As the whole world is paralyzed by the Covid-19 pandemic, the news is timely. American researchers have announced that they have manufactured a portable monitoring device capable of detecting cough and crowd size to monitor the spread of a virus in real time, the University of Massachusetts announced. States) March 19 in a press release. This device, FluSense, will be used in hospitals, health service waiting rooms and large public spaces. Ultimately, it could help predict seasonal flu and other viral respiratory epidemics like the one we are currently facing.
“I’ve been interested in non-vocal body sounds for a long time. I thought if we could capture the sounds of coughing or sneezing in public spaces where a lot of people gather naturally, we could use this information as a new source of data to predict epidemiological trends ”says Tauhidur Rahman, assistant professor of computer and information science, co-author of the study, in the preamble.
He and his colleagues therefore began by developing a cough model in the laboratory. Then, they trained a calculation model to draw boundaries and count thermal images of people. “Our main objective was to build predictive models at the population level, not at the individual level ”, details Tauhidur Rahman, ensuring that this platform does not store any personally identifiable information.
Accurately predict daily disease rates
The scientists then placed the devices, enclosed in a rectangular box, in four waiting rooms for health services at the University of Massachusetts in Amherst, United States. From December 2018 to July 2019, the platform collected and analyzed more than 350,000 thermal images and 21 million non-voice audio samples from public waiting rooms.
Result: FluSense successfully predicted the clinic’s daily illness rates. “Early information related to symptoms captured by FluSense could provide valuable additional and complementary information to current flu forecasting efforts ”, like the FluSight network, a multidisciplinary consortium of influenza forecasting teams, note the researchers. “This can allow us to predict flu trends much more precisely ”, welcomes Rahman.
According to Forsad Al Hossain, the study’s lead author, FluSense is a good illustration of the possibilities of artificial intelligence in the healthcare field. “We try to bring machine learning systems to the forefront, he explains, showing the press the compact components inside the FluSense device. All the processing is done here. These systems are becoming cheaper and more powerful. ”
Determine vaccination campaign schedules
From now on, FluSense should be tested in other public areas and geographic locations. “We have initial validation that coughing does have a correlation with flu-related illnesses (…) Now we want to validate it beyond this specific hospital framework and show that we can generalize in other places ”says Andrew Lover, vector-borne disease expert.
Just over a month ago, a Canadian company called BlueDot had already been talked about for somewhat similar reasons. Using online data and machine learning systems, she had detected the first signs of a coronavirus infection in Wuhan, southern China, as early as December 31, well before WHO alerted. the whole world on the subject. She then planned that the infectious agent would pass the following days from the cradle of contamination to Bangkok, Seoul, Taipei and Tokyo.
Ultimately, models of this kind could help health authorities in different countries determine the calendar for vaccination campaigns, possible travel restrictions or the allocation of medical supplies.
Covid-19: Our audio podcast, new daily news meetings