The Future of Humans and Artificial Intelligence in ECG Interpretation

Abstract

 By conducting electrocardiography in medical checkups, various information such as arrhythmia, myocardial abnormalities, and the possibility of myocardial ischemia can be obtained, leading to early detection and treatment of heart diseases, which is highly significant. On the other hand, the electrocardiogram reflects the electrical phenomena of the heart, and the cases in which the name of the disease can be diagnosed based on this information alone are limited. The presence or absence of coronary risk factors and findings on chest x-ray may be important in making a decision. In addition, there are often cases in which diagnosis is difficult without observing changes over time in electrocardiograms without information on clinical symptoms and their progress. Therefore, in ECG judgment, it is desirable to make a judgment after taking into account information from medical interviews, examinations, and examinations other than ECG waveforms in addition to interpretation of ECG waveforms, and final confirmation by a specialist is essential.

 Under these circumstances, in order to maintain the trust of general health checkup participants, the facility is required to ensure that the electrocardiograms performed are of a level that is reliable. For that reason, it is important to confirm the level of the own facility by participating in external quality control as well as internal quality control. However, there are also differences in the circumstances of each facility, such as who performs the electrocardiogram evaluation at what timing for him, and how the evaluation results are reflected in the health checkup results. In this paper, I would like to describe the problems and issues surrounding electrocardiogram accuracy control.

 In addition, the application of computer deep learning to electrocardiograms has been progressed in recent years, and I would like to introduce some of them in this article. For example, a study to discriminate paroxysmal atrial fibrillation from an electrocardiogram during sinus rhythm reported a sensitivity of 82% and a specificity of 83%. Similarly, a model predicting a decrease in left ventricular ejection fraction from an electrocardiogram has also been reported. In this way, research that makes it possible to step into the prevention of disease from what was conventionally judged as a normal electrocardiogram is exactly the ideal of preventive medicine and will lead to further increasing the importance of medical examinations in the future Seem.