American and Chinese researchers announced Monday that they have developed a tool using artificial intelligence to predict which coronavirus patients will develop serious lung complications.
Once deployed, this tool could allow doctors to prioritize certain patients as health systems in many countries around the world run out of steam, said Megan Coffee, of the Grossman School of Medicine at New University. York, in the journal Computers, Materials & Continua. The tool has uncovered several indicators that strongly presume that a patient may develop acute respiratory distress syndrome (ARDS), a complication of Covid-19 that fills the lungs with fluid and kills about 50% of people who develop it.
Analysis by an intelligent algorithm of data from 53 patients with coronavirus in two hospitals in Wenzhou, China, showed that changes in the level of alanine aminotransferase, an enzyme present in the liver, the level of hemoglobin and in pain reports, were the most specific indications of a complication. With other factors, the tool made it possible to diagnose a risk of ARDS with an accuracy of up to 80%.
The symptoms considered to be markers of Covid-19, such as fever, a special image of the lungs on CT, and strong immune responses, by contrast, did not predict whether patients with a mild form of the disease could develop an ARDS. Gender or age are not useful indicators, although other studies have indicated that patients aged 60 and over are a high risk group.
“Much of the data used by the machine to influence decisions is different from what a doctor would normally see“Megan Coffee, co-author of the study, told AFP. Artificial intelligence is already used by dermatologists to predict which patients are at risk of developing skin cancer.
In the case of Covid-19, a disease still poorly understood, the tool can lead doctors in the right direction to know which patients to treat first if hospitals are overloaded with patients, assured the other responsible for the study , Anasse Bari, professor of computer science at New York University. The team is now trying to refine its tool with data from New York, the epicenter of the pandemic in the United States, hoping that it will be ready to be deployed in April.
– .
