According to the original plan, the Qualcomm Tricorder XPrize should be announcing its grand prize winner right about now. The proud organizers should be bestowing a US $10-million check on the team that successfully built a consumer device capable of monitoring vital signs and diagnosing a variety of diseases—just like the tricorder wielded with such gusto by the medical officers of Star Trek.
Instead, after the finalist teams began consumer testing of their devices last year, the XPrize decided that everything needed to be pushed back—because those initial tests didn’t go well. “The teams fell short in terms of performance,” says Grant Campany, director of the prize. Now the teams are improving their technology for a new round of testing in September 2016, and a winner will be declared in early 2017.
In an interview with IEEE Spectrum before a panel discussion about the tricorder prize at SXSW Interactive, Campany forthrightly explained the reason for the delay.
When the teams delivered their devices for testing at UC San Diego’s Clinical and Translational Research Institute, the evaluators recruited real patients with the diseases that the tricorder is meant to diagnose. The tricorders were required to monitor 5 vital signs and identify 16 conditions. “The devices did do well in vital signs and user experience,” Campany says, “but they fell short in the diagnostics.” Which was the most important part.
The main problem was that the devices’ algorithms couldn’t deal with the complexity of humans, Campany says. Remember, these devices aren’t meant to be used with a doctor’s help; instead consumers should be able to work their tricorders on their own. So when a user fires up a tricorder, it first asks about symptoms. Based on those answers it will interpret the user’s vital signs and instruct the user to conduct various tests to arrive at a diagnosis.
But two people with hypertension might describe their symptoms in completely different ways, Campany says. As testing began at UC San Diego, it quickly became clear that the devices’ decision-making software wasn’t reliably arriving at the right diagnoses.
After about 50 tests, the XPrize organizers called a halt. “We still think that these teams are two years ahead of where the industry is,” Campany says, but the teams needed more R&D time if they were going to meet the XPrize’s ambitious goals.
There have been a few other changes to the competition as well. A few teams dropped out: Of the original ten finalists, seven teams now remain. The organizers also scaled back the number of conditions that the tricorders are required to diagnose, from 16 to 13. Campany says they removed some of the conditions from the list because they couldn’t recruit enough patients in the San Diego area with that condition (such as hepatitis A). They also removed the requirement to detect stroke because “the risks outweighed the benefits,” Campany says. The organizers didn’t want patients at risk of stroke to rely on experimental devices when their lives might be at stake.
The remaining list of conditions is far from comprehensive, but Campany says it provides a representative sample, including acute illnesses like pneumonia and urinary tract infections and chronic ailments such as diabetes and pulmonary disease. The long-term goal, he says, is to create a device that can serve as a general platform and that can be updated routinely to diagnose more diseases. If one of the teams succeeds in meeting the short-term goal of producing a device that provides accurate diagnoses, they can start on that next challenge.
What’s it like for the teams competing in the Tricorder XPrize? Read more here: The Race to Build a Real-Life Version of the “Star Trek” Tricorder
Eliza Strickland is a senior editor at IEEE Spectrum, where she covers AI, biomedical engineering, and other topics. She holds a master’s degree in journalism from Columbia University.