This article was originally published on GeekWire.
BOSTON – Microsoft’s robotic mosquito trap is so smart it can tell one insect species from another – and that’s good news for scientists fighting the Zika virus, dengue fever and other mosquito-borne maladies.
It can also tell if you’re buzzing an electric toothbrush in its vicinity.
The toothbrush was used to dramatic effect today by Ethan Jackson, a Microsoft researcher from Redmond, Wash. Jackson heads up Project Premonition, a research effort aimed at giving epidemiologists smarter tools for tracking disease outbreaks. Today he showed how the high-tech trap works at the American Association for the Advancement of Science’s annual meeting in Boston.
Jackson used a smartphone app to set up the breadbox-sized device, then flipped on his electric toothbrush and brought it close to one of the trap’s 64 compartments. After a couple of seconds, the compartment’s door shut with a snap.
What set it off? It wasn’t the buzzing noise, and it wasn’t the shape of the toothbrush. “It doesn’t look anything like a mosquito,” Jackson noted.
Instead, the trap was programmed to recognize the fluttery pattern of the toothbrush’s moving bristles, illuminated by infrared light. “It’s really using the shadow caused by the bristles,” Jackson said.
Thanks to machine learning, the trap can be programmed just as easily to discriminate between different species of mosquitoes.
During a weeks-long test last summer in Houston, Texas, the trap could determine the species successfully 80 percent of the time through infrared scanning, Jackson said. And when other factors were added to the algorithm – for example, the time of day and the amount of ambient light available – the accuracy rate rose to better than 90 percent.
The system is designed to automate the painstaking job of sorting through the different types of bugs that are caught in mosquito traps, and zero in on the species that could cause a crisis.
The prime example is the Zika virus, which can cause microcephaly and other brain abnormalities. Zika set off a global health emergency last year when it rapidly spread through a host of countries in the Americas, starting with Brazil and ending up in the southern United States.
Project Premonition’s smart traps could become part of an early warning system to watch for specific mosquito species – for example, Aedes aegypti, the primary carrier for the Zika virus.
“The device could really pinpoint where Zika vectors were entering the environment,” Jackson said. Fortunately, none of the mosquitoes caught during the test was actually carrying the virus.
Researchers also are working on a system for rapid genetic sequencing of the mosquitoes that are trapped. Such a system could look for the DNA signatures for Zika or other bug-borne pathogens such as West Nile virus, chikungunya virus or the virus that causes dengue fever. If those signatures are found, epidemiologists might be able to head off outbreaks before they start.
That part of Project Premonition is handled by Microsoft’s partners at academic institutions, including Johns Hopkins University, Vanderbilt University, the University of Pittsburgh and several University of California campuses.
The trap-and-sequence process can generate a huge amount of tracking data for public health officials. “They have never seen data like this before,” Jackson said. “There could be several terabytes of data about mosquito behavior in one season.”
Now Project Premonition’s scientists are trying to figure out the most effective ways to deploy the smart traps, and are also working to reduce the cost of the hardware to mere hundreds of dollars. Jackson said the aim is to provide public health agencies with robotic field biologists that are “no more expensive than the traps they use today.”
Eventually, the project’s researchers plan to use robotic drones to monitor sampling sites and even check the traps remotely – but Jackson said the flying robots haven’t yet been deployed for those purposes.
Jackson and his colleagues came across some surprises in the course of their experiments in Texas, including several cases where mating pairs of mosquitoes were found caught together in the traps.
“One of the things we learned as computer scientists working on this project is that nature does what it does,” he joked.