GeoVin: public participation in the fight against Chagas disease

COCHERO J; PATTORI LORENZO; BALSALOBRE A; CECCARELLI S; MARTI G A

Trieste

Conferencia; ECSA Conference; 2020

The commonly called ?kissing bugs? (triatomines) are insects involved in the vectorialtransmission of the parasite (Trypanosoma cruzi) that causes Chagas disease. Chagas is animportant public health issue in the Americas and, in the last two decades, came to be alsoan important issue in European and Western Pacific regions because of the non-vectorialroutes of parasite transmission (mother to child and transfusional transmission). The controland monitoring of kissing bugs is one of the current policy for all health-related actions,primarily conducted by governmental agencies, which depend on updated distribution mapsfor the insects and on their correct identification in the field by technicians.The aim of GeoVin project is to gather information about the distribution of kissing bugs inArgentina through citizen participation, and to develop tools that field technicians can use toquickly identify the vector. For this purpose, the community is engaged through a mobile appor through the project?s website, which citizen scientists can send photos of the insects theyfind and a panel of specialized reviewers from the Centro de Parásitos y Vectores from theUniversity of La Plata (CEPAVE-UNLP), that notify the user directly with the correctdetermination of the kissing bug, and instruct the citizen on the actions to perform based onthe health risk it poses. All data is georeferenced and openly shown in the project?s websitein real time, and distribution maps of the vector are supplemented with the research center?sup-to-date database of over ten thousand distribution reports from the last 20 years. TheGeoVin app was launched in May 2018 and has over 690 registered users, including agentsfrom two governmental vector monitoring agencies from the areas of Argentina mostaffected with Chagas disease, and has been endorsed by the World Health Organization inLatin America. With all the information gathered through this approach, we used aninnovative machine-learning approach (Fastai+Pytorch) to train a Convolutional NeuralNetwork (CNN) to recognize true kissing bugs directly from photos taken with cellphonecameras. The CNN results were validated using over 560 photos sent by citizen scientists,and was able to achieve correct identification rates of 93.75%, identifying true kissing bugsfrom other insects even with blurred images, poor lighting or with the presence of othersubjects and objects in the photo. These results show great promise in the development of apractical and accurate system to recognize the vector in the field without major technologicallimitations, also highlighting the potential in the contribution of citizen scientists whendeveloping tools to tackle major health threats.