Artificial Intelligence model shows Covid transmission chances in rooms, Health News, ET HealthWorld

Roorkee: Researchers at Central Building Research Institute (CBRI) in Roorkee have developed an Artificial Intelligence (AI) model to predict the “transmission probability” of Covid-19 in a closed space in a building.

The model uses an electronic device to detect the carbon dioxide concentration, temperature and humidity of a room. These and other input parameters are used to show the probability of the presence of the Covid-19 virus in an office, classroom or any other closed space in a building.

After computing the parameters, the software determines the transmission probability and displays the results in the form of a text alert on the screen. The study of the findings, called “Transmission Probability Of SARS-Cov-2 In Office Environment Using Artificial Neural Network”, was recently published by IEEE Access, an open-access scientific journal in America. Anuj Kumar, principal investigator and one of the authors of the study, said the research is “a first-of-its kind.”

As per the study, 11 input parameters were used to predict the R-value, which refers to “expected number of new infections that arise in any event occurring over a total time in any space”. The parameters are listed as: indoor temperature (TIn), indoor relative humidity (RHIn), area of opening (AO), number of occupants (O), area per person (AP), volume per person (VP), CO2 concentration (CO2), air quality index (AQI), outer wind speed (WS), outdoor temperature (TOut) and outdoor humidity (RHOut).

“Using a datalogger, we take the readings of CO2 concentration, temperature and humidity. Other equipment provides us readings of wind speed and AQI. Some of the other parameters are calculated manually,” said Kumar, who is also a consultant in the Building Energy Efficiency department of the institute.

According to the study, “real-time data for the office environment was gathered in the spring of 2022 in a naturally- ventilated office room in Roorkee under composite climatic conditions”.

The data was fed into two models, artificial neural network (ANN), a computer-based mathematical modelling, and one that uses more traditional techniques, the curve-fitting model, a mathematical analysis tool.

“We determined correlation coefficients for both models, 0.9992 for ANN and 0.9557 for curve fitting. As the value of these coefficients go down (say 0.90992 for ANN), the chances of the virus’s transmission go up,” Kumar said. “We established a link between CO2 concentration and R-event as a model for prediction purposes,” he added.