IONM Founder Sounds The Alarm on Persistent Quality and Safety Issues
Over the next several weeks, we will be publishing several posts to help with understanding the various types of noise that can contaminate our evoked potential data. We will learn to identify the noise, understand the source of the noise, and learn strategies for mitigating the noise from our evoked potential signal. This week we discuss what noise is, the two major categories of noise, and then begin to focus on several patient generated noise sources and strategies to eliminate them.
When monitoring somatosensory evoked potentials, it is important for the technologist to be able to recognize the various possible recording related noise sources that they may observe in the operating room or the clinical setting. Recording related noises can be divided into two primary categories: patient generated noise sources and externally generated noise sources. Noise issues can be observed in the live data and the live data can be extremely useful in determining the type of noise present in our recordings. We need to be able to recognize the various types of noise, understand the general source of the noise, and develop strategies for dealing with the various noise types that we encounter. One should alway keep in mind that our goal is to minimize noise amplitude and maximize the evoked potential amplitude in and effort to optimize the signal to noise ratio.
Before we discuss how to resolve noise contamination, we must first understand what is meant by the term “noise”. In relation to evoked potentials, noise is any electrical artifact that does not contain our evoked potential of interest. This could include patient generated electrical activity or externally generated electrical activity. Anything that produces a recordable electrical signature that does not contribute to the evoked potential signal would qualify as noise. Noises can be continuous or intermittent and they can be high amplitude or low voltage. Because noise types and sources vary, the techniques for dealing with them vary as well. Continuous noise sources are generally dealt with using appropriate recording filters, common mode rejection, signal averaging, and optimized stimulation rates. Intermittent noise sources are generally eliminated by taking advantage of artifact rejection. We approach each noise type differently which means that it is necessary for us to be able to recognize the type of noise contamination that is present in the recording.
Ideally, we eliminate the noise at the source when possible. In the operating room, every step that we take in our setup should be aligned with the goal of eliminating noise contamination before it ever becomes a problem. For example, braiding recording electrodes, separating stimulation and recording electrodes from each other, utilizing well placed isolation grounds, being cognizant of the location of certain electrical devices in the operating room, and other techniques can help us eliminate noise contamination before it becomes problematic.
Raw SSEP Data
There are multiple patient-generated electrical noises or artifacts that we may be forced to deal with when monitoring our patients in the operating room. These can be divided into physiologically generated noise sources and non-physiologic volume-conducted noises that are conducted through the patient’. Physiologic artifacts can be separated further into continuous and intermittent artifacts. EEG is an example of continuous physiologic artifact while EKG, EMG, and respiration artifact are all examples of intermittent physiologic noise sources. Other artifacts can include retinal potentials, pulse artifact, and even sweat artifact.
During evoked potential monitoring, EEG is treated as a random irregular source of noise that we must work to eliminate from our data. Because of its random nature, we can use signal averaging to eliminate much of the EEG contamination from our evoked potential data. Signal averaging functions under two basic assumptions.
Signal amplitude is constant in latency and amplitude so long as we begin recording our sampled data at the same time that we initiate the electrical stimulation of the peripheral nerve. In most instances we will stimulate the peripheral nerve and record for a sampling interval of 50 or 100 millisecond depending upon the expected latencies of the evoked potential data. If we practice this technique of stimulating and recording at the same time that we initiate the stimulus, the signal amplitude and latency should be constant.
EEG on the other hand is random with reference to amplitude and latency. EEG can be divided into Delta, Theta, Alpha, and Beta frequencies, but relative to the electrical stimulus that is applied, the EEG is in an unknown positive or negative phase at any given time during the sample period. Therefore, EEG is treated as a random noise source who’s average of positive and negative random electrical potentials approaches zero with every subsequent trial included in the average. Consequently, EEG artifact approaches an average of zero while the signal amplitude remains constant in both latency and amplitude permitting us to realize a significant improvement in the signal-to-noise ratio as we include more samples in our average.
The second technique that we use to deal with EEG is in the application of filters. We know that our evoked potential signal does not include the same frequency components as EEG which ranges from 0 to 30 Hertz, with Delta activity ranging from 0 to 4 Hertz, Theta from 4 to 7 Hertz, Alpha 8 to 13 Hertz, and Beta ranging from 13 to 30 Hertz. Because our evoked potential signal does not contain many of EEG’s component frequencies, we can apply a low-cut filter (high-pass filter) that attenuates or eliminates these frequencies from the sample. Applying a low-cut filter with a cutoff frequency of 30 Hz will attenuate many of the lower frequency components that compose EEG but not have a significant impact on our evoked potential response.
Optimized electrode placement is the most important technique that we can apply to aid in the elimination of EEG contamination. While any electrode placed over the patient’s scalp will record EEG activity, placing electrodes in homologous locations on the head will benefit from recording similar EEG frequencies and in similar phases. This technique provides for the possibility of common mode rejection in some recording channels. Secondarily, and of more importance, optimized recording electrode placement will yield larger amplitude recordings of the evoked potential signal of interest, thereby improving the signal side of the signal to noise ratio calculation. Because electrical dipoles obey the inverse square law, with every unit of distance our recording electrodes are from the dipole source, the amplitude of the recorded potentials will decrease by the square or cube of the distance. Therefore, electrodes placed accurately over the somatosensory cortex will yield much larger amplitude evoked potential dipoles than those placed with little care to measurement accuracy.
EMG artifact is random with respect to voltage and frequency, but it is of significantly higher amplitude signal than EEG. For this reason, we cannot rely on signal averaging to eliminate EMG artifact as it will require thousands of averages to address high amplitude artifact. EMG is considered intermittent in nature, therefore we can utilize artifact rejection to eliminate EMG contamination from our evoked potential signal.
We utilize artifact rejection because a little bit of high amplitude EMG goes a long way in contaminating the evoked potential signal average. Therefore, we set the artifact rejection threshold to tell the neuromonitoring system to eliminate the sample from the signal average thereby eliminating the need to acquire many, many, many, many averages. Consequently, we use artifact rejection to completely eliminate muscle and other high amplitude artifacts from the average.
Artifact rejection can be an extremely effective method to eliminate intermittent, high amplitude electrical contamination. And, we must be aware that turning off artifact rejection can be extremely damaging to your signal average. Additionally, one must keep in mind that setting the reject threshold too narrow can make it difficult to obtain evoked potentials as the evoked potential system would reject too many potential trials from your average.
EMG Artifact in the Raw Data