Short-term heart rate variability as a general indicator of health estimated by ballistocardiography using a hydraulic bed sensor in elder care

Abstract

A new approach for an accurate heartbeat detection algorithm for ballistocardiographic signals using combined clustering techniques (k-means and fuzzy C-means) for template extraction followed by template matching is presented, which delivered a 99.9 % accuracy of heartbeat detection. Improved heartbeat detection was performed using a representative template based on the segment of the ballistocardiographic signal containing the IJK complex. A clinical study with 61 volunteers was conducted using an existing hydraulic bed sensor system based on ballistocardiography in order to evaluate the performance of the new algorithm in regard to automated generation of the short-term heart rate variability measure LF/HF (low/high frequency). Results showed a high correlation for beat-to-beat intervals, heart rate, and LF/HF for a sampling frequency of 2 kHz and 100 Hz for five-minute recordings compared to electrocardiography as gold standard. For both sample frequencies, Bland-Altman analysis showed close agreement for beat-to-beat intervals, heart rate estimates, and LF/HF between the two measurement methods at both sampling frequencies. Short-term heart rate variability measure LF/HF based on ballistocardiographic signals obtained at a sampling frequency of 100 Hz, which is the operating sampling frequency of the existing hydraulic bed sensor system, provides enough reliable information since it corresponds with a small deviation to electrocardiographic LF/HF. Thus, the existing hydraulic bed sensor system based on ballistocardiography can be used to automatically compute the short-term heart rate variability measure LF/HF, which can be taken as an unobtrusive reliable prognostic factor and general health indicator for in-home monitoring, especially in elder care.