The AI-enabled diagnostic already improving cardiovascular care
Cardiovascular disease causes almost 20 million deaths annually(opens in new window), with heart attacks the most time-critical as every minute of delayed treatment risks greater heart muscle loss, prompting interest in innovative faster diagnostics. “Rapidly acquired electrocardiography is critical for early triage. Yet current criteria, reliant on a simple pattern, can miss or delay roughly one third of severe heart attacks. Additionally, around 40 % of emergency cardiology referrals for suspected heart attacks are false positives,” according to Robert Herman, coordinator of the OMI AI ECG Model project. The project used artificial intelligence (AI) to identify serious heart attacks from the initial standard 12-lead electrocardiograph (ECG) test. European Innovation Council (EIC(opens in new window) Accelerator) support enabled validation of the project’s CE-certified medical device, Queen of Hearts – a module now integrated into the PMcardio(opens in new window) platform, also partially developed by the project and deployed across over 150 hospitals, and used by over 100 000 clinicians to analyse millions of ECGs. While Queen of Hearts received FDA De Novo authorisation(opens in new window) in 2026, the project’s host company Powerful Medical, partnering with Roche, also won the 2025 MedTech Innovator Mid-Stage Grand Prize(opens in new window). “The Roche partnership integrates PMcardio, including Queen of Hearts, into Roche’s cobas pulse(opens in new window) point-of-care device, already available in EU emergency departments,” adds Herman.
Integrating automatic analysis and alerts into existing workflows
PMcardio integrates Queen of Hearts AI into hospital and ambulance ECG workflows, including through a mobile app, allowing clinicians to upload a digital ECG or photograph a printout. Within seconds, the AI offers analysis, based on patterns associated with acutely blocked or severely compromised coronary arteries, learned from a dataset of over 18 000 ECGs from over 10 000 patients. The AI was independently validated in a study of over 2 200 patients(opens in new window), comparing its performance with conventional ECG criteria and experienced doctors. According to Herman: “The AI identified around 81 % of patients with an acutely blocked coronary artery, compared with around 33 % using traditional criteria, achieving approximately 91 % overall accuracy.” With EIC support, the technology was prospectively tested in over 700 patients across nine centres in Austria, Belgium and Italy in the AI-ECG TIMI study(opens in new window). ECGs recorded alongside coronary angiography allowed comparison of the heart’s electrical activity, rate and rhythm with coronary blood flow almost simultaneously, accounting for arteries that can block, partially reopen and block again. “The AI was trained and validated against what was actually happening inside the patient’s heart, using heart catheterisation outcomes and cardiac biomarkers, to detect patterns that traditional ECG rules can miss,” explains Herman.
Evaluating the impact of AI-assisted ECG pathways
By the end of 2025, over 40 hospitals across Europe were already using PMcardio in routine chest-pain pathways. “In Slovakia, all three national cardiovascular institutes and six of the country’s eight primary PCI networks are using PMcardio, translating to approximately 75 % of heart attack patients benefiting,” says Herman. The broader PMcardio evidence programme now spans over 25 clinical studies, including DIFOCCULT-3(opens in new window), a 6 000-patient randomised trial across 18 PCI centres in Türkiye, and TITAN-OMI(opens in new window), testing an AI-assisted strategy against standard care in Italy, the Netherlands and Switzerland. “Evidence from such studies can inform European guidelines on AI-assisted triage in acute coronary syndrome,” notes Herman. “For example, in a multicentre US study of over 1 000 suspected heart attacks(opens in new window), AI increased sensitivity from 71 % to 92 % and reduced false-positive calls from 41.8 % to 7.9 %. For patients detected by AI but missed by standard care, the average time saving was almost six hours.”