AI to Prevent Sudden Death: Detecting Heart Disease with 99.6% Accuracy

Artificial intelligence software that can determine if a patient is suffering from a heart attack is being trialled at Scottish hospitals in an effort to cut accident and emergency response times.

A heart attack can be hard to spot because its symptoms – including chest pain, dizziness and shortness of breath – are linked to many other conditions. Doctors often miss other vital signs. The British Heart Foundation (BHF) claims that if misdiagnosed and untreated, patients have a “70% increased risk of death after 30 days”.

Meanwhile, the AI system being tested can “rule out heart attacks in more than twice as many patients, with 99.6 percent accuracy,” the nonprofit suggests.

“Chest pain is one of the most common reasons why people turn up at A&E,” commented Sir Nilesh Samani, medical director of the BHF. “Every day, doctors around the world face the challenge of separating patients in pain due to a heart attack from patients in pain due to less serious medical conditions.”

The charity helped fund the development of the CoDE-ACS (Collaboration for the Diagnosis and Assessment of Acute Coronary Syndromes) tool to diagnose heart attacks. The system is powered by machine learning algorithms that predict how likely a patient is to have a heart attack.

CoDE-ACS analyzes a patient’s age, gender, medical history, examines electrocardiogram data, and uses a blood test to look for troponin, a protein that appears when heart muscle is damaged, to calculate a perfect score of 100. A higher score means a higher chance of heart attack. The researchers believe the algorithm could figure out whether some people who end up in A&E because of their symptoms actually have heart disease, allowing doctors to more quickly identify those at higher risk.

“For patients who develop acute chest pain as a result of a heart attack, early diagnosis and treatment can save lives,” explains Nicholas Mills, professor of cardiology at the Centre for Cardiovascular Sciences at the University of Edinburgh, who led the study published in Nature.

“Unfortunately, many conditions can cause these common symptoms, and diagnosis is not always straightforward. Leveraging data and AI to support clinical decision making has tremendous potential to improve patient care and increase efficiency in our busy emergency departments.”

The BHF said the technology was being trialled again in Scotland to see if it could improve care in accident and emergency departments. “The CoDE-ACS clinical decision support system, if adopted in practice, can reduce time spent in the emergency room, prevent unnecessary hospitalizations in patients who are less likely to have a myocardial infarction and at low risk of cardiogenic death, and improve the identification and treatment of those who have a myocardial infarction. Myocardial infarction rather than myocardial injury is beneficial for both patients and healthcare providers, “the study concluded.