AI tool could calculate Covid-19 variant death risk

The technology could tailor treatment and give people the best chance of recovery, according to the research funded by the British Heart Foundation (BHF).

Severe cases of Covid-19 have been associated with a cytokine storm, whereby the spike protein of the virus causes the immune system to go into overdrive and produces a surge of damaging molecules called cytokines.

By using routine chest CT scans, researchers at the University of Oxford have developed a Covid-19 signature using machine learning. It detects biological red flags in the fat surrounding blood vessels in the chest to measure the level of inflammation driven by cytokines in people infected with the virus.

“We have built an incredibly adaptable AI platform that tracks vascular disease by decoding information from blood vessel images obtained routinely during hospital admission, and integrating it with a large RNA bio-resource from human tissue biopsies,” said Charalambos Antoniades, professor of cardiovascular medicine at the university.

He added: “By simply adding in one extra step to the routine care of people admitted to hospital with Covid-19 who already have a CT scan, we can now detect patients at high risk of life-threatening complications and might tailor their treatment to aid long-term recovery.”

Antoniades, who is also a BHF senior clinical research fellow at Oxford, said that the benefits don’t stop at that. “We know that this exaggerated immune response to the virus can also cause abnormal blood clotting, and so we are developing this AI platform to identify Covid-19 patients who are most at risk of having a future heart attack or stroke.”

He added the team can also pivot their AI-based platform with ease to develop a new scanning ‘signature’ to better understand future viruses and diseases that take hold of our population.

For the study, the researchers applied the Covid-19 signature to CT chest scans of 435 people admitted to hospitals in Oxford, Leicester, and Bath, and compared the degree of inflammation and risk of death in people with and without Covid-19.

They found that for patients admitted to hospital, the level of cytokine-driven inflammation in the blood vessels was much higher in those with Covid-19, and even greater in patients infected by the Kent, or alpha, variant.

They found that those with a high level of vascular inflammation were up to eight times more likely to die in hospital and were most likely to respond well to the anti-inflammatory drug dexamethasone.

Covid-19 patients with high vascular inflammation treated with dexamethasone had a six-fold reduction in risk of dying compared to Covid-19 patients who were not given the drug, the researchers added.

The team said that by using this tool to obtain an inflammation score, patients found to have a lot of inflammation in their blood vessels, and therefore increased risk of death, might be given anti-inflammatory drugs to reduce their risk and help their long-term recovery.

Professor James Leiper, associate medical director at the BHF, said: “Over the past year we have supported our scientists to direct their expertise to help the global effort in understanding Covid-19.

“This research clearly shows that Covid-19 is a powerful virus that can wreak havoc on our circulatory system, and that we associate different variants with different levels of risk.

“There are still a lot of unknowns relating to how the virus can impact our health in the long term, but this AI tool could ultimately help to save lives.”