Artificial intelligence predicts critical illness in COVID-19 patients

March 31, 2020 -NYU researchers have used artificial intelligence to accurately predict whether newly diagnosed patients with COVID-19 will develop severe respiratory illness, Investigation Was announced on Computer, material, continuum.

For updates on the coronavirus, Resource page, Updated twice daily by Xtelligent Healthcare Media.

The rapid spread of COVID-19 around the world has created an urgent need. Identify which case The research team said it would escalate to a serious illness. More than 80% of cases appear to be mild, but more severe symptoms require oxygen and prolonged ventilation.

Acute respiratory distress syndrome (ARDS), a potentially fatal accumulation of fluid in the lungs in the elderly, is an important feature of COVID-19 patients who experience diminished outcomes. The team wanted to determine if AI technology could help accurately predict which patients with the virus would develop ARDS.

The researchers collected laboratory, demographic, and radiological findings from 53 patients who tested positive for COVID-19 at two hospitals in China, with a mean age of 43 years. Next, the team used that data to train AI models designed to be smarter with more data to consider.

read more: Artificial intelligence research joint target COVID-19

“Our goal was to use AI capabilities (primarily predictive analytics) to design and deploy decision support tools to flag future clinical coronavirus severity.” Said Co-author of PhD, Anasse Bari, clinical assistant professor of computer science at the NYU Courant Institute of Mathematical Science.

“I hope this tool, when fully developed, will help physicians evaluate medium patients who really need beds and who can safely return home when hospital resources are scarce. . “

The results indicate that the characteristics that may be characteristic of COVID-19 (specific patterns of lung imaging, fever, strong immune response, etc.) determine which patients with the first mild symptoms will develop severe lung disease Was not useful in predicting.

Age and gender also did not help predict serious illness, but previous studies have shown that men over 60 are at increased risk.

of AI tools Found that alterations in three functions, including levels of the liver enzyme alanine aminotransferase (ALT), reported muscle pain and hemoglobin levels, and were the most accurate predictor of subsequent serious illness. These factors allowed the team to predict ARDS risk with up to 80% accuracy.

read more: Artificial intelligence helps predict cancer treatment response

Researchers have noted that ALT levels rise dramatically when diseases such as hepatitis damage the liver. ALT levels were only slightly higher in patients with COVID-19 but still played a major role in predicting disease severity. Myalgia, or deep myalgia, is also more common in COVID-19 patients, and past studies have linked myalgia to higher inflammation in the body.

High levels of hemoglobin, an iron-containing protein that enables blood cells to carry oxygen to body tissues, are also strongly associated with subsequent dyspnea. The team noted that this could be explained by other factors, such as unreported tobacco smoking, which is associated with increased hemoglobin levels.

Some limitations of the study included a relatively small dataset and limited clinical severity of the disease in the studied population. The latter may be due to an unknown lack of elderly patients admitted to the hospital during the study.

The team said that using more data from different settings to further refine the model would help improve predictive power.

“There is still work to further validate our model, but it looks promising as another tool for predicting the most vulnerable patients to the virus, but we have drawn on physicians’ struggling clinical experience in treating viral infections. We just help, “said corresponding research author Megan Coffee. , PhD, Ph.D., Clinical Associate Professor of Medicine, and a member of NYU Langone’s Department of Infectious Diseases and Immunology.

read more: Predictive analysis, risk factors may help identify Crohn’s disease

This study is AI capabilities Predictive analytics is intended to augment and support clinicians, especially in the midst of a global health crisis.

“This study shows that predictive analytics can play a role in enhancing clinical skills in distinguishing between” disease “and what is not” disease, “the researchers said.

“ The goal is to create a model that can provide insight, rather than create a black box to replace clinical inference, just as predictive text aims to increase rather than replace authors .Clinical insights are based on both individual and collective professional learning. Machine learning can add further insights. “

The advent of AI in healthcare has opened new possibilities for tracking and monitoring pandemics. This will help improve the outcome and response of clinical care.

“In my practice, I pay more attention to the data points and watch the patient closely, for example, if she complains of severe muscle pain,” Coffee said. “It may be useful to be able to share data in the field in real time. For all epidemics in the past, journal articles have been published only after the infection has subsided.”