RSNA: IBMs Watson has a future in radiology
van Terheyden described the process behind Watson’s artificial brainpower, dubbed DeepQA, and elaborated how the technology could impact the delivery of care. DeepQA reflects a significant advancement in artificial intelligence and could allow radiologists to retrieve pertinent clinical data that otherwise may not be available due to researching time constraints.
Watson works by analyzing an initial question and posing several chains of hypotheses based on a semantic framework of data. van Terheyden loaded massive amounts of clinical data into the supercomputer and tested the technology by posing a number of questions related to care. For the most part, Watson was impressive.
Asked what hormone deficiency was associated with Kallmann’s syndrome, Watson answered correctly with GnRH. Several other questions were answered correctly as well. However, when asked a question about obesity hypoventilation, Watson answered incorrectly.
After the kinks are worked out, van Terheyden anticipates the technology will be useful in the care setting, particularly with the exponentially growing amounts of healthcare data. “We see this as being able to provide knowledge at the point of the radiology read,” he said.
The challenge is pulling non-standard systems together to retrieve information, many of which do not have an analog to connect to Watson, he noted. Importantly, Terheyden predicted that Watson will never replace the human doctor.
"The ideal clinician is confident, empathetic, humane," he said. "Watson is the antithesis."