J Wave Detection Algorithm of the BCG in Chair and Bed using Continuous Spline Wavelet Transform

Abstract

This study presents an algorithm for unsupervised beat-to-beat detection of the J-wave of the ballistocardiogram (BCG) in records of both lying (bed) and seated (chair) persons. The algorithm is based on the continuous wavelet transform (CWT) with splines, which offers the advantage of using a wide range of scales and the reduction of noise and mechanical interference. For J-wave detection, the most prominent negative modulus of the CWT is detected using adaptive time windows (the negative modulus provides more information about the location of the J-wave), and then a confirmation is performed from temporal and amplitude parameters. Seven records from a chair database and fifteen records from a bed database were used to evaluate the algorithm. To assess the J-wave detection, the Bland Altman test was used, measuring the heart rate (HR) from the ECG as a reference and considering a 95% confidence interval (±2 SD). For the bed database the mean error was -0.03 beats/min with a confidence interval of ±3.87 and for the chair database the mean error was -0.05 beats/min with a confidence interval of ±3.48 beats/min. Results satisfied the standards for HR meters recommended by the Association for the Advancement of Medical Instrumentation (AAMI).