Averaging algorithm based on data statistics in magnetocardiography

An algorithm for automatic averaging of a magnetocardiogram (MCG) is described. Due to the relatively low signal-to-noise ratio in the MCG, the measured MCG data are often averaged to be analyzed. Generally, R-peaks are used as trigger points, which become anchors for superposition and we can obtain an averaged epoch eventually. However, we have to determine several parameters, such as the threshold magnitude for recognizing R-peak, the time-period of the epoch window, and which channel has dominant R-peaks. In order to determine these parameters automatically, we utilize the magnitude histogram of the root-mean-square waveform of all the channels. We can determine the threshold magnitudes for recognizing R-peaks and T-peaks, respectively, by using the characteristic distribution of the MCG signal histogram. Peak detection procedure using these thresholds records all the locations of the R-peaks and T-peaks, thus we get the average latencies of the R-T intervals and the R-R intervals. From these latencies, we estimate the full width of the epoch window. By adding a routine for processing double R-peaks, our algorithm could conduct the MCG averaging sequence fully automatically. The algorithm has been tested on recordings of 40 normal subjects and 15 patients suffering from myocardial ischemia, and we conclude that this algorithm reliably performs the averaging sequence. The MCG recordings are measured by our 62-channel planar gradiometer system in a magnetically shielded room.