AI-powered ECG Interpretation System for Arrhythmia Prediction and Hyperkalemia Detection Using Machine Learning.

BACKGROUND

Artificial intelligence (AI) has emerged as a powerful tool for electrocardiogram (ECG) analysis, potentially surpassing conventional interpretation in detecting arrhythmias and metabolic disturbances such as hyperkalemia. This study evaluated the diagnostic performance and real-world applicability of an AI-powered ECG interpretation system for arrhythmia prediction and hyperkalemia detection.

MATERIALS AND

METHODS

In this prospective observational study, patients presenting to the emergency department or cardiology clinic between March 2024 and February 2025 were enrolled. Standard 12-lead ECGs were analyzed in parallel by a validated deep-learning AI model and board-certified cardiologists blinded to AI results and clinical data. Primary outcomes included sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for arrhythmia and hyperkalemia detection. Secondary outcomes assessed the impact of AI-driven alerts on workflow efficiency and time to therapy.

RESULTS

Among 4,200 patients (mean age 56, 39% female), the AI-ECG system demonstrated 97.2% sensitivity and 96.1% specificity for arrhythmia detection, outperforming manual interpretation for atrial fibrillation, atrial flutter, and ventricular ectopy (p < 0.001). For hyperkalemia, sensitivity and specificity were 83.5% and 87.3%, respectively, reliably flagging all critical potassium elevations within 1 minute. AI alerts reduced median time-to-therapy by 19 minutes. Subgroup analysis showed robust performance in patients with chronic kidney disease and acute cardiac conditions.

CONCLUSION

AI-powered ECG analysis enables highly accurate, rapid detection of arrhythmias and hyperkalemia, supporting earlier clinical intervention and improving workflow efficiency. These findings advocate for broader multicenter validation and integration of AI-ECG tools in routine cardiology practice.