Electroencephalographic (EEG) monitoring has become increasingly commonplace in the anesthesiologist’s practice since the Bispectral IndexTM (BISTM) Monitor (Medtronic) was approved by the U.S. Food and Drug Administration (FDA) in 1996 for monitoring anesthetic depth. Today, several other intraoperative EEG devices exist, including the Masimo SedLine®, GE EntropyTM Module, and the MonitorTechnik NarcotrendTM monitor. Each of these monitors provides processed EEG (pEEG) indices, such as the BIS number or the SedLine Patient State Index (PSI), to estimate depth of anesthesia.
However, the algorithms used to derive these numbers are proprietary and vary from device to device, yielding different outputs, even with identical EEG signals (Br J Anaesth 2023; 130: 536-45). Moreover, pEEG indices influenced by physiologic conditions, including age and neuromuscular blockade, do not correlate well with end-tidal volatile concentrations, and have even been shown to be unreliable at detecting episodes of connected consciousness demonstrated via the isolated forearm technique (J Clin Anesth 2021; 73: 110325; Br J Anaesth 2015; 115 Suppl 1: i95-i103; Anesthesiology 2011;115: 1209-18; Br J Anaesth 2018;121:198-209). Accordingly, several questions have been raised about the clinical utility of pEEG indices when used in isolation, as well as with the clinical interpretation of large-scale clinical trials that stratify clinical outcomes on pEEG indices alone (Br J Anaesth 2018;121: 191-3; J Neurosurg Anesthesiol 2018;30: 193-4; Br J Anaesth 2019;123: 464-78).
Thus, while processed EEG indices have been advertised to encapsulate frontal EEG activity into a single number for ease of clinical interpretation, the variability of such indices has raised awareness for the incorporation of raw EEG in clinical interpretation of brain states during anesthesia.
Despite the aforementioned problems with dimensionless pEEG indices, the raw EEG signal is complex and, like an electrocardiogram (ECG), requires training to interpret. Indeed, this complexity was a main driver behind the generation of pEEG indices, easing the burden of raw EEG interpretation. However, intraoperative EEG monitors have recently incorporated the spectrogram as a useful tool to encapsulate the features of the raw EEG in a more parsimonious representation while preserving its characteristics. The spectrogram is simply a visual representation of frequencies in a waveform – it is commonly used in sound applications, including speech and animal calls, but can be applied to any wave, including radio waves and ultrasound waves (Journal of Voice 2022; PLoS ONE 16: e0218006, 2021).
Currently, in the United States, the SedLine, BIS Advance, and BIS Complete Monitors provide a spectrogram for EEG interpretation.
How does one interpret the spectrogram? The “raw EEG” trace is the fluctuation of voltages recorded by a scalp electrode after ground/reference subtraction and minimal filtering; it represents the summation of multiple simultaneous and dynamic oscillations by neuronal populations, resulting in a complex waveform. Like an ECG, raw EEG data are presented in the “time domain,” with time on the x-axis and amplitude (typically in microvolts) on the y-axis. Using a technique known as the Fourier transform, the spectrogram decomposes the raw EEG signal into its component frequencies over time (i.e., transforms the signal to the “frequency domain”), which allows for dynamic estimation of brain states. (The clinical interpretation of such brain states is beyond the scope of this article; however, multiple reviews exist for the interested reader (BJA Educ 2020;20: 166-72; Anesthesiology 2015;123:937-60)). In the spectrogram, time remains on the x-axis; the y-axis, however, represents the component frequencies of a complex wave at a given time window (measured in Hz, equivalent to oscillations/second). A third (z) axis represents the power of each of the component frequencies for the given time window, which is proportional to the square of the amplitude of the raw EEG signal at a given frequency. By convention, spectrograms are plotted in two dimensions, with the z-axis being plotted as color – warmer colors (i.e., orange/red) represent higher power and cool colors (i.e., blue/aqua) represent lower power.
