COVID-19 has proven extremely difficult to contain for many reasons, but one of the most prominent factors is how differently it manifests itself in various people. Some will never even notice they contracted the virus, many will only exhibit mild symptoms, and others will develop severe respiratory disease and breathing difficulties. The entire medical community is scrambling to understand and counteract the coronavirus, and it’s been frustrating over the past few weeks to see so many doctors, scientists, nurses, and other health workers express a sentiment of helplessness.
On a positive note, researchers from New York University’s Grossman School of Medicine and Courant Institute of Mathematical Sciences have developed an experimental new artificial intelligence tool that was able to accurately predict which newly infected COVID-19 patients would end up developing severe respiratory symptoms.
This project was conducted in collaboration with two medical facilities in Wenzhou, China; Wenzhou Central Hospital and Cangnan People’s Hospital.
You may be thinking that this AI tool probably just picked out older individuals or patients with prior health conditions, but that wasn’t the case. Even the technology’s creators were surprised by both the device’s accuracy and its discovery that three main unexpected markers appear to indicate future severe COVID-19 symptoms: high levels of the liver enzyme alanine aminotransferase (ALT), self-reported muscle pain, and high levels of the protein hemoglobin.
“While work remains to further validate our model, it holds promise as another tool to predict the patients most vulnerable to the virus, but only in support of physicians’ hard-won clinical experience in treating viral infections,” says corresponding study author Megan Coffee, MD, Ph.D., clinical assistant professor in the Division of Infectious Disease & Immunology within the Department of Medicine at NYU Grossman School of Medicine, in a press release.
“Our goal was to design and deploy a decision-support tool using AI capabilities – mostly predictive analytics – to flag future clinical coronavirus severity,” adds co-author Anasse Bari, Ph.D., a clinical assistant professor in Computer Science at the Courant Institute. “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.”
In combination with other predictors, researchers believe the AI is capable of assessing an individual’s odds of developing ARDS with up to 80% certainty. ARDS, or Acute Respiratory Distress Syndrome, is the medical term for fluid build-up in the lungs that often proves fatal among the elderly.
Scientific, radiological, and demographic information was compiled from 53 COVID-19 patients that tested positive for the virus in January at the two aforementioned Chinese hospitals. In the majority of these patients, their early symptoms were very mild. A few, however, developed severe issues like pneumonia within a week.
A series of computer models were then designed to make patient outcome predictions based on that data. The AI was even constructed in such a way to become “smarter” if given access to more information. The methodology behind the AI is incredibly complex. Essentially, it follows a “decision tree” process that models predictions and consequences based on prior choices and patient profile options.
As mentioned earlier, the study’s authors were surprised that the AI didn’t consider obvious markers like lung image patterns, fever, or overall immune system strength as useful predictors of future severe lung disease among COVID-19 patients. The AI didn’t even focus all that much on age and gender.
Let’s take a closer look at the three factors, according to the AI, most reliable when it comes to predicting severe COVID-19 induced lung disease; high ALT levels, muscle pain, and high hemoglobin levels.
ALT levels usually rise significantly when the liver is damaged, but in COVID-19 patients ALT levels were only slightly higher. Nonetheless, these fluctuations were still identified as a key predictor of symptom severity. Meanwhile, deep muscle pain and aches were probably the least surprising of the three main predictive elements. Muscle pain has long been established as a coronavirus symptom.
That brings us to the final main predictive factor, high levels of hemoglobin. Hemoglobin is an iron-carrying protein that carries oxygen throughout the body. The research team theorizes that this factor may have been influenced by some studied patients being cigarette smokers. Cigarettes have been linked to increased hemoglobin levels for some time, and such a habit would predispose a COVID-19 patient for lung problems.
These findings, as well as the invention of the AI tool in general, are incredibly promising. But, the study’s authors caution their work was fairly limited in its scope due to the small number of studied patients. Additionally, none of the studied patients were all that old, with their median age being 43.
“I will be paying more attention in my clinical practice to our data points, watching patients closer if they, for instance, complain of severe myalgia,” Dr. Coffee concludes. “It’s exciting to be able to share data with the field in real-time when it can be useful. In all past epidemics, journal papers only published well after the infections had waned.”
The full study can be found here, published in Computers, Materials & Continua.