A research team in Wuhan has built a machine learning model to predict individual mortality in COVID19 patients.
The model has selected three biomarkers: lactic dehydrogenase (LDH), lymphocytes and high-sensitivity C-reactive protein (hs-CRP) to predict mortality rates and so far the accuracy has surpassed 90 percent, according to a paper published in Nature Machine Intelligence on Thursday.
The prediction model is aimed at distinguishing potentially severe COVID-19 patients and giving them early and immediate medical attention, said the research team.
To identify meaningful markers of mortality risks, Yan Li and her colleagues from Tongji Hospital in Wuhan, Central China’s Hubei Province, analyzed the blood samples of 485 patients collected in Tongji Hospital from Jan 10 to February 18.
The authors input basic information, symptoms, blood samples and the results of laboratory tests, including liver, kidney and coagulation functions along with electrolytes and inflammatory factors, taken from originally general, severe and critical patients, and then associated the inputs with the outcomes—either surviving or dying at the end of the examination period.
It found LDH, hs-CRP and lymphocytes are the most crucial biomarkers in distinguishing patients at imminent risk.
The results are reliable given that there are hundreds of samples being tested, Wang Zhongyuan, a professor of computer science at Wuhan University told the Global Times on Saturday.
“Machine learning studies the statistical relationship between the three biomarkers and the mortality rate. It would be more scientifically valuable to reveal the relationship from the pathological point of view. It is also important to include external intervention such as treatments into the study,” he added.
photo:web
