Electroencephalogram (EEG)
Written by Terence W. Picton, M.D., Ph.D.What is an electroencephalogram?
When the neurons of the human brain process information, they do so by changing the flow of electrical currents across their membranes. These changing currents generate electric and magnetic fields that can be recorded from the surface of the scalp. The electric fields are measured by attaching small electrodes to the scalp.
The potentials between different electrodes are then amplified and recorded as the electroencephalogram; (EEG), which means the writing out of the electrical activity of the brain (that which is inside the head). The human EEG was first recorded in the 1920s by a German psychiatrist named Hans Berger.
The tiny magnetic fields that are generated by the neurons of the brain can be measured with very specialized detectors that became available in the 1970s. These are called superconducting quantum-interference devices; or SQUIDs. These devices are based upon the idea that the resistance of a conductor becomes very low when it is cooled by liquid helium to within a few degrees of absolute zero. These superconducting devices can then pick up the very tiny currents that are induced by the magnetic fields. A set of these detectors (usually mounted in a helmet) brought close to the human scalp, can record the magnetoencephalogram or MEG.
How does an electroencephalogram work?
EEG recordings show the overall activity of the millions of neurons in the brain. The recordings show fluctuations with time that are often rhythmic in the sense that they alternate regularly (Figure: Frequency Analysis). The most prominent rhythm in the EEG is the alpha rhythm which has a frequency between 8 and 13 cycles per second and is recorded mainly over the posterior regions of the scalp close to the places in the brain that process visual information. When the eyes are open the alpha rhythm is very small and when the eyes are closed it becomes large (Figure_Human EEG).
EEG is used extensively to assess neurological disorders. Abnormal decreases of brain activity, usually associated with large slow EEG waves, can occur with brain damage. After very extensive brain damage there may be no electrical activity recorded from the brain. This occurs in the state known as brain dead. An abnormal excitability of the neurons in the brain, which can occur in epilepsy, is usually associated with spike discharges in the EEG.
During sleep, the EEG shows many specific patterns of activity. These can be used to define various sleep stages. During a normal night of sleep, the brain will cycle through these various stages once every 90 minutes or so. In one stage there are many rapid eye movements (REM). This stage of sleep is often associated with vivid visual dreams.
The EEG patterns change when external stimuli (such as sounds or pictures) are presented. These stimuli cause or evoke a particular pattern of brain activity, called the evoked potential.
Evoked Potentials
The evoked potential is the change in brain's electrical activity that follows a stimulus such as a click. The evoked potential is usually too small to recognize in all of the other activity recorded from the brain. When I hear a click, my brain may also be evaluating what my eyes are seeing, thinking about what to have for supper, or controlling my hand as it is writing. Each of these activities will be associated with its own particular electric and magnetic fields. The specific patterns associated with the click can be measured if the click is repeated many times. Each time the click occurs the evoked potential to the click will be the same, but all of the other activity going on in the brain will change. If we average all of the recordings together, the other activity, being different every time the click occurs, tends to cancel itself out, leaving the response to the click (Figure: Averaging).
By choosing different time and voltage scales many different waves can be seen in the auditory evoked potential. These waves track the auditory signals as they are received in the ear and then transmitted up to the cerebral cortex, the specialized outer layer of the brain. This technique of averaging is very powerful. If one averages several thousand responses together one can even recognize the tiny potentials that are generated in the pathways deep within the brain (in a region called the brainstem).
The averaging technique is possible in any situation where in there is some repeating timing-cue. Averaging can therefore be applied to EEG signals that occur before a behavioural response such as pressing a button. This allows us to examine the processes in the brain that lead to the action. The average potentials occurring before, during or after a timing-cue are called event-related potentials (ERPs). According to the formal definition, evoked potentials are one kind of event-related potential. Event-related potentials are used extensively to evaluate what is going on in the human brain when we are perceiving the world, remembering and thinking.
Auditory Brainstem Evoked Potentials
In the first 10 milliseconds after a brief sound such as a click, a series of tiny waves can be recorded. These represent the responses of the auditory nerve and the auditory pathways in the brainstem. These responses are often called the ;auditory brainstem response; (ABR) (Figure: Auditory Brainstem Response). They can be used to make sure that the auditory pathways are working correctly. For example, the normal patterns of response would be disrupted if a patient had a tumor pressing on one of the auditory pathways. The auditory brainstem responses can also be used to see whether sounds have been properly received by the ear. If a patient has a hearing impairment, the responses may be absent or otherwise abnormal. Since the response is generated without the participant having to say ;Yes, I can hear that;, the evoked potentials can be used to measure hearing in participants who cannot respond normally. Most importantly, these responses can be used to assess the hearing of infants. In this way, a hearing impairment that is present at birth can be detected early and treatment initiated so that the child can learn to speak.
Frequency Analysis
EEG signals are usually plotted as changes in voltage over time. However, because of the rhythmic nature of the signals, they can also be plotted according to the frequencies that are present in the recordings. This change from the time domain to the frequency domain is accomplished using a Fourier transform (Figure: Frequency Analysis). The frequency spectrum of the EEG signal then shows various peaks that denote the particular rhythms in the EEG signal.
