The Covid-19 pandemic has undermined many health care systems around the world, resulting in the emergence of overloaded hospital environments. Researchers from all over the world are still working on efficient detection methods of this SARS-Cov-2 virus. Health professionals currently use several approaches, such as Nucleic Acid Amplification Tests (NAATs), which have a high sensitivity to detect the virus, and Serological Tests (ST) looking for antibodies in our bodies. “However, these methods require a physical consultation, which increases the risk of infection for staff and patients, and consume significant amounts of health system resources”, explains Muhannad Ismael.
In this context, the CDCVA research project, funded by the National Research Funds, aims to develop an unprecedented classification system detecting COVID-19 infection probability based on audio signature. “Respiratory conditions, such as dry cough, sore throat, excessively breathy voice and dyspnoea, caused by Covid-19, can make patients' voices distinctive, creating identifiable voice signatures, that may be discovered using our system”, details Muhannad.
Consequently, the CDCVA project, coordinated by LIST, would open the path to a better remote diagnosis for health professionals, while also minimising physical contacts associated risks between staff and patients. “Our classification system could also be applied in emergency services, which receive a large number of calls during such epidemic, in order to identify critical cases requiring a rapid intervention”, explains Muhannad.
As coordinator of the CDCVA project, LIST is closely working with the University of Luxembourg, and their non-contracting partner, the Luxembourg Institute of Health (LIH), which provides data through its “Predi-COVD cohort study”.
In order to develop such audio COVID-19 signature detection system, researchers must collect a large dataset of voices and coughs recordings. These data are processed to remove any noise and to identify cough and voice patterns, using advanced machine learning methods.
“By making use of our strong expertise in data processing and machine learning, our team will work on the classification system detecting COVID-19 infection probability. We will also develop a web-based platform to collect the data, considering both privacy and ethical issues”, adds Muhannad.
“I am the Principal Investigator (PI) of the CDCVA project. I have therefore various management tasks through this research, including coordinating between partners. Moreover, I am closely working with our partner, the University of Luxembourg, in a specific task, which consists of designing and developing a classification system based on artificial neural network. In other words, this is an artificial intelligence technology inspired by our human brain to learn.”
Muhannad Ismael joined LIST as junior Research & Technology Associate in 2019 within the IT for Innovative Services department. He holds a bachelor’s degree in Computer Science and Artificial Intelligence from the University of Aleppo (Syria) and a master's degree from the University of Limoges (France). Muhannad also has a PhD’s degree in Computer Vision from the University of Reims (France). He worked as a postdoctoral researcher at École Centrale de Nantes (France) until 2018. His professional experience in working for private and public sectors led his research interests to be more diverse, ranging from mixed reality and machine learning to computer vision.