Researchers at the University of Massachusetts Amherst have invented a portable surveillance device powered by machine learning – called Fluison – that can detect cough and congestion sizes in real-time, then on flu-like illnesses and influenza trends. Analyzes data directly for tracking.
The creators of Flissance say that new edge-computing platforms for use in hospitals, the concept of healthcare waiting rooms and large public spaces, health monitoring used to forecast seasonal flu and other viral respiratory outbreaks, such as epidemics or SARS. Can expand the arsenal of devices.
Such models can be life-saving by directly informing the public health response during a flu pandemic. These data sources can help determine the time for flu vaccine campaigns, potential travel restrictions, allocation of medical supplies and more.
“We can allow predicting flu trends in a much more accurate way,” says co-author Tauhidur Rahman, assistant professor of computer and information science, who is a Ph.D. Student and lead author Forsad Al Hussain. The results of his Fluento study were published on Wednesday in Compute Machinery Proceedings on interactive, mobile, wearable and ubiquitous technologies.
To give his invention for a real-world effort, the Fluence inventors hired University Health Services Executive Director Dr. Partnered with George Corey; Biomastician Nicholas Reich, director of the UMass-based CDC Influenza Forecast Center of Excellence; And epidemiologist Andrew Lover, a vector borne pathologist and assistant professor in the School of Public Health and Health Sciences.
The Fluison platform processes low-cost microphone arrays and thermal imaging data with a Raspberry Pi and neural computing engine. It does not store personally identifiable information, such as speech data or iconic images. At Rahman’s Mosaic Lab, where computer scientists develop sensors to observe human health and behavior, researchers first developed a laboratory-based cough model.
They then trained deep neural network classifiers to create bounding boxes on thermal images representing people, and then to count them. “Our main goal was to create a predictive model at the population level, not the individual level,” says Rahman.
He placed the fluency devices enclosed in rectangular boxes about the size of a large dictionary in four healthcare waiting rooms at the healthcare clinic of UMus University.
From December 2018 to July 2019, the Fluence platform analyzed and collected more than 350,000 thermal images and 21 million non-speech audio samples from public waiting areas.
Researchers found that Fluison at the University’s clinic was able to accurately estimate the rates of daily illnesses. Many more complementary sets of Fluinson indications are “strongly correlated” with laboratory-based testing for flu-like diseases and influenza.
According to the study, “early symptomatic information captured by Fluitsen can provide valuable additional and complementary information for current influenza prediction efforts,” such as the FluSight Network, a multi-disciplinary consortium of flu prediction teams, including Reich Lab is also included. UMass Amherst.
“I was interested in non-speech body sounds for a long time,” says Rahman. “I felt that if we could catch the sound of coughing or sneezing from public places where a lot of people naturally congregate, then we use this information as a new source of data to predict epidemiological trends can do.”
Al-Hussein states that Fluison is an example of the power of combining artificial intelligence with edge computing, the frontier-pushing trend that enables data to be collected and analyzed at the source of data. “We’re trying to bring machine-learning systems to the edge,” says Al Hussain, pointing to compact components inside the Fluence equipment. “All processing happens right here. These systems are becoming cheaper and more powerful. ”
The next step is to test Fluison in other public areas and geographic locations.
“We have an initial belief that coughing is actually correlated with influenza-related illness,” says Lover. “Now we want to verify this beyond the establishment of this specific hospital and show that we can generalize across locations.”