Novel algorithm for beat-to-beat heart rate measurement from the BCG in seated, standing and supine positions: Towards an universal algorithm
The ballistocardiogram (BCG) is a cardiac mechanical signal whose waveform morphology varies between subjects, depending on the mechanical interface and body position. Because of this, most algorithms are tailored to specific types of BCG signals, as there are not many publicly available BCG datasets. To overcome this drawback, this work describes an algorithm for beat-to-beat heart rate detection from ballistocardiogram (BCG) in sitting, standing, and supine positions with different sensors both in the longitudinal and transverse directions. The algorithm is based on the continuous wavelet transform (CWT) with splines and its envelope to estimate the J-wave peak (Jp) position in the BCG. The BCG and CWT and its envelope are then analyzed for BCG morphology type and a backward or forward search is performed to reduce false positives to find Jp. The algorithm was evaluated in recordings of three BCG databases: longitudinal BCG of chair (7 seated recordings), transverse BCG of bed (17 supine recordings) and longitudinal BCG of weighing scale (14 standing recordings), obtaining the following values of sensitivity and positive predictivity in the BCG databases: chair (99.87% and 98.48%), bed (99.64% and 97.60%) and weighing scale (99% and 88%) over 9266 beats. The 95% confidence intervals of the Bland-Altman plots in beats per minute (bpm) obtained were: chair ± 3.4 bpm, bed ± 4.18 bpm and weighing scale ± 4.01 bpm, show values that met the accuracy limits recommended by the AAMI standard for heart rate meters.
Autores:
Revista: Biomedical Signal Processing and Control
https://doi.org/10.1016/j.bspc.2024.106641