Fig. 1
Sample 30 min QEEG panel and corresponding raw EEG. (a) QEEG panel consisting of the following QEEG tools: rhythmicity spectrogram (displayed for the left and right hemispheres), CDSA (displayed for the left and right hemispheres), aEEG (displayed for the left and right hemispheres), and asymmetry index (displayed as both absolute and relative values). Vertical black arrows denote electrographic seizures. (b–d) Consecutive ictal EEGs (10 s each) corresponding to the first seizure marked on the QEEG panel demonstrating a left central electrographic seizure
Basic Principles
QEEG was initially developed in the 1960s with the development of compressed spectral array (CSA). There are now several types of QEEG trends available for clinical use as part of commercial QEEG software packages. The primary advantage of QEEG is that it allows for a large amount of data to be displayed on a single screen in contrast to only 10–20 s of data with raw EEG. QEEG also simplifies the information that is displayed, in such a way that it may be amenable for interpretation by non-neurophysiologists. One study found that there was no significant difference in the ability of neurophysiologists, EEG technologists, and neuro ICU nurses to detect seizures on QEEG panels alone [1]. This makes QEEG particularly attractive as a potential bedside patient monitor as a component of ICU multimodality monitoring.
QEEG has several putative advantages over raw EEG review. First, it may reduce the time required for data review. Indeed, raw EEG review is quite labor intensive. One study found that QEEG-guided review of the raw EEG was able to shorten the review process time by 78 % [2]. A survey showed that approximately half of neurophysiologists utilize QEEG as part of their ICU continuous EEG (cEEG) protocol [3]. The usage of QEEG will vary between institutions and readers (various trends used, derivations of trends, frequency of review, and amount of data that is reviewed only by QEEG).
Another potential advantage of QEEG in the ICU setting is that it could allow for real-time data transmission to the treatment team. EEG data obtained by conventional raw EEG review by neurophysiologists is always relayed to the ICU team in a post hoc fashion. Real-time review of the raw EEG is very difficult. In a 2014 survey, the majority of neurophysiologists review each record two or more times a day [3]. Therefore, with conventional EEG-only review, up to 12 h may pass with seizures being undetected. This could lead to delays in treatment of seizures in critically ill patients and potentially adverse outcomes. To date, no clinical studies assessing the role of QEEG on outcomes in the ICU environment are available.
Trends Used for QEEG Seizure Detection
Although many QEEG trends are available for use, this chapter will discuss the trends that have been studied for seizure detection in critically ill patients. These include envelope trend (ET), color density spectral array (CDSA), rhythmicity spectrogram, asymmetry index, amplitude-integrated EEG (aEEG), and automated seizure detectors. This section will describe these trends and provide examples of seizure appearance for each trend. The QEEG samples in this chapter for CDSA, rhythmicity spectrogram, and aEEG will be displayed for the left and right hemispheres, as this is the preference at the author’s institution and was also recently validated in a retrospective trial [1]. It is important to note that other QEEG trends and other derivations may be used. Instead of displaying the left and right hemispheres for each trend, the QEEG trend display may be modified to display individual channels separately or by quadrant. Furthermore, the asymmetry index and aEEG examples in this chapter will be displayed as separate trends, but other institutions may choose to display these as overlapping trends. All QEEG panels displayed in this chapter were created from Persyst (Persyst Development Corporation, Prescott, AZ).
Automated Seizure Detection
Automated seizure detectors are typically part of QEEG software packages and will vary between manufacturers. The algorithms recognize rhythmic patterns based on waveform morphology, distribution, and evolution over time [4]. Once a certain threshold is reached, the software program assigns a pattern as a seizure. The Persyst 12 automated seizure detector has two types of outputs: a binary output of yes/no based on the detection of discrete electrographic seizure events lasting ≥11 s and a seizure probability curve that displays the probability of each 1 s epoch as being categorized as a seizure (Fig. 2). Of note, most automated seizure detection algorithms (ASDA) are trained on a sample of seizures obtained from various EEGs pooled from the epilepsy monitoring unit (EMU), ICU, and ambulatory EEGs. The complex interictal patterns and sometimes subtle nature of seizures in critically ill patients combined with numerous sources of ICU artifact lead to challenges in successful identification by automated seizure detectors (Fig. 2).
