Heart Anomaly Detection System For Ambulatory Electrocardiogram

Abstract

Cardiovascular diseases are the leading cause of death globally, causing nearly 17.9 million deaths per year. Early detection and treatment are critical for improving this situation. Today's wearable medical devices are becoming popular because of their price and ease of use. Many manufacturers have developed products to continuously monitor patients' heart conditions as they perform their daily activities. However, one major challenge of collecting and analyzing heart data using mobile ECG is baseline wander and motion artifacts created by the patient's motion, resulting in false diagnoses. In addition, very few monitor devices can diagnose complex heart anomalies beyond detecting rhythm fluctuation. The research reported in this thesis proposes an anomaly detection system that could compensate for the motion noise and give a reliable diagnosis based on a single lead ECG signal. The noise removal algorithm could automatically remove the baseline wander and suppresses most motion artifacts in mobile ECG recordings. This algorithm shows a significant improvement over conventional noise removal methods. Two signal quality metrics are used to compare a reference ECG with its noisy version: correlation coefficients and mean squared error. For both metrics, the experimental results demonstrate that the signal filtered by our algorithm can improve the signal-to-noise ratio by ten. In addition, this thesis describes a new algorithm that combines a Short-Time Fourier Transform (STFT) spectrogram of the ECG signal with handcrafted features that outperform commercial product capabilities. The new algorithm can detect 16 different rhythm anomalies with an accuracy of 99.79% with a 0.15% false alarm rate and a 99.74% sensitivity. In addition, the same algorithm can also detect 13 heartbeat anomalies achieving 99.18% accuracy, 0.45% false alarm rate, and 98.80% sensitivity.