Post 2: Acquisition and processing of the EAdi signal
The EAdi signal processing is a fascinating area of development where new materials and increased processing power have allowed us to make endless decisions in real time. The blog below describes the main steps of acquiring and processing the EAdi signal (or Edi on the Servo-i). Note that decisions are made every 16ms during NAVA, and monitoring neural respiratory drive can be performed with the same resolution.
The EAdi catheter
To obtain the EAdi signal, we designed an array of 9 electrodes (+1 reference/ground) organized to obtain 8 bipolar recordings. This bipolar recording was preferred to minimize the influence of common mode disturbance on the EAdi signal. The array design was simply to cover respiratory movements of the diaphragm. To simplify the use of these sensors and their introduction into the esophagus we designed electrodes that could be applied on an ordinary nasogastric tube. Due to anatomical changes with growth, we adjusted the interelectrode distance to cover the range of diaphragm movement observed at various body heights.
Signal acquisition and filtering
Acquisition of EAdi (and any EMG signal) follows certain rules. The so called Nyqvist’s theorem (http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem) states that a signal needs to be acquired at a frequency of at least twice the highest frequency content it contains. This is one step in order to prevent so-called “aliasing”. (http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem#Aliasing). Aliasing can be referred to as an effect that causes sinusoidal signals to alias to another sinusoid of the same frequency, but with a different phase and amplitude. The trans-esophageally measured EAdi is acquired at 2KHz, which is at least 4 times higher than its highest frequency component. A second step in the process is filtering of the signal: Low-pass filter allows low frequencies to pass and are used to prevent aliasing, thus it is also called an anti-aliasing filter. High pass filters allow high frequencies to pass and are first applied to stabilize the signals’ DC level and avoid saturation of the amplifiers. High pass filters are also applied to eliminate low frequency signals induced by changes in tissue-electrode interface so-called “electrode motion disturbances”, esophageal peristalsis, lower esophageal sphincter activity, and cardiac activity all of which have a frequency content that is mainly lower than that of the EAdi. However the low frequency of the EAdi spectrum overlaps with the high frequency e.g. cardiac spectrum, such that a tailored filter is required to optimize the signal-to-cross-talk ratio. So called notch-filters are implemented to eliminate disturbances related to specific frequencies and their harmonics e.g. 50 or 60 Hz AC. Later in the signal processing, logical algorithms are implemented on every 16 ms of EAdi signal to detect non-diaphragmatic signals and replace them by predicted values. To dampen stochastic variability of the EAdi signal and to extrapolate the trend of the EAdi without causing runaway, recursive filtering of the EAdi can be applied.
Electrode filtering
One major problem of the EAdi signal processing was to overcome so called electrode-filtering. As the diaphragm moves with respect to the esophagus and the EAdi catheter can move within the esophagus, the positions of electrodes vis-à-vis the diaphragm changes continuously.
Using bipolar electrode amplification, the signal amplitude is affected by the position of the electrode pairs vis-à-vis the diaphragm. For example if the diaphragm (i.e. its electrical activity) is centered between electrodes, the bipolar electrode arrangement will cancel the signal since the same signals are recorded on both electrodes and then subtracted. This will result in an EAdi signal with reduced amplitude and increased frequency content. The strongest EAdi signal using a bipolar electrode arrangement is obtained when one electrode is close to the diaphragm (high EAdi) and the second is away from the diaphragm (low EAdi). This bipolar electrode position will allow the highest amplitude of EAdi (with relatively lower frequency content) and still providing good common mode suppression cancelling crosstalk signals.
Also, it should be noted that the interelectrode distance acts as a high pass filter i.e. the smaller the interelectrode distance the less low frequency waveforms can be measured.
Detection of diaphragm position on electrode array (cross-correlation technique)
Due to the design of the EAdi electrode array, having the same order of polarity for all bipolar electrode recordings, EAdi signals obtained above and below the diaphragm are inverse to each other.
We therefore developed the so-called cross-correlation technique: successively correlating one electrode pair to the second next electrode pair. This method would provide positive correlation i.e. signals are of the same phase if both are obtained above (or both below) the diaphragm, whereas signals obtained from electrode pairs located above and below (i.e. over the diaphragm) will show opposite phase and provide the most negative correlation coefficient.
Double-subtraction technique
To avoid electrode filtering i.e. to reduce the influence of (bipolar) electrode pair position vis-à-vis the diaphragm on the EAdi amplitude, we developed a method called the “double-subtraction technique”. The double-subtraction technique uses the cross-correlation technique to determine the two electrode pairs (one to the second next) that overly the diaphragm. Since these signals are reversed in phase i.e. due to the bipolar electrode array arrangement signal above the diaphragm are inverted to those obtained below the diaphragm, subtraction of these signals creates a summed signal that reduces electrode filtering and improves signal to noise ratio. The RMS of the double-subtracted signal is then added to the RMS of the signal obtained in the center (electrode pair located between the electrode pairs used in the double subtraction). The RMS value obtained with double-subtraction technique is calculated in real time on 16 ms signal segments and constitutes the EAdi signal. In other words this was the story behind the signal waveform representing the neural respiratory drive to the diaphragm used for monitoring and to control NAVA.
Next post will start to address the physiological interpretation of the EAdi signal.