'Cutting edge' AI could help diagnose heart attacks

Key takeaways:

• An AI-developed algorithm can help physicians quickly and accurately diagnose myocardial infarction.
• The technology has potential to benefit both patients and physicians and reduce hospitalizations.

An algorithm that was developed with the help of artificial intelligence could help physicians diagnose myocardial infarction more quickly and accurately, according to the results of research published in Nature Medicine.

“For patients with acute chest pain due to a heart attack, early diagnosis and treatment saves lives,” Nicholas Mills, PhD, British Heart Foundation Professor of Cardiology at the University of Edinburgh’s Centre for Cardiovascular Science, said in a press release. “Unfortunately, many conditions cause these common symptoms, and the diagnosis is not always straightforward.”

Guidelines currently recommend diagnosing myocardial infarction with fixed cardiac troponin thresholds, Mills and colleagues wrote, but troponin concentrations can be influenced by comorbidities, sex, age and time from symptom onset.

To improve diagnosis, the researchers developed machine learning models that integrate clinical features with cardiac troponin concentrations at presentation or on serial testing and compute the Collaboration for the Diagnosis and Evaluation of Acute Coronary Syndrome (CoDE-ACS) score, which indicates one’s probability of having MI.

“Harnessing data and artificial intelligence (AI) to support clinical decisions has enormous potential to improve care for patients and efficiency in our busy emergency departments,” Mills said in the release.

The researchers trained the models on data from 10,038patients, and the models’ performance was externally validated using data from 10,286patients.

They found that CoDE-ACS “had excellent discrimination” for MI (area under curve = 0.953; 95% CI, 0.947-0.958) and performed well across subgroups, with an accuracy of 99.6%

The algorithm was additionally able to identify fewer patients at presentation as high probability of having MI than fixed cardiac troponin thresholds (10% vs. 16%) with a greater positive predictive value and more patients as low probability of having MI (61% vs. 27%) with a similar negative predictive value.

The patients who were identified as having a low MI probability had lower rates of cardiac death than those who had intermediate or high probability 30 days (P<0.001) and 1 year (P<0.001) from patient presentation.

“Chest pain is one of the most common reasons that people present to emergency departments,” Sir Nilesh Samani, MD, FRCP, FMedSci, medical director of the British Heart Foundation, said in the release. “Every day, doctors around the world face the challenge of separating patients whose pain is due to a heart attack from those whose pain is due to something less serious.”

Mills and colleagues wrote that if their algorithm, CoDE-ACS, is used as a clinical decision support system, it could potentially have major benefits for both health care providers and patients while reducing hospital admissions. Samani agreed.

“CoDE-ACS, developed using cutting edge data science and AI, has the potential to rule-in or rule-out a heart attack more accurately than current approaches. It could be transformational for emergency departments, shortening the time needed to make a diagnosis, and much better for patients,” Samani said in the release.

References:

• Artificial intelligence could improve heart attack diagnosis to reduce pressure on emergency departments. https://www.eurekalert.org/news-releases/988985. Published May 11, 2023. Accessed May 22, 2023.
• Mills NL, et al. Nature Med. 2023;doi:10.1038/s41591-023-02325-4.