “Voracious” is how Dr. Hildebrand describes Watson’s appetite for data. “It can read 40 million documents in about 15 seconds,” he says. “That means it can read the entire annual ophthalmic literature in less than a second.”
Dr. Hildebrand says the development team envisions three features: a single-screen presentation of chart summaries and key patient information; a tool that identifies gaps in care and provides algorithms that include clinical guidelines; and a clinical decision support system that provides probabilities and optimal pathways for differential diagnoses. “This is intended to be a clinical tool, not a physician replacement,” Dr. Hildebrand says.
The final piece of the software development is creating a comprehensive knowledge base of ophthalmology. “That’s probably the more challenging piece to build,” says Dr. Hildebrand. “That’s a matter of gathering all the concepts and linking them in a knowledge tree so that that knowledge base can help drive all the other applications.”
Dr. Hildebrand expects the first iteration of the Watson Health cognitive imaging tool in ophthalmology to be available by this time next year, depending on the Food and Drug Administration, which may have to review and approve some of the applications before they hit the market. The IBM commercialization team is working on that regulatory aspect, he says. RS
