By Steve Hamm
Dr. Jose Morey has a full-time job as a radiologist with the U.S. Veterans Administration in Hampton, VA. He also teaches part-time at the University of Virginia and Eastern Virginia Medical School. As if that wasn’t enough, he is helping IBM develop a system, Medical Sieve, aimed at assisting doctors to interpret medical images.
Why does he do it? “I have an eight-year-old son,” Jose says. “I tell him that someday a computer might help save his life. I’ll play a little part in that. And even when I’m gone it might help his kids. It’s a legacy thing.”
Jose is one of eight practicing physicians who are helping out with the Medical Sieve project, a three-year-old effort involving IBM scientists in San Jose, Calif., and Haifa, Israel. The project emerged in the aftermath of IBM Watson’s victory on the TV quiz show, Jeopardy! While the original Watson tapped written language, the Medical Sieve system adds another dimension to the Watson technology family: vision.
Medical Sieve is intended to help radiologists and other specialists spot anomalies in medical images ranging from X-rays and MRIs to angiographs. The computer reviews hundreds or even thousands of images so humans don’t have to do so–selecting a few images the human should concentrate on. It also interprets the images, providing background information that the physician might find useful.
Even though smart machines don’t really think like humans, they need to know how humans think and what they know. And they need to be trained to do their jobs well. That’s why experts in a wide range of domains are helping IBM develop systems designed to help people like them do their jobs better or faster. We’re working with oncologists, scientific researchers and financial advisors, to name a few. “I’m teaching Watson to think like a radiologist,” says Jose.
The consulting physicians in the Medical Sieve project have a range of “training” duties. They help IBM engineers write algorithms used for spotting anomalies. They contribute their knowledge and experience to the vast Medical Sieve knowledge base. They review sets of images that the computer system also reviews–so we can benchmark the system against human experts. And they help us prepare question-and-answer pairs that we use to train the system. “We couldn’t do this without them,” says Tanveer Syeda-Mahmood, chief scientist and overall lead of the Medical Sieve project.
There’s another important role for the consultants: As evangelists for the technology within their professions. Some physicians react negatively when they first hear about the Medical Sieve project. How will it affect their jobs? Jose shares his views with colleagues at the VA and at conferences. He tells them the system isn’t intended to replace them–but to help them do their jobs better. “People fear change and disruption,” he says. “I tell them why I believe this is where we should be going as a society and as physicians.”
The great promise of cognitive computing is that by combining the efforts and brainpower of humans and machines, we’ll get better results than either one could achieve separately. Jose is fully committed to delivering on that promise.