Machine learning model detects heart attacks faster and more accurately than current methods

A new machine learning model uses electrocardiogram (ECG) readings to diagnose and classify heart attacks more quickly and accurately than current approaches, according to a study led by researchers at the University of Pittsburgh. . natural medicine.

“When a patient comes in with chest pain, the first thing we ask is if they’re having a heart attack. It seems simple, but when it’s not clear from the ECG, It can take 24 hours to complete additional tests,” said lead author Dr. “Our model can help address this major challenge by improving risk assessment and ensuring patients receive appropriate care without delay.”

Within the peaks and troughs of the ECG, clinicians can easily recognize distinct patterns indicative of the worst type of heart attack called STEMI. These severe symptoms are caused by complete blockage of the coronary arteries and require immediate intervention to restore blood flow.

The problem is that nearly two-thirds of heart attacks are caused by severe blockages, yet have no obvious ECG pattern. This new tool helps clinicians detect subtle cues in ECGs that are difficult to spot, improving the classification of patients with chest pain.

The model was developed by co-author Dr. Ervin Saiditch, associate professor of electrical and computer engineering at the University of Toronto, Edward S. Rogers, and chair of the Artificial Intelligence for Health Outcomes Research Committee at North York General Hospital. In Toronto, Pittsburgh, he collected electrocardiograms of 4,026 patients with chest pain at three hospitals. This model was then externally validated on his 3,287 patients from another hospital system.

Researchers compared the model to three gold standards for assessing cardiac events. It is an experienced clinician’s interpretation of her ECG, a commercially available her ECG algorithm, and a HEART score that considers medical history when presented. interpretation of the electrocardiogram, age, risk factors – smoking, diabetes, high cholesterol, etc. -; and blood levels of a protein called troponin.

This model outperformed all three and correctly reclassified 1 in 3 chest pain patients as low, intermediate, or high risk.

We dreamed of matching the accuracy of HEART, but were surprised to find that the machine learning model was based solely on ECG. Beyond this score. “

Salah Al-Zaiti, first author

According to co-author UPMC’s Emergency Medical Services (EMS) Division Director Christian Martin Gill, MPH, M.D., the algorithm is designed to help EMS personnel and emergency department health care providers identify people who are having a heart attack or who have heart problems. It helps identify people with low blood flow. Detects the heart in a more robust manner compared to conventional His ECG analysis.

“This information will help guide EMS medical decisions, such as initiating specific treatments on site or alerting hospitals that high-risk patients are arriving,” said Martin- Gill added. “Conversely, it may also be of interest to help identify low-risk patients who do not need to go to hospitals with specialized cardiac facilities, which may improve pre-hospital triage.”

In the next phase of this research, the team is working with the Pittsburgh Department of Emergency Medical Services to optimize how the model is deployed. Al-Zaiti said he is developing a cloud-based system that integrates with hospital command centers that receive ECG readings from EMS. This model analyzes his ECG and sends back a patient risk assessment to guide medical decisions in real time.

Other authors who contributed to this study are Zeineb Bouzid, Stephanie Helman, MSN, RN, Nathan Riek, Karina Kraevsky-Phillips, MA, RN, Gilles Clermont, MD, Murat Accakaya, Ph.D., and Susan Sereika, Ph.D. , Samir Saba, M.D., and Clifton Calloway, M.D., all in Pitt. Dr. Jessica Zegle Hemsey, University of North Carolina, RN. Ziad Faramand, M.D., Northeast Georgia Health System; and Dr. Mohammad Alrawashdeh, Harvard Medical School. Richard Gregg of Phillips He Healthcare, MS, Peter Van Damme of Utrecht University Medical Center. Stephen Smith, MD, Hennepin Healthcare and University of Minnesota. and Dr. Yochai Birnbaum of Baylor College of Medicine.

This work was supported through the National Heart, Lung, and Blood Institute, the National Center for the Promotion of Translational Sciences, and the National Institute of Nursing Research grants R01HL137761, UL1TR001857, K23NR017896 and KL2TR002490.