Extracting a clean ECG from a noisy recording: a new method based on segmented-beat modulation

Electrocardiogram (ECG) testing is a common non-invasive diagnostic tool to identify heart diseases. Generally, an ECG recording is corrupted by different types of noise which may jeopardize clinical evaluation. Thus, aim of the present study is to introduce the new segmented-beat modulation method (SBMM) for extracting a clean ECG from a noisy recording. Each cardiac cycle (CC) is segmented into the QRS segment and the TUP segment, respectively independent and proportional to preceding RR interval. TUP-segment modulation is initially used to force all CC to have the same duration, in order to allow a template-beat computation as the median beat. Then, a clean ECG is obtained by concatenating the template-beat after TUP-segment demodulation in order to obtain reconstructed beats whose duration matches that of the corresponding beats in the original noisy recording. Optimization procedures are performed to compensate for small inter-beat morphological ECG variations independent from actual heart rate. SBMM was tested in two applications respectively involving an ECG corrupted by motion artifacts and an abdominal recording from a woman in labor. Results clearly demonstrate the SBMM ability to provide a clean, and thus clinically useful, ECG tracing from a noisy recording.