Interpreting Electroencephalogram (EEG) – An Introductory Review of Assessment and Measurement Procedures

This work aims to compile various measurement and calibration techniques for Electroencephalogram (EEG). The objective is to provide an assessment of different components of an EEG and their significance in understanding different types of responses to stimuli. This review provides an overview of recommended electrode positions and lead systems for EEG with calibration and activation procedures. The different components of an EEG according to frequency bands – Gamma, Beta, Alpha, Theta and Delta have been discussed with emphasis on their interpretation and their role in several applications. The motivation behind this study is to provide a basic understanding of the composition of EEG, understanding the significance of each brain wave and the locations at which they are most prominent.


Introduction
Hans Berger, a neuropsychiatrist from the University of Jena in Germany introduced the first Electroencephalogram (EEG) machine in year 1929. According to him, a change in the brain functions such as sleep, epilepsy caused a change in the brain currents. This led to the creation of a new branch of medical science called neurophysiology. A small number of disk electrodes were glued on the scalp and the signal received from the electrodes were amplified using operational amplifiers to observe the change in brain's activity. EEG monitoring is used to record the electrical activity or voltage fluctuations of the brain generated due to the ionic flow of current within the nerve cells. EEG finds its application in the diagnosis of epilepsy, brain tumours, stroke, sleep disorders, etc.
EEG is usually preferred for brain function assessment, along with several other neuroimaging techniques such as Magnetic Resonance Imaging (MRI) and Positron Emission Topography (PET), as it is the safest non-invasive technique and can be easily used for long-term bedside monitoring in intensive care units (Gasser and Molinari, 1996). Moreover, EEG has a higher time resolution (milliseconds) as compared to the PET or MRI (seconds to minutes). EEG analysis can be a time-series or a spectral assessment. Stimulated EEG responsesalso called as Evoked Potentials (EPs) are the time-series events which can be used for analysis of cognitive skills and psychophysiology and are based on the neuron excitation due to a visual or auditory response. EPs have a much smaller amplitude than EEG and cannot be differentiated from raw EEG, therefore, they are extracted by averaging the epochs of EEG, and contain significant patterns of the evoked neural activity (Teplan, 2002). Spontaneous EEG responses are analysed in frequency domain in terms of spectra and coherences. EEG is often affected by artefacts due to facial muscle activity (Electromyogram-EMG) or eye movement (Electrooculogram-EOG).
This paper describes an overview of the methods adopted for EEG assessment from various parts of the brain and the significance of brain waves of various frequencies. This review will first discuss the sensory positions, lead systems and calibration methods used for EEG assessment in clinical practice and then shift its focus on the brain oscillations, their significance and the locations they can be extracted from.

Background
EEG arises from a cortical neural cell composed of 70-80% water and 20-30% protein and lipid, and can be classified on the basis of its structuring and functioning as (i) soma-cell body, (ii) dendrites-numerous short impulses of soma, (iii) axon-signal transporter or carrier from soma to another nerve or muscle cell (Malmivuo and Plonsey, 1995). The gap between the axon of one neural cell and the dendrites of another is called a synapse, and neurotransmitters also called as chemical messengers are responsible for the signal transportation across neurons (Maas, 1979).
A diagram depicting a cortical neuron which receives impulses from several thousand neurons is shown in figure  1.
The bioelectricity generated due to the neural excitation can be measured on the scalp as EEG using wet or dry electrodes. Figure 1. A cortical neuron with soma, dendrites and axon EEG is measured as an electrical activity generated by similarly oriented groups of cortical neurons near the scalp where the electrodes are placed (Louis et al., 2016a). Action potentials are generated when neurons communicate with each other causing the membrane potential of soma to rise and fall rapidly. An action potential is only fired when a certain threshold voltage is reached and depends on the average of excitatory and inhibitory potentials. Action potentials are too short impulses to be measured and hence EEG is measured as a summation of excitatory and inhibitory postsynaptic potentials (Louis et al., 2016b).

