Mosquitoes spread disease, but this is how AI is biting back

Machine learning and satellite comms are helping identify mosquitoes and their infection potential.
Written by Anna Solana, Contributor

Mosquitoes aren't just a nuisance. Along with their itchy bites, they spread serious diseases, like malaria, dengue or zika, which, as Microsoft co-founder Bill Gates has pointed out, kill more people every day than sharks have managed to do in an entire century.

A Yale report released last year also suggests that climate change is going to make the problem worse. So, having early detection systems to promptly deploy preventive controls is crucial to protecting people.

The Institute of Agrifood Research and Technology (IRTA) in Catalonia, Spain, has started to use artificial intelligence (AI), sensors, and satellite communications to automate the process of trapping mosquitoes and classifying them according to species, sex, age, and their potential for causing infection.

SEE: How to implement AI and machine learning (ZDNet special report) | Download the report as a PDF (TechRepublic)

Traps are not a new control technique. In many countries, manual field trap inspections are conducted by trained researchers or technicians.

However, these inspections are time-consuming, costly, and classifying mosquitoes accurately by eye requires a great deal of experience, Dr Sandra Talavera, researcher at the Animal Health Research Center IRTA-CreSA, tells ZDNet.

That is one of the reasons why IRTA, owned by the government of Catalonia, has decided to adopt the Vectrack system, funded by the EU Horizon 2020 program.

The first test phases with traps are being carried out by the Barcelona Public Health Agency (ASPB), which is responsible for monitoring and controlling mosquitoes in the Mediterranean city's urban environment.

Vectrack has been developed by Irideon, a Spanish-German company headquartered in Spain, which sells sensor-based products for various sectors. The system enables researchers to combine traditional sampling processes with remote sensing and modeling techniques to develop risk maps.

The traps are similar to those already in use but they add optoelectronic sensors, enabling remote and automated counting and classification of the targeted mosquitoes "according to the frequency of flight and shape of their body", Talavera says.

Fellow researcher Dr Carles Aranda explains that mosquitoes are attracted to the traps because the devices "emit carbon dioxide as an entree and then suck the specimens inwards so that they do not escape".

The sensors also collect data on temperature and humidity related to the GPS position of the trap. They turn the morphological, physiological, and flight kinetics of mosquitoes into a digital fingerprint of the insect.

João Encarnação, Irideon chief business officer, explains that the use of sensors is more efficient than using photos or video from the traps and then analyzing the information.

The high cost of data delivery and power consumption involved in using photos or video would make the technology too expensive to implement in field surveillance.

In addition, another advantage of the Vectrack sensors is that they can be installed almost anywhere in the world as they are compatible with a range of communication protocols, including 2G, 3G, 4G, Wi-Fi, LPWAN technologies NB-IoT and LoRA, and satellite IoT.

The information obtained is sent to the cloud, analyzed by algorithms, and then processed in geographic information systems, provided by Belgian company Avia-GIS Software. The data is then integrated with satellite information from the European Space Agency (ESA).

SEE: Google reveals new Python programming language course: Scholarships for 2,500

This process allows researchers to develop real-time risk maps for regional, national, and international public-health bodies, such as the European Center for Disease Control. The maps can potentially help authorities prevent diseases and contain epidemics more quickly.

In a previous experiment, the optoelectronic sensors were already able to differentiate the species, sex, and age of the mosquitoes in an efficiency range from 61% to 99%. So, hopes are high for the new system.

And that is good news because since the Protocol for the Surveillance and Control of mosquito-borne diseases was launched in 2014, a total of 507 cases of dengue have been detected in Catalonia, two of which were autochthonous – that is, not brought in from overseas – and in Spain overall there have been seven cases of autochthonous dengue. 

Editorial standards