Brazilian AI uses drones to map water reservoirs and fight dengue fever

Brazilian researchers have developed a computer program that is able to identify, through aerial photographs, water tanks on rooftops or panels and outdoor swimming pools. Artificial intelligence. The suggestion is to use this type of image as an indicator of areas that are particularly prone to mosquito infestation. Aedes aegyptiwhich transmits diseases such as dengue fever, Zika and chikungunya.

In addition, the strategy emerges as a potential alternative to dynamic social and economic mapping of cities – a win-win for various public policies.

search, supported By FAPESP, it was carried out by specialists from the University of São Paulo (USP), the Federal University of Minas Gerais (UFMG) and Endemic Disease Control Supervision (Sucen) of the Ministry of Health of the State of São Paulo. The results were published in the magazine PLUS ONE.

“What we did in this first moment was to create a model based on aerial images and computer science to detect water tanks and swimming pools, and use it as a socio-economic indicator,” says the engineer. Francisco Chiaravaluti NettoProfessor, Department of Epidemiology, USP College of Public Health.

In the published article, he and his colleagues note that previous surveys have already shown that poorer areas of municipalities are often more susceptible to dengue. That is, using a dynamic model to update the socio-economic situation – particularly in comparison with the 10-year-delayed census – would help prioritize prevention efforts for dengue, Zika and chikungunya.

“This is one of the first steps of a broader project,” highlights Chiaravalloti Neto. Among other things, the group aims to incorporate other elements to be detected in the images and to determine the true infection rates in Aedes aegypti In a specific area to improve and validate the model.

“We hope to create a flowchart that can be applied in different cities to find areas at risk without the need for home visits, a practice that takes a lot of time and public money,” says Chiaravalloti Neto.

machine learning

at Previous studyThe group previously used artificial intelligence to identify water tanks and swimming pools in Belo Horizonte (MG). The researchers began by submitting satellite images of the mining town to a computer algorithm and indicating which ones had these facilities.

through the process deep learning (or deep learning), the program has begun to identify patterns in images that indicate the presence of a swimming pool or water tank. Over time, the system was able to distinguish these structures in the images on its own.

“It’s really a process of machine learning, a sub-area of ​​artificial intelligence,” explains Jefersson Alex dos Santos, professor in the Department of Computer Science at UFMG and founder of the Earth Observation Pattern and Recognition Laboratory.

For the current research, specialists have identified four regions of Campinas characterized by different socio-economic contexts, according to the Brazilian Institute of Geography and Statistics (IBGE). A drone with a high-resolution camera flew over these areas and took a series of photos. Therefore, a database was created for water tanks and another for swimming pools.

The next step was to make a learning technology transfer. “We trained this model in Belo Horizonte and applied it in Campinas,” Santos explains. With the images obtained in the city of São Paulo, the models became more reliable for the area, with the detection accuracy of swimming pools reaching 90.23% and 87.53% for the detection of exposed water tanks.

Socioeconomic Indicator

With a properly trained algorithm, the researchers used other images to calculate the concentration of exposed water reservoirs and ponds in those four preselected regions of Campinas. Crossing this information with the IBGE data, it was observed that socioeconomic indicators were lower in regions with greater concentration of water reservoirs and higher where there are more swimming pools.

Because areas that are less well organized are more susceptible to Aedes aegyptiThis model will actually help fight the diseases it spreads. “Although it is not yet the definitive methodology, we can already consider a practical and relatively simple use of software development for large-scale use, with the goal of mapping neighborhoods with the greatest risk of dengue spread,” Santos stresses.

Chiaravalloti Neto points out that the models created can be useful beyond the control of dengue fever, Zika virus and chikungunya: “Social and economic indicators for different parts of Brazil are updated every ten years, through the census. In a more flexible way, which can be used to counteract Various diseases and problems.

According to him, future work may find other signs from aerial images, thus improving these algorithms to ensure greater reliability.

Drone or satellite?

Although the aerial images of Campinas were obtained with a drone, it is expected that in the future, the strategy tested in this research will only use satellite imagery. “We used a drone because it was a study, but scanning with this equipment is expensive,” Chiaravalloti Neto analyzes.

“They also have less autonomy. To implement a large-scale project, involving large cities, we will need satellite images,” Santos continues.

In the study in Belo Horizonte, satellite images were used successfully – they need high resolution so that a computer can identify patterns. Fortunately, access to this type of image is expanding, according to Santos.

Although this type of methodology appears costly, it generates potential savings by eliminating the need for face-to-face visits to map, house to house, dengue-prone areas. Instead, health agents will take advantage of information obtained remotely – and processed using artificial intelligence – to go to priority locations with more packets.

Next steps

The current model is able to detect water tanks, but not if they are properly closed. Something similar applies to swimming pools: he defines them, but he doesn’t know if they are well maintained or closed. “This methodology can be improved to distinguish structures that are well preserved from those that could actually be breeding grounds for mosquitoes,” says Chiaravaluti Netto.

Charging these and other combinations of patterns associated with increased mosquito prevalence would make the algorithm more reliable for identifying public health measures.

Currently, researchers are installing a series of traps for Aedes aegypti In about 200 buildings in Campinas, the conditions of the property and the presence of different mosquito breeding sites were evaluated in detail. Social and economic indicators will also be examined.

The next step will be to evaluate the aerial photographs of these areas with the same logic used in the research mentioned above to classify the degree of danger to the existence of these areas. Aedes aegypti and the diseases it transmits.

“By monitoring these blocks, we intend to build a model for prioritizing dengue control for the entire city, and later for the rest of Brazil,” concludes Chiaravaluti Neto.

In addition to funding from FAPESP, the researchers had resources from the Serrapilheira Institute, the National Council for Scientific and Technological Development (CNPq), the Dean of Research at the USP and the Minas Gerais Research Support Foundation (Fapemig). Sucen also provided structural support.

The authors involved are: Hegor Souza Cunha, Brenda Santana Schlauser, Pedro Fonseca Wildemberg, Eduardo Augusto Militao Fernandez, Jefferson Alex dos Santos, Mariana de Oliveira Lage, Camila Lorenz, Gerson Lorendo Barbosa, Jose Alberto Quintanilla and Francisco-Cheravalotti.

Article Water tank and pool discovery based on remote sensing and deep learning: relationship to socioeconomic level and applications in dengue control. It can be accessed at: https://journals.plos.org/plosone/article/authors?id=10.1371/journal.pone.0258681.

* This text was originally posted by FAPESP . agency according to Creative Commons CC-BY-NC-ND license. Read the original here.