AUTOMATIC ASSESSMENT OF ELECTROMYOGRAM QUALITY

Power spectrum analysis of the diaphragm electromyogram (EMGdi) is tim e consuming, and no criteria have been developed to objectively quanti fy contamination of the signal. The present work describes a set of co mputer algorithms that automatically select EMGdi free of the electroc ardiogram and numerically quantify the common artifacts that affect th e EMGdi. The algorithms were tested 1) on human EMGdi (n = 5) obtained with esophageal electrodes positioned at the level of the gastroesoph ageal junction, 2) on EMGdi obtained in mongrel dogs (n = 5) with intr amuscular electrodes in the costal diaphragm, and 3) on computer-simul ated power spectra. For authentic and simulated power spectra, indexes were obtained by the algorithms and were able to quantify signal dist urbances induced by noise, electrode motion, esophageal peristalsis (i n humans), and non-QRS complex-related electrocardiogram activity. Wit h the index inclusion thresholds set to levels that allowed for a high signal acceptance rate with relatively small artifact-induced fluctua tions (10-15%) of the EMGdi center frequency, the computer algorithms were found to be as reliable as or more reliable than other methods, i ncluding careful visual selection of the time domain signals by experi enced analysts. In conclusion, the frequency domain application of com puter algorithms offers a reliable and reproducible means to objective ly quantify the sources that contaminate the interference pattern EMG.