Welcome Your IET account
crowd

Flu-like trends detected from coughing crowds

University of Massachusetts-Amherst scientists have created a portable device which can detect coughing and crowd size, and analyse this data to monitor trends in flu-like illnesses.

The platform, which is called 'FluSense', uses machine learning to detect patterns from the data it collects. It is intended for use in hospitals and medical waiting rooms, as well as in larger public spaces.

The researchers hope that FluSense will expand the health surveillance toolkit available for forecasting seasonal flu and other viral respiratory outbreaks, such as the Covid-19 pandemic. As governments struggle to contain the virus through extensive testing and drastic lockdowns unprecedented outside wartime, the need for tools to monitor and predict infection trends has been thrown into sharp relief.

Models informed by FluSense could directly inform public health responses during viral epidemics, such as by determining the timing for flu vaccine campaigns, travel restrictions, and allocation of essential medical supplies like masks, testing kits and ventilators.

“[FluSense] may allow us to predict flu trends in a much more accurate manner,” said Professor Tauhidur Rahman, assistant professor of computer and information science at the university.

The platform includes a low-cost microphone array and thermal camera. Data from these sensors are processed with a Raspberry Pi and neural computing engine, storing no identifiable information. The researchers developed a lab-based cough model, then trained an onboard artificial neural network to identify the areas on thermal images corresponding with people, and to count them.

They placed the devices in four healthcare waiting rooms at the university’s Health Services clinic from December 2018 to July 2019. During this time, they collected and analysed more than 350,000 thermal images and 21 million audio samples.

The researchers found that FluSense was able to make accurate predictions about daily illness rates at the clinic, with its findings “strongly correlated” with lab-based tests for flu-like illnesses, including influenza itself.:

“The early symptom-related information captured by FluSense could provide valuable additional and complementary information to current influenza prediction efforts,” the researchers wrote.

According to lead author Forsad Al Hossain, FluSense is an example of what can be achieved by combining AI with edge computing (gathering and processing data at its very source). “We are trying to bring machine learning systems to the edge. All of the processing happens right here. These systems are becoming cheaper and more powerful,” he said.

Next, the researchers plan to test FluSense in other public areas to demonstrate that it works just as well in non-hospital settings.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles

Info Message

We use cookies to give you the best online experience. Please let us know if you agree to all of these cookies.


Learn more about IET cookies and how to control them