AI Trained On Vast ECG Data Helps Save Lives In Breakthrough Study

The use of an artificial intelligence (AI) system has proved to be a lifesaver in a clinical trial held across two hospitals in Taiwan.

The study used AI-enabled electrocardiogram (ECG) to identify hospitalised patients who were at a high risk of dying. The AI system threw up an alert, including an AI report and warning messages to the attending physicians, who subsequently attended to the patient.

It helped cut down mortality among high-risk patients by as much as 31 per cent. The overall reduction of mortality was 17 per cent.

“That’s remarkable and as good or better than our most effective medical treatments,” Eric Topol, MD, professor, and executive vice president at Scripps Research in California, wrote in his blog “Ground Truths.”

An ECG is a quick test that records the electrical activity of one’s heart. Doctors use it to make an assessment of how the heart is functioning.

The randomised clinical trial, or RCT, with AI-ECG involved 39 physicians and 15,965 patients with an average age of 61 years. RCTs are considered the gold standard for medical research.

The AI-ECG RCT was conducted between December 2021 and 2022. The study, led by the National Defense Medical Center, Taiwan, was published in the Nature Medicine journal on 29 April 2024.

The AI system was trained on more than 450,000 ECG tests along with the survival data of the patients whose ECG was taken. Its superpower is to pick up on subtle signs that humans might miss. It came up with a percentile score representing each patient’s risk of death. A percentile of 95 or higher was deemed high risk.

Patients participating in the trial would get their ECG test done. The AI would analyse the results. About half of the patients were evaluated by physicians alone. Roughly the other half got the assistance of AI alerts. Once a patient was identified as high-risk and, therefore, requiring prompt attention, the physician would be alerted on their smartphone.

The message would say that an ECG received for a certain patient indicates a high risk of mortality and that the physician should intensively attend to the patient’s conditions.

For further verification, the physician could cross-check the ECG with the result of the AI-ECG prediction if there was a need for it.

Comparison of the death rates in the two groups at 90 days showed that while 3.6 per cent of participants in the AI group had died from any cause, 4.3 per cent died in the control group.

Notably, the death rate was 31 per cent lower among high-risk cases where the AI alert system intervened.

A significant implication of using AI was that patients predicted to be in the high-risk category received swift attention and further testing and medical treatment, including a transfer to the intensive care unit.

Upon receiving an AI-ECG alert, doctors in this study were asked “to take the high-risk alert seriously and decide on the best care for patients by themselves.”

It led to a dramatic reduction in deaths from heart issues among high-risk patients by over 90 per cent. Patients who showed less obvious signs of illness benefitted especially greatly.

The paper in Nature Medicine said that the observed “results indicate that such implementation assists in the detection of high-risk patients, prompting timely clinical care and reducing mortality.”

Fourteen military hospitals in Taiwan are already making use of the AI alert system due to the “impressive results and potential to greatly improve medical care.”

“This represents a major milestone in medical A.I. While previous randomized trials have shown improved detection or accuracy endpoints, this one takes it a big step further,” Topol writes.