UC Irvine

Among all cardiac arrhythmia diseases, atrial fibrillation (AF) is the most prevalent and is associated with a chaotic and fast heartbeat, which often increases the risk of cardioembolic stroke and other heart-related problems, including myocardial infarction and progressive heart failure. Thus, it is important to diagnose AF in patients in the early stages and to have them receive proper treatment before the condition worsens. Surface electrocardiogram (ECG), implantable cardiac monitor (ICM), and Holter monitor analyses by doctors are the standard methods to diagnose AF in clinics. However, such analyses/diagnoses are time-consuming and sometimes difficult to interpret due to noise or data contamination. In this thesis, a new AF detection algorithm is proposed and evaluated using four available databases. Before discussing the new algorithms developed in this thesis, a basic introduction of the heart and its arrhythmia are reviewed in Chapter 1. An overview of existing AF detection methods and algorithms used in clinical and academic research is provided in Chapters 2 and 3. Chapter 4 is dedicated to exploring the real-life factors that impact AF detection. The new QRS template-based AF detection method is introduced and discussed in Chapter 5 through 7. It is shown that the new AF detection algorithm improves detection accuracy over standard methods in Chapter 8.