AI Excludes Heart Attack Better Than Specialist Criteria

A machine-learning–based artificial intelligence (AI) program developed by German researchers rapidly and easily rules out a heart attack in patients presenting with chest pain in the emergency department. In a retrospective study, the program named ARTEMIS performed significantly better than traditional diagnostic criteria, the authors reported in a paper published in The Lancet Digital Health.
The AI program has the potential to make a real difference in often overwhelmed emergency departments, according to Dr Andreas Zeiher, associate professor of cardiology at Johann Wolfgang Goethe University in Frankfurt am Main, Germany. “The study results show that with ARTEMIS, twice as many patients can be confidently sent home, as opposed to using the criteria of the European Society of Cardiology (ESC) or the American College of Cardiology (ACC),” he told the Medscape German edition.
Many ‘False Alarms’
Chest pain is one of the main symptoms of a heart attack and the most common reason, in every country, for admission to a hospital’s emergency department. But only 5%-25% of these patients have an acute heart attack that requires immediate treatment. Other diagnoses behind the complaints include spinal complaints or muscular tension, heartburn, a stomach hernia, or a pulmonary embolism.
The ESC and ACC guidelines recommend measuring high-sensitivity cardiac troponin-T (hs-cTn) to rule out an acute coronary event. “Clinical decision pathways do not systematically account for important clinical variables that are known to influence hs-cTn concentrations and pretest probability of myocardial infarction, such as age, sex, symptom onset, electrocardiographic findings, and cardiovascular risk factors,” wrote the study authors, led by Dr Betül Toprak of the Clinic and Polyclinic for Cardiology at the University Heart and Vascular Center of the University Medical Center Hamburg-Eppendorf in Germany.
Relevant Risk Factors
“However, the ARTEMIS program includes a whole slew of clinical risk factors in the evaluation,” said Zeiher. “Nothing additional needs to be measured.”
ARTEMIS has already been validated in several German cohorts, and Toprak and her colleagues applied the program to two retrospective patient cohorts from the United States and Australia. Of 2560 patients who came to the emergency department with suspected heart attack, 166 (6.5%) were ultimately diagnosed with an acute heart attack.
Point of Care
The ARTEMIS program also works with the troponin value. “The great advantage here is that all available troponin measurement methods can be integrated,” said Zeiher. Therefore, a 24-hour standby laboratory is not necessary; a point-of-care rapid test can be used instead. Laboratory-based troponin tests, as recommended by the guidelines, take up to 60 minutes.
“As a decision-making aid in clinical practice, the combination of ARTEMIS and troponin rapid test could enable a safe and fast triage of patients suspected of a heart attack even in preclinical, outpatient, or geographically isolated care areas,” said senior author Dr Stefan Blankenberg, medical director of the University Heart and Vascular Center of the University Medical Center Hamburg-Eppendorf.
ESC and ACC
With the combination of the troponin rapid test and ARTEMIS program, researchers ruled out a heart attack in 35.1% of the patients, with a high negative predictive value of 99.96% and a sensitivity of 99.68%. For Type 1 myocardial infarction, both values were 100%. Only 0.05% of acute heart attacks and 0.07% of heart attacks in the following 30 days were missed.
ARTEMIS achieved exclusion rates well above those of traditional diagnostic criteria. For comparison, Toprak and her team also applied the cutoff values of the two major cardiology societies to the patient cohort. Using the ESC criteria, a heart attack was ruled out for 15.2% of patients, and with the ACC criteria, for 13.8%.
Even among the 806 patients who presented to the emergency department within 3 hours of symptom onset, the ARTEMIS program allowed for the safe exclusion of a heart attack in 153 (19%) of the cases. “In contrast, the ESC and ACC guidelines do not allow for the direct exclusion of a heart attack based on a single troponin measurement in these patients,” the study authors wrote. In the group of “early presenters,” a second troponin measurement after 1 or 2 hours traditionally has been obligatory.
Prospective Validation Required
“Ultimately, the use of ARTEMIS in combination with rapid testing could contribute to relieving hospital emergency departments,” said Blankenberg. The authors pointed out, however, that ARTEMIS has not yet been prospectively tested in clinical practice.
“To exploit its full use case potential in different settings including in-hospital, ambulatory, and preclinical care, randomized controlled clinical trials evaluating the implementation of the ARTEMIS algorithm as a medical decision support device are warranted,” they wrote. Currently, the ARTEMIS program is in the process of certification as a medical product.
This story was translated from the Medscape German edition using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.
 
Send comments and news tips to [email protected].