Fig. 2
Seizure identification on seizure probability trend and corresponding EEG. (a) Seizure probability trend containing one electrographic seizure (approximate onset marked by the vertical black arrow). The seizure probability trend does identify the seizure, but is not able to discriminate it from numerous non-seizure events. (b) Corresponding rhythmicity spectrogram (displayed for the left and right hemispheres). (c) Ictal EEG corresponding to the time point on the QEEG trends as marked by the vertical blue line. This EEG sample contained abundant artifact (most notably in the T6 electrode), rhythmic delta activity (RDA), and brief rhythmic discharges (BRDs) resulting in poor seizure identification on QEEG
Frequency-Based Trends
Color Density Spectral Array
CDSA is known by several other names: Color spectral array (CSA), fast Fourier transform (FFT) spectrogram, and density spectral array (DSA). CDSA displays a three-dimensional, frequency-based graphical display of the EEG data over time. Time is shown on the x-axis, and the EEG frequency is shown on the y-axis. The various colors represent the power of various frequency bands. The power is the area under the Fourier spectrum curve within a given frequency range (i.e., delta power). In other words, the power is the amplitude (or voltage) of the EEG within a specific frequency range. The power is represented by color. The colors used in the graphical display of the power in the CDSA trend will vary between QEEG software programs. Each program will display a color scale with the CDSA trend. The CDSA trends shown in this chapter were created from Persyst with cooler colors (blue and green) indicating lower power and warmer colors (red, yellow, pink) indicating higher power.
Seizures often consist of an increase in frequency and amplitude and therefore will appear on CDSA trend as a paroxysmal event with increased power. Warmer colors will take the place where cooler colors previously were seen. Additionally, the characteristic seizure evolution in terms of amplitude and frequency can be appreciated on CDSA as an upward arch shape (Fig. 3). Some seizures in critically ill patients consist of little or no increase in amplitude and/or frequency and therefore might be missed on CDSA.
Fig. 3
Seizure appearance on the CDSA trend, 0–20 Hz (displayed for the left and right hemispheres) for two different patients. Vertical black arrows denote the approximate onset of electrographic seizures. The upward arch shape of seizures can be appreciated on both patients. (a) Recurrent right hemispheric seizures seen as an increase in power (represented by warmer colors). Note the evolution of power increase (shown by the red and yellow colors). Soon after the onset of the seizure, there is a gradual decrease in frequency, then increase, and then decrease again before cessation. This is superimposed on a diffuse mild increase in power (shown by green and teal colors during seizure activity). (b) A single right hemispheric seizure on the CDSA trend. Aside from a brief increase in high power (denoted by diagonal black arrow) in mid-frequency range, the majority of the seizure consists of highest power (red, pink, and white) in the delta frequency range. This is superimposed on a diffuse mild increase in power (shown by green color)
Rhythmicity Spectrogram
The rhythmicity spectrogram, rhythmic run detection and display, is a proprietary tool developed by Persyst, Inc. An example of a rhythmicity spectrogram is shown in Fig. 4. Like CDSA, the rhythmicity spectrogram is a three-dimensional display. Time is on the x-axis and frequency is on the y-axis (but on a logarithmic scale to accentuate lower frequencies). Although the power is displayed by color-coding (darker blue color indicating more power), it differs from CDSA by only displaying the power in components that have a high degree of rhythmicity, instead of displaying all the power. Seizures will present as areas that are darker in color. The rhythmicity spectrogram is particularly helpful in displaying the evolution of seizures (Fig. 4).
Fig. 4
Seizure appearance on the rhythmicity spectrogram, 0–25 Hz (displayed for the left and right hemispheres) for two different patients. Vertical black arrows denote the approximate onset of electrographic seizures. The evolution of the seizure can be appreciated on both patients. (a) Recurrent right hemispheric seizures beginning with an increased power (darker blue coloration) in alpha activity. As the seizure progresses (shown by the red arrow), there is gradual evolution of increased power into lower frequency ranges before cessation. (b) Three generalized seizures (with left hemisphere predominance) beginning with a subtle, increased power in the delta frequency range that gradually increases in power (light blue becoming darker blue). As the seizure progresses, an increase in power is seen in the alpha and beta frequency ranges as well followed by abrupt cessation
Subtle seizures can often be seen only on the rhythmicity spectrogram while not appearing on other trends. However, the rhythmicity spectrogram is prone to highlighting interictal periods and artifact that are easily mistaken for seizures. Examples of these will be discussed later in the chapter.
Asymmetry Index
The asymmetry index compares the difference in power between homologous electrodes (i.e., the difference in power between F3 vs. F4 and O1 vs. O2, etc.). The difference is represented in a graphical display. Typically, there are two graphs that are separate or overlapping: the absolute asymmetry index and the relative asymmetry index (Fig. 5). The absolute asymmetry index (yellow trace) calculates the absolute difference, always displaying a positive score. There is an upward deflection with increasing asymmetry and a downward deflection with decreasing asymmetry. The relative asymmetry index (green trace) is able to show lateralization for the asymmetry. An upward deflection represents more power in the right hemisphere, and a downward deflection represents more power in the left hemisphere. This trend is particularly helpful for focal or lateralized seizures. However, a bilateral or generalized seizure with similar power in both hemispheres will likely not show up well on the asymmetry index.
