AI Finds New Predictive Markers of COVID-19 Severity
Researchers and healthcare organizations are putting their heads together to consider how technology could ease the ever-worsening COVID-19 global crisis. A partnership between NYU and the hospitals in Wenzhou, China are turning to artificial intelligence (AI) to help predict exactly which COVID-19 patients go on to develop severe respiratory disease.
“We hope that the tool, when fully developed, will be useful to physicians as they assess which moderately ill patients really need beds and who can safely go home, with hospital resources stretched thin.”
Symptoms for COVID-19, including cough, fever, and difficulty breathing start to appear after an incubation phase that can take as long as two weeks. According to research, around 80 percent of infected individuals experience the mild form of the coronavirus infection. Individuals at risk for developing severe complications from infection are those with underlying health complications, such as heart disease, diabetes, and cancer. However, correlating the severity of the disease with health status at the time of infection may not be as straightforward as initially thought.
Data collected from over 50 COVID-19 patients admitted to Wenzhou Central Hospital and Cangnan People’s Hospital revealed an unexpected trend. A small subset of patients who were healthy at the time of infection and experienced mild symptoms subsequently progressed to a very severe, life-threatening form of the disease. Caught off guard, physicians struggled to improve patient outcomes in these cases.
The study, published in Computers, Materials & Continua, aimed to determine whether AI-powered tools could assist in pinpointing exactly which patients are most likely to develop Acute Respiratory Distress Syndrome (ARDS) — a common cause of death in critically ill COVID-19 patients.
The research teams fed their AI-driven computer model with COVID-19 data, in the form of demographic, laboratory, and radiological findings, with the program “learning” to pick out hidden patterns connecting this data to clinical outcomes. Interestingly, features once thought to be strong predictors of…