Electrode Positions and Lead Systems
The 10-20 system is an internationally recommended method by the International Federation of Clinical Neurophysiology to describe the locations of twenty-one electrodes on the scalp. It is generally used to measure EEG from four main parts of the brain-frontal, central (top), temporal (sides) and occipital (back). The electrodes are equally interspaced according to skull size and shapes and are placed at 10% and 20% intervals from each other (Jasper, 1958;Tyner and Knott, 1983). Depending on type of application, the EEG can be extracted from 8 to 32 channels. EEG measurements can be bipolar or unipolar. The former refers to potential difference between a set of electrodes and the latter is the potential of each individual electrode with respect to reference electrode.
The Queen Square System (QSS) is preferred over the 10-20 system to measure Evoked Potentials (EPs) as the lateral occipital leads are placed farther from the centre, thereby allowing an improved recording of the scalp distribution of the patterned visual EPs (American Clinical Neurophysiology Society, 2006a). If electrodes are closer to each other, then the lead field current flows more and more within the skin than between the electrodes, thus decreasing the electrode sensitivity (Malmivuo and Plonsey, 1995). Both 10-20 system and QSS are depicted in figure 2. Reference points nasion on top of the nose and inion at the base of the skull are used as landmarks to place electrodes. At least four channels should be recorded for the EPs employing QSS. In both 10-20 and QSS, even numbered and odd numbered electrodes refer to right and left side of the head respectively. G (in QSS) and Cz & Fz (in 10-20 system) are the ground electrodes, whereas A1 and A2 are used for contralateral referencing. A minimum of six EEG derivations are required using the 10-20 system to obtain activity from the frontal, central and the occipital regions. The recommended measurement channels for 10-20 system and QSS are F3-A2, C3-A2, O1-A2, F4-A1, C4-A2, O2-A1 and LO-MF, MO-MF, RO-MF, MF-A1 respectively ((American Clinical Neurophysiology Society, 2006a;Kushida et al., 2005)).

EEG Measurement, Calibration and Activation Procedures
According to 10-20 standard system, all twenty-one electrodes should be used for diagnostic EEG analysis, however, smaller number of electrodes are adequate for some specific analysis such as sleep study. The interelectrode impedances should not be more than 5000 Ohms (American Clinical Neurophysiology Society, 2006b). EEG are measured as peak to peak voltages and range from 0.5-100µV.
According to the guidelines proposed by the American Clinical Neurophysiology Society, the minimum sampling frequency to acquire EEG signal should be three times the higher frequency filter setting to avoid aliasing (American Clinical Neurophysiology Society, 2006c). The low frequency filter setting higher than 1Hz and the high frequency filter setting lower than 70Hz is not recommended to obtain all brain waves. Due to low amplitudes of EEG waves, a 12-bit analog-to-digital conversion is recommended. A notch filter to reject the electromagnetic interference can be implemented. A grounding electrode should always be used except when other instruments are also attached to the patient, so as to avoid double grounding (American Clinical Neurophysiology Society, 2006b).
The effect of stimuli on the EEG should be recorded when the eyes are open and closed as some information is masked by the alpha activity and is only visible on the attenuation of alpha waves by opening eyes. Similarly, hyperventilation should be performed for a minimum of 3 minutes for adequate activation of EEG. The quality of hyperventilation, level of drowsiness and awareness in patients is also recorded by an electroencephalographer. Hyperventilation is performed to evoke elliptical abnormalities or seizures and to increase EEG sensitivity. The EEG should normally return back to its baseline level one minute after the conclusion of hyperventilation. If the mas.ccsenet.org Modern Applied Science Vol. 14, No. 6; 2020 return of EEG occurs after a protracted period, it may indicate an abnormality (Marcuse et al., 2016). For calibration purposes, it is the job of the technologist to assure that the patient is relaxed and maximally alert during certain epochs of assessment.
Intermittent Photo Stimulation (IPS) is also performed in EEG to detect photosensitivity (Kasteleijn-Nolst Trenité et al., 2012). IPS is performed by strobe light flashing between 1 to 35 Hz which evokes rhythmic waves in the occipital region of the brain. Lack of rhythmic waves do not indicate any abnormality, however, absence of waves on one side of the brain may indicate a unilateral structural disease involving the occipital region. This test is useful in patients with seizures where flickering light gives a photoparoxysmal response.
Sleep and sleep-deprived EEG are other standard tests that are used to detect unusual brain activity. Sleep EEG tests are generally conducted in children under five years of age to diagnose periods of epilepsy during sleep ((Arbasino et al., 2015;Kaminska et al., 2015;Kurahashi and Hirose, 1993;Weber et al., 2017;Werhahn, 2009)). On the contrary, sleep deprived tests are mostly conducted in adults where an increase in epileptiform activity is observed in about 30% patients with epilepsy ( (Marcuse et al., 2016;Renzel et al., 2016;Wirrell, 2010)). Epileptic events are also recorded during ambulatory monitoring, hospital and home video-telemetry ( (Lawley et al., 2015;Seneviratne et al., 2013;Shafer et al., 2012)).
Other standard EEG tests involve the use of Neurofeedback systems which acquire the brain waves simultaneously and aid in self training to alter and improve the brain activity. These tests have shown huge improvements in children suffering from Attention Deficit Hyperactivity Disorder (ADHD) and in patients with eating disorders, anxiety and depression ((Alhambra et al., 1995;Arns et al., 2013;Boutros et al., 2005;Gordon et al., 2010;Laporte et al., 2002;Patrick, 1996;Snyder et al., 2008;Swartwood et al., 2003)).

