EEG

3.1.1 Origin of the EEG

The cortex is organized in cortical columns with different layers and some of which contain pyramidal cells aligned perpendicularly to the cortical surface. This organization is characterized by synaptic connections between neurons of different layers and structures.

Neurons generate electrical currents when activated, which are due to ionic currents at the level of the cellular membranes. There are two types of neuronal potentials: the fast action potentials at the level of the axon, and the slower postsynaptic potentials at the level of the dendrites or soma of the neuron. Action potentials consists in a fast depolarisation of the neuronal membrane (the intracellular potential changes from negative to positive due to the flow of Na⁺ inwards) followed by a flow of K⁺ outwards (repolarisation) and then a rapid return to the resting intracellular negativity. An impulse is then generated which propagates along the neuron, generating local currents outside of the cell which facilitate the propagation of the signal along the neuron. However, these currents are too small to be detected by EEG and the axons are arranged randomly so many of the currents cancel each other out. When the action potential arrives to the axon terminal, neurotransmitters are released into the synaptic cleft, and then bind to the receptors of a postsynaptic neuron, creating a postsynaptic potential, which is caused by ionic channels in the membrane that open up. Depending on the type of neurotransmitter, its receptor and interactions with specific ionic channels, the postsynaptic potential can be excitatory or inhibitory: the membrane of the postsynaptic cell becomes depolarised (thus, more likely to generate an action potential) or hyperpolarised (less likely to generate an action potential). If the postsynaptic potential is excitatory, at the level of the synapse there is a flow of Na⁺ or Ca²⁺ into the neuron, while if it is inhibitory, there is a transference of Cl^- inwards or K⁺ outwards. In the former case, the influx of cations from the extracellular space into the neuron creates a negatively charged local extracellular space (sink).At the distal part of the neuron, there is an outflow of cations from the intracellular to the extracellular space (passive current), which then becomes positively charged (source). In the case of the inhibitory postsynaptic potentials, there is an extracellular source at the level of the soma and a passive sink at the basal and apical dendrites. Therefore, the postsynaptic activity at the soma-dendritic membrane causes a sink-source configuration in the extracellular space around the

neuron, which forms a current dipole between the apical and distal part of the neuron (Figure 2). The neuron can be then approximated as a microscopic electrical dipole perpendicular to the cortical surface [10, 45-47].

However, the current dipole of a single neuron is very small, so individually, they are undetectable by the EEG. Indeed, the activity of a large number of neurons with parallel orientation needs to be synchronised in order to have a measurable electrical field in the scalp. The dendritic trees of pyramidal neurons are parallel to each other and perpendicular to the cortical surface, and thus, are believed to be the main generators of the EEG signal. In order to have a measurable electrical field at the surface about 40-200 mm² of cortex has to be synchronously activated [48]. When the pyramidal neurons are activated by a postsynaptic potential, the longitudinal components of the intra and extracellular currents sum up while the transverse components cancel out. This results in a current along the neurons main axis.

EEG records, thus, the sum of excitatory or inhibitory postsynaptic potentials of a large population of synchronized pyramidal cells that are oriented perpendicularly to the cortical surface (Figure 2) [45].

Therefore, due to the brain geometry, the electrical activity measured by the EEG is mainly generated in the gyrus parallel to the surface, while the sulcal activity is usually unseen by the EEG. The postsynaptic potential should consist of the same type, i.e. excitatory or inhibitory.

The measured EEG signal is influenced by 1) the electrical conductive properties of the tissues between the electrical generators (sources) and the EEG electrode (for instance, the brain parenchyma, the duramatter, the cerebroencephalic fluid, the skull and the scalp) as well as the impendances between the head and the electrode; 2) the orientation of the cortical sources to the recording electrode; 3) the conductive properties of the recording electrode (e.g. size, material, resistance).

Figure 2 – Illustrative representation of a pyramidal cell excitatory postsynaptic potential.

Simultaneous intracranial and scalp recordings have shown that synchronous cortical activity over at least 6-10 cm² of cortex is necessary for the detection of pathological events in scalp EEG [49].

3.1.2 EEG acquisition and scalp field maps

EEG captures the time-varying potentials at certain electrodes placed on the scalp at certain predefined standard positions. An internationally accepted standard is the 10-20 system, in which the electrodes are placed proportionately (10% and 20% spacing) between certain bone landmarks (normally, inion, nasion and preauricular points) in a way that the entire brain is equally covered (Figure 3). The amount and exact placement of the electrodes depend on the application. In the clinical EEG, normally 21-32 electrodes are placed on the scalp using a conductive gel to reduce the scalp-electrode impedance. Impedances should be kept as low as possible (typically <20kΩ is recommend). By convention, the electrodes over the left side of the head are attributed an odd number and those on the right side of the head an even number. The letters of the electrode placement reflect the relative position over the head: Fp (frontopolar), F (frontal), C (central), P (parietal), T (temporal), O (occipital) and A (auricular) and Z stands for midline. The electrodes are then connected to an amplifier and the output reveals a variation in voltage over time: the EEG signal.

EEG is commonly recorded with a sampling frequency between 250 and 2000 Hz, and thus, it has a temporal resolution in the millisecond range. This high temporal resolution of the EEG turns it very suitable to study dynamic and fast-occurring processes in the brain, such as IEDs (more on this in section 2.1.4). The spatial sampling of the EEG depends on the number of electrodes. When the number of recorded electrodes is equal or superior to 64 channels, it is called high-density EEG recording.