Fig. 5
Example of three left hemispheric seizures on asymmetry index and asymmetry spectrogram (approximate onset marked by vertical black arrows). There is a subtle, upward deflection of the absolute asymmetry index (yellow trace) indicating a period of increased asymmetry. There is a corresponding downward deflection of the relative asymmetry index (green trace) indicating increased power in the left hemisphere. Interictally, there is equal power in the left and right hemispheres, as seen by equal red and blue coloration on the asymmetry spectrogram. The seizures appear on the asymmetry spectrogram as a period of dark blue indicating higher power in the left hemisphere. There is increased power in the right hemisphere after each seizure due to postictal left hemispheric suppression
The asymmetry spectrogram (Fig. 5) also displays similar information regarding the power in homologous electrodes. Colors indicate where more power is present (red = more power in the right hemisphere and blue = more power in the left hemisphere). The degree of asymmetry is represented by the darkness of the color. In addition to seizure detection, the asymmetry index and asymmetry spectrogram are also particularly helpful for ischemia detection.
Amplitude-Based Trends
Envelope Trend
The envelope trend (ET) is a QEEG trend that is based only on amplitude. The raw EEG is divided into 10–20 s epochs. For each epoch, the median amplitude is calculated and plotted over time, creating the ET display. This trend is often displayed separately for the left and right hemispheres, but can be customized to separately display the ET for a specific set of electrodes. By plotting only the median amplitude, the ET has the advantage of being able to filter out short-duration artifacts. Conversely, it may miss very brief seizures due to the fact that the ET is calculated in 10–20 s epochs. Seizures on ET are visualized as an upward deflection in the trace (Fig. 6).
Fig. 6
Example of three generalized seizures on envelope trend. The blue trace corresponds to the left hemisphere and the red trace corresponds to the right hemisphere. Vertical black arrows mark seizures. For each seizure, there is a clear, upward deflection in both the red and blue traces. Seizure duration is approximately 5 min
Amplitude-Integrated EEG
The amplitude-integrated EEG (aEEG) trend is another trend calculated only by amplitude. For each data point, the raw EEG is filtered and rectified (all values made positive). The amplitude-integrated EEG (aEEG) trend is displays the minimum and maximum amplitude of the raw EEG signal in a predefined time frame (typically 1–2 s) on a semilogarithmic scale. Seizures appear as an increase in the minimum amplitude, creating an upward arch shape (Fig. 7). There is often a corresponding increase in the maximum amplitude. This trend is also known as a cerebral function monitor (CFM) and has been utilized extensively for seizure detection in neonates. The original CFM display represented EEG data from one raw EEG channel placed over the parietal regions (P3 and P4). To have the ability to detect lateralized abnormalities, it is now common for CFM machines to display two channels of data (C3-P3 and C4-P4). Commercial QEEG software, such as Persyst, has the ability to display aEEG trends by any group of electrodes and is often displayed separately for the left and right hemispheres, incorporating all lateralized electrodes from the standard 10–20 montage.
Fig. 7
Example of seizures on the aEEG trend (displayed for the left and right hemispheres) for two different patients. Approximate seizure onset is marked by vertical black arrows. (a) Bilateral seizures are represented by a large, upward deflection in the minimum and maximum amplitudes of the baseline of both traces. The gradual increase in amplitude (evolution) can be appreciated well. (b) Right hemispheric seizures are represented by an increase in the minimum amplitude of the red trace, without a notable change in the maximum amplitude. This subtle seizure appearance on aEEG is more common in critically ill patients than the seizures shown in panel (a)
Data for Quantitative EEG Utilization in Seizure Detection
Sensitivity of Quantitative EEG Used in Isolation for Seizure Detection
The majority of studies on QEEG for seizure detection have been in the pediatric and neonatal population, although there are an increasing number of studies evaluating QEEG in critically ill adults. Beyond just patient’s age, there is significant heterogeneity in these studies. Some utilize QEEG trends obtained from full-montage cEEGs, while others are obtained from limited channel cEEGs. Furthermore, even when a full cEEG montage is used, the QEEG trend studied may be derived from all channels or from a limited number of channels. Although certain QEEG trends are studied more often than others, the type of QEEG trend studied (commercially available vs. a novel QEEG algorithm) often differs between studies. Some studies may employ only one trend while others use more than one. Another potentially confounding variable in QEEG studies is the variability in expertise of QEEG readers and the extent of QEEG training provided. Studies may use neurophysiologists as readers, but they may not be considered “experienced readers” as many neurophysiologists have not had training/experience with QEEG trends. Conversely, many of the studies in the neonatal population utilize neonatologists as readers since they are more likely to be the ones interpreting the bedside CFM. Neonatologists may not have experience in reading raw EEGs, but they might be considered “experienced readers” since some have had several years experience in interpreting CFMs. Furthermore, the manner in which sensitivity and specificity are calculated (scoring based on capturing individual seizures or scoring based on the presence/absence of seizures in patients or epochs) vary between studies. However, in actual clinical practice, knowing the exact number of seizures present may not be necessary, and simply knowing if seizures are present or not may be sufficient to guide therapy.
Related posts:
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
