Evaluation of diaphragm electromyogram contamination during progressive inspiratory maneuvers in humans

The diaphragm electromyogram (EMGdi) is susceptible to contamination by non-diaphragm related electrical signals such as the ECG, electrode motion artifacts, and other sources of noise. It is difficult to distinguish between these contaminating signals and those that are representative of the non-contaminated EMGdi, especially during periods when the EMGdi amplitude is relatively small, as during mild contractions of the diaphragm. The aim of the present study was to evaluate how contaminating signals influence the EMGdi power spectrum center frequency (CF) during progressive inspiratory maneuvers. EMGdi and transdiaphragmatic pressure (Pdi) were measured via an esophageal electrode in eight patients with cervical cord injury performing inspiratory capacity (IC) maneuvers. The influence of the contaminating sources on CF was evaluated by two spectral deformation indices, one which is sensitive to both high and low frequency spectral deformation (omega index), and the other which is sensitive to high frequency deformation only (CF1000/CF500 index). The results indicated that EMGdi CF values scattered over a wide frequency range, particularly when the signals were obtained at Pdi levels less than 15% of Pdimax, or at lung volumes less than 30% of IC. When the spectral deformation indices were applied, the scattering in CF values was drastically reduced. This was expressed by a factor of 4 reduction in the coefficient of variation of the CF values. The majority of the excluded EMGdi signals (i.e. not satisfying the spectral deformation index inclusion levels), had low CF values mainly due to the presence of electrode motion artifacts. It was concluded that: 1) The majority of EMGdi power spectrums are deformed early on during unloaded inspirations, and their CF values should be carefully interpreted as being representative of diaphragm function. 2) The relative contribution of contaminating signals in the EMGdi decreases proportionally throughout the first two thirds of an inspiration to IC. 3) The use of visual inspection of the signal in the time domain is questionable as a method to discriminate non-contaminated signals. 4) Analysis of the signal in the frequency domain makes it possible to detect the influence of signal contamination.