R-wave detection for pacemakers using a matched filter based on an artificial neural network

A matched filter composed of a time lagged feedforward artificial neural Network (TLFN) and a pulse-shaping filter is used in a noisy environment to detect R-waves for pacemakers. The TLFN reduces the influence of lower frequencies in the invasive electrogram (EGM) signals, e.g. P and T waves, and conditions the EGM to optimize the performance of the dynamically updated matched filter. Detector performance is studied by means of databases containing electrograms as well as different types of noise and interferences, which are added to the signals. The results show that reliable detection can be obtained for relatively high noise levels.