Algorithms created by DCC researchers use satellite images and can assist public health policies for epidemiological control and surveillance
With the coronavirus pandemic, many stopped worrying about diseases that for many years have also brought discomfort, sequelae and even death. In the years 2010, 2013, 2016 and 2019, Minas Gerais faced Dengue epidemics, and also outbreaks of Zika and Chikungunya, which took the lives of many miners. According to data from the State Department of Health of Minas Gerais (SES-MG), from 2019 to November 2021, the numbers of infected and dead people have been falling, but this year alone, we have already had 84,636 probable cases (notified cases except those discarded) of dengue. Of this total, 15,255 cases were confirmed for the disease and 15 deaths were confirmed. In relation to Chikungunya Fever, 6,323 probable cases were registered and, of this total, 5,344 cases were confirmed, with one death confirmed. Regarding Zika, 95 probable cases were registered and, of this total, 25 were confirmed. No deaths from Zika have been confirmed in Minas Gerais to date. There is also concern about Yellow Fever, which is also transmitted by Aedes aegypti. Even with better data than in previous years, according to experts, this does not mean that the population should leave care aside, as the outlook across Brazil is worrying.
Thinking about the public health problems related to these diseases and ways to identify areas at risk for Aedes aegypti infestation, researchers from the Faculty of Public Health (FSP) and the University of São Paulo (USP) established a partnership with scientists from the Department of Sciences of Computing at the Federal University of Minas Gerais (DCC/UFMG), coordinated by professor Jefersson Alex dos Santos. The article that describes a first phase of the study, entitled “Water tank and swimming pool detection based on remote sensing and deep learning: Relationship with socioeconomic level and applications in dengue control” has just been published in the prestigious online scientific journal PLOS ONE, under the responsibility of the Public Library of Science.
Started at the end of 2020, the analyzes are initially being carried out in the municipality of Campinas, in the state of São Paulo. “This first article covers four regions of Campinas, São Paulo, characterized by different socioeconomic contexts. With image mosaics obtained by a camera carried by unmanned aerial vehicles, we developed algorithms based on deep learning to detect water tanks and swimming pools. An object detection model, initially created for areas in Belo Horizonte, Minas Gerais, was improved with transfer learning techniques, which allowed us to detect objects in Campinas with fewer samples and more efficiency,” said Jefersson.
To develop the software, the DCC team uses Artificial Intelligence and deep learning, classifies remote sensing images, in addition to Bayesian modeling that relates the number of Aedes aegypti females, as well as cases of Dengue, Zika and Chikungunya with socio-environmental characteristics. . “Studies have shown that areas with lower socioeconomic classification are often more vulnerable to dengue fever and other similar deadly diseases that can be transmitted by mosquitoes. This first study aimed to detect, in digital images, water tanks installed on roofs and swimming pools to identify and classify areas based on socioeconomic index, to assist public health programs in controlling diseases linked to the Aedes aegypti mosquito”, says the teacher.
According to the researchers, studies carried out (bv-cdi fapesp) in the state of São Paulo are extremely fragmented, almost always without relating the vector, the population and the environment. Furthermore, the Aedes aegypti mosquito plays a fundamental role in the spread of all these diseases and there is great difficulty in identifying risk areas, based only on traditionally used entomological indicators (Breteau, Building and Containers). “Our objective in this work is to develop a model to identify high-risk areas for infestation by Aedes aegypti and the occurrence of arboviruses (DEN, ZIK and CHIK) based on the quantification of adult females of the vector, the physical, economic, social and climatic characteristics of the regions”, describe the associated researchers.
The research aims to develop methodologies to identify high-risk areas, in addition to finding a spatial pattern. In this way, these methods, as well as the results, after validation, can be used in public health management, optimizing resources and time in identifying areas where diseases occur, in addition to applying surveillance and control measures in these regions. . “With the help of computing through deep learning, we intend to create a useful tool for controlling Aedes aegypti and helping in efforts to prevent diseases caused by mosquitoes. With this first study, we concluded that it is possible to detect objects directly related to the socioeconomic level of a given region from digital images, which encourages the practicality of work aimed at public health”, celebrates the professor.
The authors of the project are researchers Higor Souza Cunha, from the Department of Electrical Engineering, Escola Politécnica, University of São Paulo, Brazil; Brenda Santana Sclauser, from the Department of Electrical Engineering, Escola Politécnica, University of São Paulo, Brazil; Pedro Fonseca Wildemberg, Department of Computer Science, Federal University of Minas Gerais, Belo Horizonte, Brazil; Eduardo Augusto Militão Fernandes, Department of Computer Science, Federal University of Minas Gerais, Belo Horizonte, Brazil; Jefersson Alex dos Santos, Department of Computer Science, Federal University of Minas Gerais, Belo Horizonte, Brazil; Mariana de Oliveira Lage, from the Postgraduate Program in Environmental Sciences (PROCAM), Institute of Energy and Environment, University of São Paulo, Brazil; Camila Lorenz, from the Department of Epidemiology, Faculty of Public Health, University of São Paulo, Brazil; Gerson Laurindo Barbosa, from the State Department of Health, Superintendency of Endemic Disease Control, São Paulo, Brazil; José Alberto Quintanilha, from the Scientific Division of Environmental Management, Science and Technology, Institute of Energy and Environment, University of São Paulo, Brazil; and Francisco Chiaravalloti-Neto, from the Department of Epidemiology, Faculty of Public Health, University of São Paulo, Brazil.
This article had repercussions in various media, see the “Press page”.