Brain Waves and What Do They Tell Us?
As mentioned before, EEG or captured brain waves are an average of postsynaptic potentials and are generally sinusoidal in shape. EEG contains of several frequencies which are extracted by Fourier transform of raw EEG.
A person can be trained to generate more or less of these specific frequencies, which enables him/her to control his/her own brain function. A neurofeedback system is used for these purposes and is a popular technique applied in sports. It also provides solutions for attention training, stress, sleep, emotional difficulties and intimacy ((Albright, 2010;Fisher, 2014;Larsen, 2012;Robbins, 2008)).
Brainwaves can be divided into five general categories ranging from the highest frequency to lowest frequency content -Gamma, Beta, Alpha, Theta and Delta

Gamma
Gamma (γ) waves are the fastest (or highest frequency>32 Hz) waves and are known to pass information rapidly. They were referred to as brain noise by some researchers until it was found to be dominant in states of higher virtues like meditation or altruism (Albright, 2010;Fisher, 2014;Larsen, 2012;Robbins, 2008). However, some studies have suggested that γ waves are more likely a product of mismeasurement or constitutes of Electromyography (EMG) waves (Whitham et al., 2008(Whitham et al., , 2007Yuval-Greenberg et al., 2008).
Gamma waves are known to originate in thalamus which lies at the top of the brain cell near the cerebral cortex (Herrero et al., 2002). The main purpose of thalamus is to relay motor and sensory signals. Low amounts of γ waves are known to be linked to low memory and learning disabilities (Başar-Eroglu et al., 1996) and high amounts are linked with anxiety, high arousal and stress (Malik and Amin, 2017). Irregular γ activity has been linked with Alzheimer's disease an epilepsy (Uhlhaas and Singer, 2006). Study conducted by Le Van Quyen M in 2016(Quyen et al., 2016 suggests that γ waves are highly correlated with the synchronous firing of inhibitory neurons. Several studies have reported the emergence of stimuli-specific γ waves between 30 to 80 Hz in the visual cortex located in the occipital lobe of anaesthetized cats and awake monkeys ( (Brosch et al., 1997;Eckhorn et al., 1988;Engel et al., 1991;Freiwald et al., 1995;Frien et al., 1994;Gray et al., 1989;Kreiter and Singer, 1996)). Spike oscillations in the above studies were more frequent when the stimuli were visually coherent than random.
γ waves have been classified into three classes in a study by as 40-Hz transient evoked response, 40-Hz steady state response and induced gamma (30-80 Hz) response (Tallon-Baudry and Bertrand, 1999). Evoked γ responses occur on the onset of an external stimuli-audio, visual or physical and has the same latency and phase in each trial. Thus, it can be analysed as an average of several trials. The evoked response is known to disappear during sleep and Rapid Eye Movement sleep (REM) (Llinás and Ribary, 1993). Steady state responses at 40-Hz are interpreted in terms of natural resonant frequencies of brain (Galambos et al., 1981)   Lo beta is associated with focuses, introvert concentration, Beta 2 with high energy, anxiety and performance, and hi beta with stress, anxiety and high arousal (Abhang et al., 2016).
β rhythms are also classified as Rolandic and frontal based on the locations they are extracted from (Kropotov, 2009). The former is found in central (C3, Cz and C4) part of the brain and the latter in frontal (F3, F4).
Rolandic or Mu wave (8-13Hz) is considered as a subharmonic of β activity (Hobson and Bishop, 2017) and was first described by Gestalt et al. ((Hf Gastaut, 1952;H. Gastaut, 1952)). Mu waves appear like pyramidal, arch shape waves which are suppressed on movement, thinking of movement and on watching others move as a sign of neuron mirroring. Mu waves are mainly highlighted in the central regions of the brain (C3,C4) and are most active when a person is at rest or in state of drowsiness with eyes open (Amzica and Silva, n.d.). Mu waves are generally studied in people with Autism Spectrum disorder (ASD). Autism is considered to be influenced by an altered mirror neural system and is viewed as disordered understanding of other people's intentions ((Bernier et al., 2007;Oberman et al., 2005;Williams et al., 2006)). In a study performed on 20 people (10 ASDs and 10 typically developed TDs) within the age group 6-47 years, ASDs showed a supersession of mu wave during own movement as opposed to TDs where mu waves were suppressed during own movement as well as while observing others move (Oberman et al., 2005). However, in another study (19ASDs and 19TDs) conducted on 6-year olds, mu suppression was found on both groups (Bernier et al., 2013).
Rolandic EEG waves are present in both alpha and beta frequencies and exhibit different dynamics in both ranges. Rolandic waves are more active in a 'resting state' (Ritter et al., 2009), while, frontal β rhythms are visible in performing cognitive tasks which are stimuli initiated or related to decision making. Both Rolandic and frontal β rhythms can be present at the same time. In trials including motor-related movement (lifting finger), Rolandic waves were suppressed during the preparation for movement, and were not suppressed when no preparation to make a movement was involved (Kropotov, 2009). Rolandic waves are also observed after an activity is performed when the neural system in the sensory motor strip is relaxing.
Rolandic β rhythms are extremely useful to study the basal ganglia -one of the most crucial areas of the brain including drug addiction, learning and memory and psychopathology ((Aron et al., 2009;Nambu, 2009;Packard, 2001;Watkins and Jenkinson, 2016)). However, β waves (of high amplitudes) are mostly picked up from the front side of the brain and their absence may indicate cortical damage (EEG Brain Signal Classification for Epileptic Seizure Disorder Detection, 2019).
The amplitude of frontal β activity (within 14-16 Hz) is increased during drowsiness or due to sedative drugs (Marcuse et al., 2016). Low amplitude and higher frequency β waves (hi Beta) on other hand are associated with active, busy or anxious thinking or traumatic brain injury (Buzsaki, 2006).
In one of the studies (Baker, 2007), the muscle cortex β activity was found to be suppressed during motion and preparation of motion. Several studies have shown high correlation of β waves in maintenance of cortical state for steady motor output (Androulidakis et al., 2007(Androulidakis et al., , 2006Gilbertson et al., 2005;Kristeva et al., 2007). β activity due to oscillatory activity is elevated in motor disorders such as Parkinson's disease (McAllister et al., 2013).
Several studies also suggest the role of β waves in strengthening and recalibrating the sensorimotor feedback, whereas β waves are reduced during a change in motion or instructions. A reduction in β activity is observed when the motion is resisted ( (Lalo et al., 2007;Zhang et al., 2008)). β activity can be useful in analysing interhemispheric asymmetry. A focal decrease or increase in β waves can relate to brain abscess, stroke, tumours, vascular malformations and cortical dysplasia (Marcuse et al., 2016).

