The Government of the City of Buenos Aires (CABA) has implemented an Artificial Intelligence system called IATos, with the purpose of improving the current COVID-19 testing system.
IATos is an AI system, which works from a neural network and is capable of classifying voice, breathing and cough sounds. Using machine learning, it can determine from the sound of a cough whether a person should be tested for COVID-19.
In this way, the Government of the CABA has incorporated IATos, in the triage process to find out if a person has symptoms of COVID-19 or had close contact with a positive case. The process is as follows, the user records and sends an audio coughing to Boti (11-5050-0147). IATos analyzes the sound and determines if the audio matches patterns of positive cases and if so, recommends the user to take a COVID-19 test.
For the training of the AI and machine learning algorithm, a database of 140,000 cough audios of negative and positive people to COVID-19 was collected. In this way, a pattern was detected in the sound of the coughs of people with positive results in their COVID-19 tests.
The technology was later evaluated in a cough audio test of 2,687 people. IATos recommended 554 people to undergo a coronavirus test, of which 436 were positive and 118 negative.
Likewise, the IATos database was collected in its entirety by the city government and is also an open domain, so that all the data related to the project can be consulted to implement similar systems in other regions.
“Innovating is taking advantage of technology to improve the lives of the residents of the City. That is what we are doing through this artificial intelligence project and in every opportunity we have to generate tools based on the latest technologies”, explained Diego Fernández, secretary of Innovation and Digital Transformation of the CABA.
The CABA Government explains that IATos is a scalable, versatile and also low-cost tool to improve the COVID-19 testing strategy in the city. It is also a complementary service for Boti's triage system, which based on questions about COVID-19 symptoms, the chatbot quickly assesses whether the user is at risk of being infected.