The evoked potentials can also be considered in the frequency domain just like the spontaneous EEG. If a stimulus is presented at a rapid and regular rate it will make the EEG follow at the same rate. This was first demonstrated many years ago with rapid flashes of a stroboscope, which evoked a driving response. Because the responses continue on with the repeating stimulus, they are also called steady state responses. Similar responses can be recorded in other sensory modalities. In the auditory system a large steady state response can be evoked by stimuli presented at rates near 40 cycles per second. This response varies in amplitude with the level of consciousness, being reduced in sleep and almost absent during anesthesia. At faster stimulus rates (for example, 80-100 stimuli per second) the response does not vary with the state of arousal, and may therefore be used to measure hearing in sleeping participants (Figure: Auditory Steady-State Response).
Event-Related Potentials
The late components of the human event-related potentials can be used to study what is going on in the brain during complex psychological processes such as attending to some stimuli and ignoring others, making decisions about stimuli, understanding speech, and learning new things.
A common stimulus paradigm that is used to study these processes is called the ;oddball; task. Stimuli in this paradigm are presented regularly. Most of the stimuli are one kind (the standard stimuli) but occasionally a different stimulus occurs, and the participant is asked to detect this target stimulus by pressing a button or by keeping a mental count of how many of them have occurred. When one detects the improbable target, a large late positive wave occurs in the ERP (Figure: P300). This is often called the P300 because its peak latency is about 300 milliseconds when the participants are young adults and the target stimulus is easy to distinguish from the standard stimuli. The peak latency of the wave increases as the participants get older or as the task of distinguishing between the target and the standard stimuli gets more difficult.
This wave is usually preceded by a negative wave called the N2 wave. The N2 wave probably represents the conscious recognition that the stimulus is a target, and the P300 probably represents the use of this information to update a model of what is going on in the world.
When the participant does not attend to the stimuli, there are still small changes in the event-related potentials as the brain detects the difference between the occasional target stimulus and the more common standard stimuli. In this context the target stimulus is often called a deviant stimulus since it is just different from the preceding stimuli and not a target to be consciously detected. The deviant stimulus elicits a small negative wave that is superimposed on the usual event-related potential. This negative wave is called the mismatch negativity or MMN. It is most easily recognized in a ;difference waveform; obtained by subtracting the response to the standard stimulus from the response to the deviant stimulus. The MMN reflects the automatic detection of something that is not predicted on the basis of previous stimuli. Although the MMN is small it has the advantage that it can be recorded when the participant is not paying attention to the sounds.
The event-related potentials may also be recorded when a participant learns a list of words that he or she must recall later. If one separately analyzes the ERPs for the words that were later recalled and the words that were not recalled, the waveforms show distinct differences (Figure: Memory). The most prominent changes are a large slow positive wave in the frontal regions, more on the right than on the left and a large slow negative wave in the posterior temporal regions. We consider these waves to represent the processes of ;elaboration;, the making of images and stories to associate with the words. This process facilitates the later recall of the words. Physiologically, this elaboration probably involves processes in the frontal regions (for narrative) and the visual association areas (for images).
Many different regions of the brain are involved in complex psychological processes. These different regions of the brain can all contribute to the event-related potentials that are recorded during the psychological processing. Disentangling the different contributions in the ERPs recorded from the scalp requires a technique called source analysis.
Source Analysis
Given a set of intracerebral currents and the conductive geometry of the head, one can calculate the electrical potentials that are recorded at the scalp. Unfortunately there is no unique solution to the inverse problem. Given the scalp-recordings and the head-geometry, one cannot uniquely determine the current generators. Nevertheless, one can model possible generators of the recorded fields by applying constraints as to what is physiologically and anatomically feasible, one can derive generators that are accurate within reasonable confidence limits. There are two main approaches to source analysis. One attempts to interpret the scalp-recorded electric potentials in terms of a small number of ;discrete; sources. These sources are modeled as ;dipoles;. A dipole is a simple separation of charge into positive and negative poles between which current flows. The second approach to source analysis allows currents to be ;distributed; over wide regions of the brain. The sources obtained using distributed solutions are very similar to the images of cerebral activation obtained by studying cerebral blood flow using PET scans. The major differences are that the electrical (or magnetic) images have a greater temporal precision but less spatial precision.
The process of source analysis can be illustrated for the ERPs evoked by a simple auditory stimulus. First, the ERPs are recorded from multiple scalp locations. This is illustrated in the first figure (Figure: Auditory Evoked Potentials). A distributed source solution can be obtained using a program called VARETA (Figure:VARETA). VARETA is an acronym for Variable Resolution Electromagnetic Tomography. This analysis is presently limited to one point in time (or latency). One such analysis is shown in the figure. This shows that most of the intracerebral activity at a latency 105 ms after the tone is occurring in the right temporal lobe in the region of the auditory cortex which is located on the upper surface of the temporal lobe. This is shown in red on the figure. Multiple analyses at different latencies can be performed and then strung together to make a movie of the brain.