EEG consists of a two-dimensional (2D) array with one of the dimensions being the number of time samples and the other the number of electrodes. For each time point, EEG potentials can also be represented as scalp field maps, also called topographies (Figure 3) [50].

The potentials at each EEG electrode are always recorded against a reference. Thus, the recorded EEG potentials represent always the difference between the potential at the electrodes and the potential at the reference. Therefore, the recorded values at the reference electrode are always zero. A common used reference is Cz (thus, the center of the head). If another reference is chosen computationally, the new values at each electrode correspond to the voltage difference between the recorded voltage and the voltage at the new reference. A common used reference here is the average reference, which is the average of the potentials in all electrodes, for each time point. However, it is important to point out that the relative potential differences between electrodes remain the same since the same value is being subtracted from the original measurements for each time point, and thus, although EEG values change, scalp topographies are not affected by the change of reference (the isopotential lines remain the same just their labelling changes) [11, 50]. It has been recommended thus, that the distribution of the scalp field rather than the scalp potential values are used for analyses so that data interpretation is independent from the choice of the reference.

Indeed, it has been recognized that scalp potential maps and whole-head sampling with a higher number of electrodes provide more information about the localisation and orientation of intracranial sources than do the sole interpretation of EEG values and a restricted number of electrodes [11, 51].

In order to visualize the EEG traces, there are several montages that can be used, such as the bipolar montage (potential difference between adjacent electrodes), the double-banana or the triple banana. Figure 3 shows an example of triple banana montage.

In order to quantify the strength of a scalp potential field map, for each time point, the Global Field Power (GFP) can be computed. It includes the differences between all possible pairs of electrodes and thus, it is reference independent. It is defined by the sum of all squared potential differences at each time point:

N t t u t

GFP

N

i







 ¹

))2

( ) ( ( )

(

 (1)

where ui is the potential at electrode i, μ is the average potential of all electrodes and N is the number of electrodes.

Figure 3 - (A) The 10-20 system (adapted from [52]). (B) Triple banana montage. (C) EEG visualised with this montage. (C) Scalp topographic map correspondent to the time point indicated with the red marker in (C)

. 3.1.3 EEG rhythms

The synchronised activity of a population of pyramidal neurons allows the EEG to record oscillations that have a certain amplitude and frequency, depending on the age, awake/sleep state and on the presence of neuronal dysfunctions. These EEG rhythms are generally divided according to the observed frequency content in mainly 5 rhythms or frequency bands [45]: 1) delta: the frequency of the signal is below 4 Hz. The EEG signals have large amplitude. It is the predominant physiological rhythm in adults during deep sleep.

When present in an awake adult it may indicate a brain disorder. 2) theta: frequencies range between 4 and 7 Hz. It is present during certain sleep stages. 3) alpha: frequencies range between 8 and 12 Hz. It is the most predominant rhythm is resting-state EEG in adults, i.e., subjects that are awake with eyes-closed and not performing any particular task. This rhythm is normally localized in the occipital lobe. 4) beta: frequencies range between 13 and 30 Hz. It normally occurs when subjects are with eyes-open during alertness but also with eyes-closed during drowsiness. 5) gamma: frequencies are higher than 30 Hz. It is related to active information processing. In general, when the cortex processes information, neuronal activity is fast but also somehow unsynchronized, which results in an EEG amplitude relatively small. Physiological and pathological high frequency oscillations (HFOs) have also been described [53]. They have a frequency range between 200 and 500 Hz. HFOs can be observed in human intracranial recordings and are supposed to reflect fields of hypersynchronized action potentials [54]. However, HFOs are only rarely recorded with scalp EEG [55] and high sampling rates are required. HFOs are outside the scope of this project and we here focus mainly on the theta, alpha and beta frequency bands.

3.1.4 Detection of epileptic activity

EEG is the most important tool for detecting epileptic activity, and thus, diagnosing epilepsy.

During ictal periods (seizures), the EEG shows a synchronous activity manifested by periodic waves with higher amplitude than the interictal periods (without interictal events).

IEDs are non-physiological transient waves that are clearly distinguished from background activity. IEDs can occur isolated or in brief bursts. In general, isolated, independent spikes do not generate clinical symptoms. When the bursts last for several seconds, they likely represent electrical seizures rather than IEDs. IEDs can be subdivided into sharp waves (duration of 70-200 miliseconds (ms)), spikes (duration of 20-70 ms), spike-and-slow-wave complexes (spike followed by a slow wave) and polyspike-and-slow-wave-complexes (two or more spikes associated with one or more slow wave).

An example of an IED is shown is Figure 4.

Figure 4 - (A) Example of a medial temporal lobe spike (spike peak indicated with a pink maker) and (B) topographic maps in the period within the red box represented in (A) (please note that not all maps occurring during this

period are represented).

Since, in some cases, seizures do not occur very often and thus, might not be detected during the patient's stay at the hospital, the detection of IEDs is the main diagnosis tool. Although the importance of these interictal spikes for epileptogenesis and seizure generation has been debated, they have always been very important in clinics for the diagnosis and localization of the epileptic foci. The emergence of spikes precedes the first seizure in animal models of epilepsy and may guide the development of aberrant neuronal circuits and the initiation of spontaneous seizures [56]. There is also evidence of IEDs related cognitive impairments in epilepsy, notably regarding memory maintenance and retrieval in the hippocampus of TLE patients [25].