Resolution of ballistographic records into cardiac and respiratory components.

Abstract

The feasibility of average transient computing (ATC) for elimination of respiratory interference and random variations of the ballistocardiogram (BCG) was investigated. It appears that at least one half of the low-frequency interference in the BCG is not random with respect to the cardiac cycle.If noise is periodic there is an ostensible rapid improvement in the ratio of signal to noise when relatively few transients have been summed, and little or no improvement with increasing number of transients.The behavior of non-random noise in ATC was further studied with the aid of simulation experiments. It was observed that the computed average can be considerably biased if noise is periodic. The error was smallest when the number of transients averaged was a multiple integer of the least common factor of the fundamental frequencies of the signal and periodic noise.Several other potential sources of errors in ATC were discussed. With proper consideration for these limitations, the ATC methods can be used for separation of cardiac and respiratory components of the ballistogram with adequate accuracy for some applications. At the same time, the importance of developing other and more sophisticated ways to eliminate respiratory interference is emphasized. A mechanical device for compensating the effect of respiration was devised.