CORDIS - EU research results

A novel smart trap station as an Internet of Things surveillance solution to remotely count and identify the species of disease-carrying mosquitoes

Article Category

Article available in the following languages:

AI improves identification of mosquitoes for better disease control

Effective control of disease vectors relies on fast, simple identification - currently missing in the regions most affected. An EU project offers an AI-driven solution.

Health icon Health

Millions of citizens globally are infected by diseases spread by infectious mosquitoes, resulting in thousands of deaths, loss of income through ill health and a drain on national health systems. There is now evidence to suggest that climate change is introducing more mosquito-borne diseases into Europe. Comprehensive attempts to tackle the problem typically rely on accurate information about mosquito populations in a given area, and which usually involve the periodic manual inspection of traps. The REMOSIS project developed smart traps to cut the investigative costs and the time commitment required. The device, called the BG-Eye, successfully used machine learning to distinguish between different mosquito species, detecting species of special interest (like invasive disease vectors), and to distinguish between the sexes of the same species - information necessary for effective preventative action. Not all species are created equal Not every mosquito species is equally interesting to researchers, especially when looking at cases of invasive species or the transmission of certain diseases. So it is important to be able to separate them. Until now, this could only be done by bringing those caught in traps to labs for manual identification. This process is clearly time-consuming and expensive, and often lends itself to inaccuracies, due to human error. BG-Eye uses Artificial Intelligence (AI) capabilities to continually improve its identification of species of interest, leaning from past efforts. As the project coordinator Dr Andreas Rose explains, “The basic training is performed in the lab with lab-reared mosquitoes, where the device learns to identify the target species. The device is then introduced into the field, where its identification capabilities are tested and adjusted, if necessary. After this, it can utilise what it has learned autonomously.” An important aspect of BG-Eye is that the device can perform when used in conjunction with commercially-available, fan-driven mosquito traps. This means that it can already be easily used alongside standard mosquito monitoring field tools. Until now, similar technologies only worked in laboratory set-ups with mosquitoes outside the airstream of a mosquito trap. The BG-Eye detector unit is an upgrade of a device already on the market called the BG-Counter, which differentiates between mosquitoes and other insects by sucking the catch into the trap through a thin light barrier. However, the resulting signal is too short to differentiate the mosquito species, and so REMOSIS set about lengthening the signal and increasing the data points, meaning the insects had to be scanned for longer than had been originally anticipated. The upgrade proved to be a success, as Dr Rose recalls, “A surprise, at least for the team biologists, was the precision with which the AI was able to separate species that are visually very hard to distinguish. After feeding the machine learning algorithms with more data, the system learned to distinguish between two Anopheles species that can only be otherwise differentiated using molecular methods.” Policing society’s immune system The monitoring of potential disease vectors is in essence protecting society’s immune system, whereby speed of threat detection results in faster control of that threat. As Dr Rose summarises, “With the REMOSIS device, the surveillance and monitoring of disease vector mosquitoes will be quicker, providing reliable information almost in real time. This allows for highly focused, efficient control activities, with greatly reduced costs, much shorter reaction times and less impact on the environment.” Currently the team are working on up to five prototypes, with the full functionality of the final product. These devices will be sent to different partners around the world to collect mosquitoes in various habitats and environments. Every count will provide a typical electronic signature for these insects, and machine learning, algorithms and AI will be used to calibrate the technology. As the mosquitoes’ signatures library grows, more complex habitats with many different mosquito species will be targeted.


REMOSIS, mosquito, malaria, disease, vectors, Zika, dengue, artificial intelligence, machine learning, algorithm, species, insects, identification

Discover other articles in the same domain of application