AI model aids early detection of severe COVID-19 infection in children

Children have been less impacted by COVID-19 caused by the severe acute respiratory syndrome coronavirus 2 (SAR-CoV-2) than adults.

But some children diagnosed with SARS-CoV-2 have experienced severe illnesses, including Multisystem Inflammatory Syndrome (MIS-C) and respiratory failure; nearly 80% of children with MIS-C become critically ill with a 2 to 4% mortality rate.

Currently, there are no methods to discern the spectrum of the disease's severity and predict which children with SARS-CoV-2 exposure will develop severe illness, including MIS-C. Because of this, there is an urgent need to develop a diagnostic modality to distinguish the varying phenotypes of disease and risk-stratify disease.

To prevent children from becoming critically ill from SARS-CoV-2, a team of Wayne State University researchers led by Dongxiao Zhu, Ph.D., associate professor of computer science in the College of Engineering, are developing an artificial intelligence (AI) model to aid in the early detection of severe SARS-CoV2 illness in children.

Zhu is working with researchers from Central Michigan University and Penn State University, who are working to define and compare the salivary molecular host response in children with varying phenotypes of SARS-CoV-2 infections and develop and validate a sensitive and specific model to predict severe SARS-CoV-2 illness in children. They are working to develop a portable, rapid device that quantifies salivary miRNAs with comparable accuracy to predicate technology (qRT-PCR). Zhu and his team will develop an AI-assisted cloud and mobile system for early recognition of severe SARS-CoV-2 infection in children.

The research team aims to develop an innovative and efficient AI model with cloud and edge intelligence-integrating noninvasive biomarkers with social determinants of health and clinical data to aid with early detection of severe SARS-CoV-2 illness in children.

Currently, severe disease is challenging to discern given the low rate of occurrence, the spectrum of symptoms mimicking other common infections, the lag period before the development of severe illness, and the absence of a sensitive diagnostic tool. This has led to an urgent need to develop a sensitive, noninvasive, and rapid modality to predict severe illness.

Our research is critical as we expect to improve outcomes of children with severe SARS-CoV-2 infection via early recognition, timely intervention, and appropriate allocation of critical resources. The successful completion of the project will also be significant, as it will lead to the development of a rapid bedside diagnostic device and creation of patient profiles based on individual risk factors which we expect to lead to personalized treatments in the future."

Dongxiao Zhu, PhD, Associate Professor of Computer Science, College of Engineering, Wayne State University