AI, NLG, and Machine Learning

How AI Can Predict the Severity of Coronavirus

In the wake of the coronavirus pandemic, artificial intelligence (AI) has stepped up in many ways to help combat the potentially deadly virus.

By Tauren Dyson
May 20, 2020

In the wake of the coronavirus pandemic, artificial intelligence (AI) has stepped up in many ways to help combat the potentially deadly virus.

Some research uses AI to sort through the socioeconomic factors that put people at a higher risk for contracting coronavirus.

About 80 percent of coronavirus patients have only mild symptoms. But for others, the virus can be debilitating.

In the past, AI tools have been used to examine retinal images to predict heart attack risk and chest X-rays for tuberculosis. The researchers combined the use of AI with decision trees, a predictive analytics method, to produce their findings.

The researchers of this study wanted to create an AI tool that uses existing patient data to help doctors make clinical decisions.

Anasse Bari, a researcher at the Courant Institute, and Megan Coffee, a researcher at New York University (NYU), worked with doctors from Wenzhou Central Hospital and Cangnan People’s Hospital in China to publish a study, titled “Towards an Artificial Intelligence Framework for Data-Driven Prediction of Coronavirus Clinical Severity.”

The team of researchers created an AI tool that could accurately forecast the likelihood of developing severe respiratory disease in patients newly infected with SARS-CoV-2, or coronavirus.

“Our goal was to design and deploy a decision-support tool using AI capabilities—mostly predictive analytics—to flag future clinical coronavirus severity,” Bari said in a press release. “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.”

The team suggested that predictive analytics could be used in practical clinical settings by doctors to figure out whether a coronavirus patient is likely to become severely ill or not.

“Predictive analytics (a form of artificial intelligence) learns from historical data to help predict future outcomes,” the researchers said. “The technology uses machine learning algorithms that can extract insights and rules from experience (historical examples) in order to determine data attributes (features) with the most predictive power for making accurate predictions.”

The study included 33 men and 20 women, at a median age of 43. None of the patients were over the age of 67. Their symptoms ranged from coughing, fever, and upset stomachs. But a small number of patients had symptoms such as dyspnea, nasal congestion, myalgia, nasal congestion, sore throat, and wheezing.

The researchers built computer models that absorbed patient data. Then the models used decision trees to follow a sequence of choices based on multiple options at every point.

They discovered that the normal coronavirus symptoms, like fever and specific characteristics in lung images—such as ground-glass opacities and strong immune responses—weren’t helpful in telling which of the 53 patients would go on to develop severe lung disease. Age and gender didn’t matter either.

Blood analytics, like immune system cell levels and hemoglobin, along with physical symptoms and radiological results, were also measured to assess the patients’ conditions.

The findings showed that slight changes in liver enzyme alanine (ALT) levels, patient-reported myalgia (muscle aches), and elevated hemoglobin levels were the best predictors of which patients would develop serious respiratory distress.

With those indicators, plus other factors, the researchers were able to predict 80 percent of the patients who went on to develop acute respiratory distress syndrome (ARDS).

“Just as predictive text is intended to augment but not replace writers, the goal is not to create a black box to supersede clinical reasoning but to create models that can provide insight,” the researchers wrote. “Further refinement of these models with more data, from different settings with different spectrums of severity, would strengthen the predictive power of the model and allow it to be a useful tool.”

The researchers acknowledge this AI tool is just a prototype, and they note about their small-scale study, “Further validation and refinement of this model for a wider clinical spectrum will require further clinical datasets with diverse disease presentations, which we look to include in the near future.

“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.”

At the time of this writing, on April 24, 2020, 2,658,062 coronavirus cases have been confirmed around the globe, including 48,061 deaths.

“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,” Coffee said. “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.”

Leave a Reply

Your email address will not be published. Required fields are marked *