We provide three examples to help illustrate basic features of the Fourier transform and spectrogram interpretation. Figure 1a, top panel, shows a sinusoidal wave with intermittent, high-amplitude bursts at a constant 10 Hz frequency. Here, only the amplitude of the signal changes over time. The corresponding spectrogram (Figure 1a, bottom panel) shows the raw EEG activity at 10 Hz, with warmer colors corresponding to the time windows of the high-amplitude bursts. Note that there is increased power of the transformed signal (yellow epochs) where the amplitude of the raw signal is increased. Conversely, Figure 1b, top panel, shows a sinusoidal wave with constant amplitude but with steadily increasing frequency in the time domain. In this example, the spectrogram shows a fluctuation of the signal frequency on the y-axis, but with the same color throughout, as the power (z-axis) is unchanged over time in the raw signal. Thus, the spectrogram is useful in identifying both changes in amplitude and frequency of a waveform over time.
The spectrogram is particularly advantageous in its ability to visually represent the myriad frequencies and respective amplitudes in the raw EEG signal. Here, the spectrogram reveals the beauty of the Fourier transform, which decomposes a complex wave in the time domain to its multiple constituent frequencies and their respective power in the frequency domain. As an example, Figure 2a shows the two waveforms seen in Figure 1, with their summation producing the waveform in Figure 2b. While the resultant waveform is more visually complex (like a raw EEG trace), the spectrogram in Figure 2c reveals the underlying structure as simply the summation of the two component waves in Figure 1 (compare Figure 2c with spectrograms in Figure 1). It is this visual representation that can give the anesthesia professional a rich assessment of a patient’s brain state and its evolution over time. Moreover, when used in combination with the raw EEG signal and processed metrics, the spectrogram can be tremendously useful in guiding a tailored anesthetic plan.
A useful feature of spectrograms as provided by the current intraoperative EEG monitors is their timescale, which allows the clinician to view EEG activity over the past 20 minutes for the BIS Complete, around one hour for the SedLine, and up to 24 hours for the BIS Advance. A useful application of this feature is detection of burst-suppression (BSP), an alternating pattern of bursts of EEG activity with electrical silence that is known to occur with deeper levels of anesthesia (Acta Anaesthesiol Scand 1993; 37: 121-3). Though BSP is fairly easy to visually detect on the raw EEG signal, raw EEG traces are typically only displayed for seconds on a scrolling screen, limiting the anesthesia professional’s visual detection of BSP to the past few seconds. This time window can be improved by using the suppression ratio, a calculated metric provided by both the BIS and SedLine monitors; however, these calculated parameters only reflect the previous minute or so of EEG activity. By contrast, BSP can be readily detected from the spectrogram over the entirety of its recording. This will be seen on the BIS monitors as a time window of no activity (dark blue) over all frequencies, as there is no frequency content in a period of electrical silence. By contrast, the spectrogram provided by the SedLine will report BSP as a black box over the BSP epochs.
Two inherent limitations of the spectrogram are that 1) it is simply a visual representation of the raw EEG in the frequency domain and 2) it is thus susceptible to anything that manipulates or distorts the raw signal. For example, it is important to set the color (z) axis on the spectrogram to a level that can reveal fluctuations in power spectra over time; axes that are set too low or high will not show color fluctuation and will limit signal interpretation. Further, as with any voltmeter (much like the ECG), the raw EEG signal records all sources of voltage fluctuations and is prone to a significant number of artifacts, often requiring training to interpret. There are several courses and learning modules to this end, including the International Council on Perioperative Neuroscience Training Intraoperative EEG Workshop, the EEG Bootcamp provided by the Safe Brain Initiative (asamonitor.pub/438Fti5), and the International Consortium for Electroencephalography Training of Anesthesia Practitioners (asamonitor.pub/3Iy9Dli). Lastly, as stated above, the spectrogram is but one tool to guide EEG interpretation, and full interpretation is most useful in combination with the raw EEG data, quantitative EEG metrics (such as spectral edge frequency and suppression ratio), and clinical context.
Leave a Reply
You must be logged in to post a comment.