In August 2016 in Florida and California, initial testing began on the most advanced prototype of an intelligent trap that, in addition to attracting and catching insects, automatically identifies their species and sex. The device, created by two Brazilian researchers and their American colleague, can simplify the process of monitoring populations of disease-transmitting mosquitoes such as Aedes aegypti, the vector responsible for spreading dengue, chikungunya and Zika.
“The trap has undergone continual refinement since we started developing it,” says entomologist Agenor Mafra-Neto, president of Isca Tecnologias, a Brazilian pest control company that is involved in the project. Mafra-Neto, along with computer scientists Gustavo Batista of the University of São Paulo (USP) at São Carlos and Eamonn Keogh of the University of California at Riverside, had already produced versions that use a laser in the detection mechanism before obtaining better results with an LED. They have also built models powered by batteries and by power cords connected to the electrical grid, and they are studying ways to connect a solar panel to the device. During the process of making these adjustments, they have tested different versions of a software program that recognizes the species and sex of the mosquitoes by their wingbeats. “By the end of the year, we hope to have a robust prototype adapted for use in the field to test in Brazil,” says Mafra-Neto.
The trap is a simple device consisting of a cylinder of black fabric nearly 60 centimeters long and closed on the bottom. Its austere appearance, however, gives no hint of the technology at its core. At the top of the cylinder, a black tube serves as an entryway for the insects. Attached to the tube is an electronic sensor that uses light to detect mosquitoes passing through. The sensor identifies the insects’ species and sex by the frequency of their wingbeats.
The insects are attracted by an artificial scent, developed by Isca Tecnologias, that resembles the aroma of fresh plants. Whenever a mosquito enters the trap, it crosses an infrared cone of light emitted by an LED. The insect’s passage through the lighted area projects a shadow onto the sensor, which is transformed into electrical signals that are then fed into the computer program that recognizes the species and sex. The time of the event and the temperature and humidity are immediately recorded. Once inside the trap, the mosquito cannot escape and dies of dehydration.
The use of light to measure wingbeats was a major design coup. Researchers have sought out strategies for automatic execution of this task ever since the 1940s. But earlier devices recorded the sound using microphones, which picked up ambient noise.
Both the detection apparatus and the recognition software were developed by Batista’s team at USP. His and Keogh’s groups, along with the team from Isca Tecnologias, used a computational approach called machine learning to develop a software program that learns to recognize the wingbeat patterns of each species after it is exposed to a few examples.
While they were working on the programming, the researchers began to create reference libraries to teach the program to identify the wingbeat frequencies of different species. The researchers have already evaluated the program’s ability to identify at least six mosquito species. The success rate was generally high, ranging from 80% for fruit flies of the species Drosophila simulans to 99% for Aedes aegypti, according to a paper published in the Journal of Insect Behavior in 2014.
Refinement and an app
Batista and Mafra-Neto calculate that they have already invested $5 million in development of the trap, with funding from FAPESP, the National Council for Scientific and Technological Development (CNPq) and the United States government. In August 2016, the design was one of the 21 chosen from among 900 competitors to receive funding from the U.S. Agency for International Development (USAID) to combat the Zika virus. The researchers will have $500,000 to refine the trap.
The money is already earmarked. Mafra-Neto and his group at Isca Tecnologias are expected to finalize the development of a long-acting bait, a blend of compounds that attracts mosquitoes and repels pollinators. In São Carlos, Batista and his team will work on refining the sensor to lower its cost, and on finalizing a mobile app through which a cell phone can receive information on mosquito species and their population density in the areas monitored by the trap. The app is also expected to provide data on insect behavior and tips for controlling them. “This type of information can give people an incentive to control mosquitoes and their eggs at home,” Batista says.
The goal is to obtain an affordable product that can be used easily by health officials and ordinary individuals. “Today, mosquitoes are counted and identified manually by experts in taxonomy and entomology,” Mafra-Neto explains. “These professionals are a scarce and costly resource, and this causes bottlenecks in detecting disease-transmitting vectors.”
The traps currently available for monitoring the insects only catch them. “I don’t know of any trap that does automatic identification,” says biologist Delsio Natal, an expert in the ecology of mosquitoes of the family Culicidae and retired professor from the USP School of Public Health in São Paulo. “If the design works, it will be a groundbreaking achievement,” he says. USP biologist Margareth Capurro, who developed an Aedes strain that is genetically altered to produce sterile males, notes that a trap that identifies mosquitoes will be able to detect when a new species enters the area. “That type of monitoring is important, although it doesn’t yet allow us to know whether the mosquitoes are infected,” she says.
The current version of the trap sells for $100, and the researchers want to lower the cost. “We’re close to obtaining a version that can go out to the international market,” says Mafra-Neto.
Intelligent sensor for controlling agricultural pests and disease-vector insects (FAPESP-PPP/2012) (nº 2012/50714-7); Grant Mechanism Regular Research Grant; Principal Investigator Gustavo Enrique de Almeida Prado Alves Batista (USP); Investment R$ 137,402.06.
CHEN, Y. et al. Flying insect classification with inexpensive sensors. Journal of Insect Behavior. V. 27 (5). p. 657-77. September 2014.