Computer processing of phrenic neurograms.

Abstract

Manual processing of large numbers of electrophysiological waveforms is a tedious process prone to subjective errors in judgment. To eliminate these problems, we developed computer algorithms and techniques to analyze cat phrenic neurograms produced in response to step changes in end-tidal PCO2. The computer analyzed the neurogram in terms of a model waveform, which consisted of 1) base-line activity made up primarily of noise, 2) an initial sharp increase from base line, 3) a slower ramplike increase in activity, 4) a peak value, and 5) a rapid decrease in activity back to base line. Parameters describing these model elements as well as inspiratory and expiratory times were calculated by the computer. Computer-produced parameters were compared with manually calculated parameters from chart recordings for over 200 individual neurograms. Repeated manual processing of the neurogram had a variability of +/- 10% in the parameter estimates. The computer-produced parameters fell within this range more than 89% of the time. Although the techniques described are directed specifically toward phrenic neurogram analysis, the methods are general enough to be useful in computer processing of other types of physiological waveforms.