A novel approach to ECG classification based upon two-layered HMMs in body sensor networks

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Traditional ECG monitoring is limited to hospital settings.
• Continuous, real-time ECG analysis is crucial for early detection of cardiac events.
• Wearable ambulatory devices offer a solution for remote patient monitoring.

Purpose of the Study:

• To develop a novel algorithm for ECG signal filtering and classification in a free-living environment.
• To enable real-time heart attack detection using wearable ambulatory devices.
• To improve the efficiency of ECG preprocessing and feature extraction.

Main Methods:

• Integral-coefficient-band-stop (ICBS) filter for efficient ECG preprocessing.
• Two-layered Hidden Markov Models (HMMs) for ECG feature extraction and classification.
• Segmentation of ECG waveforms into ISO intervals, P subwave, QRS complex, and T subwave using expert-annotation assisted Baum-Welch algorithm.

Main Results:

• The ICBS filter omits time-consuming floating-point computations.
• The two-layered HMMs successfully categorize ECG signals into normal or abnormal types (PVC, APC).
• An ECG body sensor network (BSN) platform was developed to demonstrate real-time signal collection, transmission, display, and classification outcomes.

Conclusions:

• The proposed algorithm is effective for real-time ECG monitoring and classification in free-living conditions.
• The integration of ICBS filters and HMMs provides an efficient approach to abnormal ECG detection.
• The developed BSN platform validates the practical application of the algorithm for remote cardiac monitoring.