Prospects of AI-ECG

In 1924, Willem Einthoven won the Nobel Prize in Physiology or Medicine for demonstrating that the ECG could detect biological signals from the heart. Since then, ECG has been used as an easily available, low-cost, and non-invasive way to detect cardiac function and diagnose heart disease risk. Since the 1950s, the emergence of analog-to-digital converters has made it possible to process and digitize ECG signals [1], and it has also made it possible to automatically extract, analyze, and interpret ECG signals with the help of computer algorithms [2]. Entering the twenty-first century, with the progress of the times, we have come to the era of big data and AI. At this time, multi-source electronic data came into being, and a variety of large-scale electronic medical record data (such as massive ECG data) appeared in the medical field to record the clinical diagnosis process. These massive ECG data have laid a solid foundation for our technological innovation in the field of data-driven heart disease risk diagnosis. The traditional heart disease analysis methods represented by statistical models are somewhat powerless in today’s new situation. With the rapid improvement of hardware computing capabilities represented by graphics processors, AI technology represented by deep learning has ushered in an era of its own. Compared with the ECG-based heart disease risk diagnosis technology used in the past, the newly applied technology not only improves the accuracy of heart disease risk diagnosis but also creates a series of new application scenarios, including intelligent monitoring based on ECG data, bracelets, etc. What is even more gratifying is that doctors can also use the new data obtained to continuously optimize and update the existing heart disease risk diagnosis model so as to obtain a dynamic disease risk diagnosis and prediction model. Through the analysis of past literature, this chapter believes that the AI technology applied to ECG has the following prospects.