A discrete solution can be obtained using a program called BESA, which stands for Brain Electric Source Analysis (Figure: BESA). This program allows the user to place dipoles in various parts of the brain and to move these around so that the surface fields generated by these dipoles fit the scalp-recorded activity. It uses the information obtained over the whole recording time instead of just analyzing the data at one latency. The analysis of the auditory evoked potential shows multiple intracerebral sources. The activation patterns of these sources shows the sequential transfer of information from one region of the cortex to another.
Figure: Human EEG
On the left is shown a schematic view of the human head viewed from above with the nose at the top. Electrodes are placed on the scalp at many locations and the electrical activity of the brain (the EEG) is recorded from each of these electrodes. When the eyes are open, the EEG signals show low-voltage random activity. When the eyes close (ath the time of the large signal in the electrode near the eyes) an alpha rhythm with a frequency of 11 cycles per second occurs at the back of the head.
Figure: Averaging
In the left column are 16 separate recordings of the evoked potentials to a sound. Each recording lasted for 1000 ms and began 100 ms before the sound came on. Although there are some similarities between the recording the specific response to the sound is obscured by other EEG activity occurring at the same time as the sound is heard. Averaging groups of 4 recordings together gives the blue waveforms shown in the middle column. Averaging these waveforms together gives the average of 16 waveforms shown on the right. The purple wavefroms provide an estimate of the ;noise; in the recording. This term does not mean sounds but describes the background EEG that obscures the evoked potential (just like auditory noise can obscure the sounds we are trying to hear). The vertical scales increase by 2 for every fourfold increase in the number of recordings averaged. The noise level stays approximately constant as the scale increases, since the noise decreases by the square root of the number of recordings averaged. The final average of 16 recordings is larger than the noise level. Most evoked potentials require averaging over many more recordings, but this does not easily fit into a figure.
Auditory Brainstem Response
The upper set of tracings show the auditory brainstem responses recorded from a normal adult in response to a click of moderate intensity. Two separate recordings have been superimposed to show the reliability of the responses. A series of small positive waves are identified using Roman numerals. The most prominent of these are waves I, III and V. Wave I is generated by the auditory nerve and wave V is mainly generated in the ascending fibres of the lateral lemniscus in the brainstem. The lower set of tracings show the response from a normal newborn infant. The response is simpler in its waveform, and the peaks generally have a smaller amplitude and a longer latency than in the adult response.
Frequency Analysis
The upper part of the figure shows an EEG signal recorded in the time domain (blue). The recorded voltage fluctuates as time passes. This signal may be converted used a Fast Fourier Transform into the frequency domain. The spectrum of activity at the different frequencies is shown at the bottom in red. The purple lines in the centre connect the signals that change at particular rates in time to their locations in the frequency spectrum. Thus rapidly changing signals in the centre of the time signal are located on the right of the spectrum.
Auditory Steady State Responses
The upper part of the figure shows the Average response to an amplitude modulated tone with a carrier frequency of 1000 Hz and a modulation frequency of 91 Hz. The main response is at the same rate as the modulation, and shows 4 complete cycles in the 44 ms time period over which the recording was made. The lower part of the figure shows the same response in the frequency domain. This spectrum was actually calculated over a much longer time period than shown in the average response shown above. The spectrum shows a prominent peak at the frequency of the modulation (labeled as the fundamental frequency or fo) and a smaller peak at the second harmonic of this frequency. The lower frequencies in the spectrum represent the activity of the background EEG.
P300
This figure shows the ERPs recorded when a participant was presented with a train of ;standard; tones in which a target tone with a different pitch occasionally occurred. There were two different experimental conditions: in one, the participant ignored the tones and read a book; in the other, the participant attended to the tones in order to detect and count the targets. When the participant ignored the tones (the left tracings) a small mismatch negativity (MMN) was superimposed on the response to the target stimulus. This is best seen in the difference waveform formed by subtracting the ERP to the standard from the ERP to the target. When the participant attended to the tones there is an additional large N2-P300 complex in the response to the target
Memory
Auditory Evoked Potentials
This figure shows the auditory evoked potentials recorded from multiple scalp locations in response to a brief tone presented to the left ear. The waveforms are displayed on a diagrammatic scalp viewed from above with the nose at the top. The waveforms at the upper left are the recordings used to monitor eye-movements. The red arrows point to three different negative peaks. The N100 is maximally recorded at the vertex, the N140 is largest overt the left temporal area, and the N160 is largest in the midfrontal regions of the scalp
Vareta
This shows a VARETA analysis of the auditory evoked potentials at a latency of 105 ms. The top of the figure shows tomographic cuts through the brain at the level where the greatest intracerebral current occurs. The letters represent the left, right anterior and posterior. The levels of the tomographic cuts are shown below on diagrammatic views of the surface of the brain from the top and from the right. The maximum activity is shown by the red color. At this latency the maximum activity occurs on the top of the right temporal lobe. This is the location of the primary auditory cortex
