Investigating the value of 24-hour electrocardiogram evaluation by artificial intelligence

Abstract

Background: Electrocardiography (ECG) is a method of recording the heart’s rhythm through connection with electrodes placed on the surface of the body with the help of galvanometer.1 Abnormal heart rate is called arrhythmia.2 Arrhythmias can be detected by 24-hour ECG analysis.3 This method It is an important, painless, non-invasive, and risk-free method for the detection of leakage, which is included in the international clinical guidelines.4–6 Cardiologists use traditional methods or processing software to process 24-hour heart rate data, which is considered a complex task that requires a lot of time and attention. In addition, errors and omissions of parameters are insufÏciently processed and directly depend on the skills and experience of the physician. Recently, artificial intelligence has been used to analyze the 24-hour heart rate, and cardiologists have been working together to confirm the results. Artificial intelligence (AI)-based 24-hour ECG is a method trained on millions of data, with low error, high sensitivity and specificity for abnormality detection.7,8 It has many advantages, such as detecting arrhythmias and saving doctors’ time. Therefore, in order to keep pace with the times, this study was conducted using an artificial intelligence-based platform.9 Aim: to compare the results of the 24-hour ECG using a traditional platform and artificial intelligence. Materials and Methods: 24-hour ECG was performed in patients with arrhythmia (n=40) using a standard BI9900 (Biomedical, Germany, 2023) device. The 24-hour ECG data was processed using the modern Ecglab (Biomedical, Germany, 2023) program. Also, the data was analysed by AI ECG analysis artificial intelligence (Cardiomatic, Poland 2024). Processing was done with SpSS 27.0 (IBM, USA). Results: A total of 40 people participated in the research. Among them, 30.8% were men (n=13), 69.2% were women (n=27), and the average age was 52±33. Since there is no specified gold standard for the diagnosis of arrhythmias, the specificity of the total arrhythmia compared to artificial intelligence was 91.3%, sensitivity 97.9%, and the kappa statistic was k=0.90, so the two methods are compatible. indicating that there is When both methods analyzed the types of arrhythmias detected in each case, the sensitivity for supraventricular arrhythmias was Sens=84.2% or inadequate, and the specificity Spec=100% or adequate. For ventricular arrhythmias, the sensitivity was 100% or adequate, but the specificity was inadequate at 71.4%. For arrhythmias such as atrial fibrillation, tachycardia, and pauses, sensitivity 100% and specificity 100%, respectively, and for bradycardia, sensitivity 100% and specificity 91.3% quality was sufÏcient. Conclusion: The AI-based platform detected more ventricular arrhythmias (p<0.001), while the traditional software overestimated artifacts with atrial arrhythmias (p<0.05). The sensitivity of artificial intelligence is 92.3%, the specificity is 97.9%, and the kappa value is 0.90, which is much better than the traditional program.