Alpha
Alpha waves (α ) (8-13Hz) (Foster et al., 2017) were first recorded EEG by the Berlin psychiatrist Hans Berger in 1929 from the occipital region of the brain (Karbowski, 2002). They aid in mental coordination and calmness and represent a relaxed state of mind where a person is a rest and not asleep. In human beings, α waves start appearing at four months and matures at 3 years of age (Niedermeyer, 1997). They are predominant when the eyes are closed and attenuated by eye opening and are also measured from the frontal-central region of the brain during REM (Rapid Eye Movement) sleep marked by intense brain activity. Alpha activity in the occipital lobe has been reported to be linked with the inhibition of planned actions and with momentary memory storage in the frontal lobe (Baars and Gage, 2013). Desynchronization (or suppression) of α rhythm due to a corresponding sensory input and synchronisation (marked by an increase in amplitude) due to inhibition of sensory input play a role in cognitive processes of the brain (Quantitative EEG, Event-Related Potentials and Neurotherapy, 2009). In one of the studies, frequencies of α rhythm were found to decrease with age from 7 to 80 years (Kropotov, 2016). However, the variations are relatively small and α rhythms below 7.5Hz should be considered abnormal. Alpha waves can be slightly asymmetric between the left and right hemispheres of the brain. However, a persistent asymmetry of more than 2 Hz is regarded as abnormal. The side with a lower frequency is generally the abnormal one (Aminoff, 2012).
Alpha waves can be classified as α1 (below 8Hz), α2 (8-10Hz), α3 (10-13hz). Some recent studies have demonstrated that an increase in the α3/ α2 ratio is a marker of mild cognitive impairment representing the "at-risk" state of developing dementia ( (Moretti et al., 2013(Moretti et al., , 2012(Moretti et al., , 2011). In another study EEG obtained from 1340 healthy children between 3-12 years indicated that α1 reduces and α2 increases with age, while no change in α3 was observed (Ogawa et al., 1989). The development of the α activity was also found to develop faster in the occipital than the central regions of the brain. Peak alpha frequency during resting state and photo stimulated evoked response was found to reduce in patients with first episode psychosis (FEP) as compared to healthy control groups in a recent study by Murphy et al (Murphy and Öngür, 2019).
In another study where α waves were evoked using a visual-semantic categorization task, it was indicated that the speed of α waves from midline occipital to right parietal sites was inversely related to the reaction time or the fast picture categorization (Fellinger et al., 2012).

