According to a new study, the use of artificial ielligence to analyze the electrocardiogram, or ECG, has improved the diagnosis of severe heart attacks. This model was able to ideify cases with unusual symptoms and atypical ECG patterns and at the same time significaly reduce false positive cases.
According to the medical technology section of Tekna news media, a heart attack with ST segme elevation, which is known as STEMI, is considered a very severe heart attack. In this case, a main coronary artery is blocked and preves blood from reaching the heart muscle. The standard of care in this situation is rapid restoration of blood flow with percutaneous coronary ierveion.
Unfortunately, the delay in reaching the recommended time for this ierveion remains a challenge. This problem is especially seen in hospitals and rural ceers that do not have expertise in this field. A delay of more than 90 minutes in restoring the blood flow increases the death rate of paties up to three times.
The study’s lead researcher, Robert Herrmann, from AZORG Hospital in Belgium, explained that ECG ierpretation based on artificial ielligence can bring the best results. According to him, this method detects real heart attacks earlier and at the same time reduces unnecessary activations of the emergency system.
He emphasized that improving the accuracy of triage at the first medical call would simplify emergency care. It also reduces stress and fatigue on clinical teams and ensures that paties who really need urge ierveion receive it without delay and in a timely manner.
In one of the first large-scale evaluations of this artificial ielligence model in the emergency room, researchers retrospectively reviewed the files of 1032 paties with suspected STEMI. These paties were the ones for whom emergency protocols to restore blood flow were activated. Data were collected from three differe specialized ceers between January 2020 and May 2024.
Each patie’s initial ECG was analyzed by the Queen of Hearts artificial ielligence model. This model was trained to detect acute coronary artery occlusion as well as detect STEMI-equivale cases and distinguish them from benign cases. The results of angiography and biomarkers confirmed that among these 601 paties actually had STEMI and 431 cases were falsely diagnosed.
The artificial ielligence ECG model performed much better than the standard triage method. The AI was able to ideify 553 of 601 real heart attacks, while standard triage only correctly ideified 427. The false positive rate for the artificial ielligence model was only 7.9%, while this rate was reported to be 41.8% for the standard triage.
The study’s senior author Timothy D. Henry of Christ Hospital in Cincinnati said the results show the poteial of artificial ielligence. He noted that AI-enhanced diagnosis at the first medical call could shorten the time to treatme and reduce false alarms. This technology is very valuable for optimizing the transfer of paties from non-specialized ceers.
Also, in an editorial accompanying this research, Mohammad Al Khouli, a cardiologist at the Mayo Clinic, praised the researchers for developing an operational model in one of the most complex areas of the heart. However, he cautioned that caution should be used in ierpreting the model because it was originally designed to detect blocked arteries. He emphasized the need for more validation in differe populations and the challenge of iegrating this technology as a compleme to human judgme.
