An Approach of Heartbeat Segmentation in Seismocardiogram by Matched-Filtering

Abstract

Heartbeat segmentation in Seism cardiogram (SCG) provides the fundamentals for automated SCG analysis. Traditionally, waveform detection of each cardiac cycle in SCG was depended on the reference QRS complex in Electrocardiogram (ECG). The dominant wave in SCG, namely W complex, could be used as the feature wave for heartbeat segmentation. In this paper, matched-filtering approach was developed and evaluated for W complex detection in SCG without noise suppression stage. Firstly, the template of W complex was selected automatically by the triangle character in SCG and then was used as the coefficients of the finite impulse response (FIR), which greatly enhanced W complexes and attenuated other regions. Subsequently, W complex was detected using the triangle structure, and cardiac cycles and triangle structures were further analyzed for the reduction of false-positive and false-negative detections. The performance of the proposed algorithm was tested on 20 records (each record lasted 50 min) of the combined measurement of ECG, Breathing and Seism cardiograms Database (CEBSDB). The detection rate of 98.74%, the sensitivity of 99.33% and the positive prediction of 99.41 % was achieved on the CEBSDB, respectively. In conclusion, matched-filtering is reliable for the heartbeat segmentation of Seism cardiogram.