Theta
Theta waves are slow waves lying within the frequency range of 4 to 8 Hz that are present in waking adults and may be completely absent in some people (Marcuse et al., 2016). Theta (θ) rhythm is found in around 35% of young adults in the frontocentral region of the brain. Absence of θ on one side of the brain may indicate an underlying cause of structural disease.
Brain's memory system can be found in three parts of the brain viz. Hippocampus, cortex and striatum (Headley and Paré, 2017). Theta waves have been linked to memory formation and navigation ((Lega et al., 2012;Tesche and Karhu, 2000)), and are the strongest in Hippocampus during movement, REM sleep and sniffing (Vanderwolf, 1969).
Some studies have suggested a stronger presence of θ waves during hypnosis or no thought meditation in humans (Aftanas and Golocheikine, 2001). Several studies have indicated relation of theta waves recorded from the frontal midline region with attention processes, memory and arithmetic tasks ( (Gevins et al., 1997;Grunwald et al., 1999;Ishihara and Yoshii, 1972;McEvoy et al., 2001;Sasaki et al., 1996;Smith et al., 1999;Tsujimoto et al., 2006)). Presence of higher frontal θ is more observed in people who are less anxious or neurotic and more extrovert (Mizuki et al., 1984). Higher θ waves can be an indication of improvement in treating patients with depression (Pizzagalli et al., 2001).
Theta waves are a tag for short term memory processing that can either appear due to voluntary (locomotion and REM sleep) or involuntary (sedative state) actions and are termed as Type I and Type II theta respectively (Kramis et al., 1975). Active movements such as running, jumping and exploration in rats was found to be highly associated with θ, where higher θ was found on running faster. During exploration a rat continually gathers and updates information which is stored temporarily in hippocampus. If the exploration bears rewards (or punishments), the inputs to the hippocampus are modulated and the short-term theta memories are converted into long-term memories (Vertes, 2005).

Delta
Delta waves (also known as Zeta waves) are the high amplitude, low frequency (>4Hz) , saw tooth shaped waves which were discovered by W. Gray Walter in 1936 (Magnus andVan der Holst, 1987). Delta waves (Δ) are found in cases of NREM (Non-REM) or deep sleep and their presence in the state of wakefulness indicates cerebral dysfunction. They form an important part of sleep in adults, are more commonly found in the occipital region of babies and are found to reduce during adolescence (Marcuse et al., 2016). Sleep walking and talking occur during high Δ activity (Pilon et al., 2006).
Sleep can be classified in four stages-I to IV. During the initial sleep stage, β waves can be observed which eventually diminish and lead to the visibility of α rhythm. Stage I embarks theta activity in which people are still somewhat awake and alert. Ten minutes after stage I marks the beginning of sleep spindles (Stage II) for around 20 minutes. Delta waves can be seen in stage III when people are less responsive and aware of their external environment. In stage III, half of the brain consists of Δ waves as compared to stage IV where more than half of the brain produces these waves. The sleep stage IV aids in healing and regeneration and is followed by REM stage where the brain begins to dream ( (Colrain et al., 2009;Sekimoto et al., 2011)). Stage IV sleep and Δ waves can be entirely absent by the age of 75 years (Colrain et al., 2010).
Delta waves are preceded by K complex which are the largest event in a healthy human EEG (Cash et al., 2009). K complex can be seen as high amplitude (>100uV) peak (see figure 5) which occurs during sleep stage II. K complex are known to protect sleep by supressing cortical arousal and engage in information processing and memory consolidation (Tononi and Cirelli, 2006).
Research studies have shown that Δ waves are more prominently found in females than males (Ehlers and Kupfer, 1997). Delta waves can be obtained from frontal, temporal and occipital lobes where each location holds significance in terms of diagnosing pathological disorders. Some studies ((Accolla et al., 2011;Watemberg et al., 2002)) suggest that Δ activity found in frontal region during wakefulness may indicate metabolic impairment (renal failure or hyperglycaemia) and brain lesions. Polymorphic Δ activity marked by sharp spikes in temporal regions of the brain is usually a sign of Type II diabetes ((Abdelkarim et al., 2002;Dijk, 2008)) brain injury (Inui et al., 1994) and temporal lobe epilepsy (Walter, 1936), whereas in occipital regions it can indicate absence seizures (Brigo, 2011