AI-Enabled ECG Screening Detects Previously Undiagnosed Heart Condition in Real-World Care

Artificial intelligence–enabled electrocardiogram (ECG) screening can be effectively deployed in everyday clinical practice to identify patients with previously undiagnosed hypertrophic cardiomyopathy (HCM), according to new prospective data from a multicentre study.

HCM, one of the most common inherited heart conditions, often goes undetected until advanced disease or sudden cardiac events occur. Traditional screening methods can miss early or atypical cases, leading to delayed diagnosis and missed opportunities for prevention.

While earlier retrospective studies suggested that AI-based ECG tools could detect HCM, evidence of real-world clinical implementation had been limited. The latest findings help bridge that gap.

Real-World Deployment Across Health Systems

Researchers implemented an AI-enabled ECG algorithm, Viz ai’s Viz HCM, across five healthcare systems between January and December 2023. The system automatically analysed standard 12-lead ECGs from adult patients without a prior HCM diagnosis and flagged results suggestive of the condition for clinician review.

Out of nearly 146,000 ECGs screened, around 3% triggered an AI alert. Clinicians reviewed close to 70% of flagged cases, indicating strong engagement even within busy clinical workflows. From these, 217 patients met eligibility criteria and were enrolled for follow-up evaluation.

The study cohort included patients from diverse racial and ethnic backgrounds, including Black, Asian, and Hispanic or Latino populations groups often underrepresented in cardiovascular research supporting the broader applicability of the approach.

Faster Follow-Up, New Diagnoses

Most enrolled patients underwent further diagnostic testing based on clinical indications. The median time from the AI-flagged ECG to confirmatory imaging suggestive of HCM was just over seven days, pointing to a significant reduction in diagnostic delay.

In total, 17 patients nearly 8% of those enrolled were newly diagnosed with HCM. Cases were identified in both inpatient and outpatient settings, suggesting that AI-based ECG screening can uncover clinically relevant disease that might otherwise remain undiagnosed.

Reducing Alert Fatigue Without Losing Accuracy

During the study, researchers refined the algorithm’s alert threshold to reduce unnecessary notifications. This optimisation nearly halved the number of ECGs flagged, without lowering the proportion of reviewed alerts that led to patient enrolment.

The finding highlights the importance of balancing sensitivity with alert burden to ensure AI tools remain practical and sustainable in routine care.

Implications for Clinical Practice

The results suggest that AI-enabled ECG screening can be integrated into diverse healthcare environments and support earlier identification of HCM. While the absolute number of new diagnoses was modest, early detection carries significant benefits, including improved risk stratification, timely family screening and prevention of adverse cardiac events.

The authors caution that additional research is needed to evaluate long-term outcomes, scalability and how AI-based ECG screening compares with standard diagnostic pathways. Still, the study marks an important step in translating AI from experimental promise into real-world cardiovascular care.