Instrumentation and Considerations When Recording

The SEEG signal is obtained by recording small voltage differences between two points in the forebrain, most often within the cerebral cortex (which includes specializations such as the amygdala and hippocampal formation), and less often subcortical structures such as the thalamus and hypothalamus. Subcortical recordings remain uncommon, with emerging clinical use of this information, such as in hypothalamic hamartoma. Recordings of deep cortical malformations resulting from impaired cellular migration are common, such as in periventricular nodular heterotopia.

While early EEG and SEEG recordings were made with analog pen-chart recorders, modern recordings are acquired digitally. The organization of hardware for typical modern digital recording is shown in Fig. 5.1 . As we will see, digital recording has several advantages related to storing raw data that can then be manipulated both online and offline with respect to montage, filters, and other signal processing.

Modern Recording Setup.

The head box typically includes most components of the acquisition hardware, with hardware high pass filter, amplifier, anti-aliasing filter, sampler, and multiplexer. The multiplexed signal is then sent to a wall unit or PCI card for de-multiplexing or sending the signal via a serial or Ethernet connection.

Electrodes

SEEG electrodes are typically made of platinum-iridium (Pt–Ir, 95% Pt, 5% Ir) contacts with plastic spacers between them. Pt–Ir contacts are chosen for their electrochemical properties, and stability is partly due to the iridium oxide layer on the surface of these electrodes. Iridium oxide is more electrochemically stable and therefore resistant to metal deposition during the application of charge (as in radiofrequency ablation and electrical stimulation).

References, Ground, and Recording Contacts

Although reference and ground can be common, they have distinct purposes. The ground electrode on the patient helps to prevent voltage offset between the patient and recording equipment. Because an electrode cannot measure absolute voltage, the EEG amplifier measures the difference between each recording contact and the chosen reference electrode. In stereo EEG, separate contacts on a subgaleal electrode at the vertex are often used for ground and reference due to their consistent signal quality and relatively low susceptibility to artifacts. This most commonly used reference is active, however, picking up vertex activity, especially prominent in sleep. Another approach is to use an extracranial ground that is sometimes switched to a relatively quiet white matter reference shortly after connecting the patient, but white matter references are not inactive and also include intracerebral field potentials. Ground and reference electrodes should have a low impedance and be composed of the same material as the recording contacts to prevent electrochemical current production. The reference often has a similar impedance to the recording contacts. The location of the system reference should not then be changed during subsequent recording, and an alternate computed reference can also be used (see below). Many centers use the system reference for a referential montage, which is often adequate.

Sampling

It is important to consider sampling rate when recording the SEEG signal. The sampling rate is the frequency at which samples are taken from each recording contact from the analog signal. While it has been suggested that the manual review of raw SEEG data can tolerate sampling rates around 100 Hz, this is not recommended for the reasons discussed below. Identifying low-voltage fast activity in SEEG, which often has frequencies above 80 Hz, and embedded higher frequency activity, is critical. To identify low-voltage fast activity, compared to other contacts (typically requiring signal processing as described below), it is critical to sample in a way that captures activity up to at least 200 Hz. Sampling at a minimum of ∼500 S/second is therefore recommended. If there is any current or future interest in high-frequency oscillations or pathological ripples in the SEEG signal, it is necessary to record frequencies of up to 500 Hz. ^, Therefore, it is recommended to sample at 2 kS/second where possible. The large data files generated are often smaller than accompanying compressed video but can make sharing and storing data more onerous.

Noise

Noise arises from many sources and is generally divided into two categories: biological and non-biological. Non-biological noise and artifacts commonly result from electromagnetic interference in the recording environment. Principal among sources is line-noise frequency artifact due to the electromagnetic fields accompanying power lines, switching power supplies, fluorescent lights, and alternating-current power equipment in the immediate environment. Sometimes, interference from cell phone networks and wireless devices is present. These sources of noise can be minimized: if wires are bundled together (so that they encounter the same ambient noise), and a driven ground is used, there can be a very clean signal despite the electromagnetically noisy environment of a hospital room. All recording equipment should also be on the same electrical circuit (or have common ground), and “wall wart” switching power supplies may need to be moved away from the patient and recording equipment. Most importantly, ensure that all recording, reference, and ground contacts have a low impedance. When contact impedance is high, the apparent voltage measured is increased (consider the equation voltage = current × impedance ) and may report high amplitude electromagnetic noise (typically line noise, which is 60 Hz in the Americas and some parts of Asia, mostly 50 Hz elsewhere), rather than biological signal. While a notch filter can be used, it should be avoided where possible given the potential for filter ringing, and removal of some biological signal at this and immediately adjacent frequencies.

Montages

EEG has 1/f amplitude characteristics, ^, meaning that lower frequencies have higher voltages than faster oscillations. Relatedly, slow potentials generally affect a larger tissue region than high frequencies. Exclusion of low frequencies by comparison of near electrode contacts helps to visually emphasize fast activity, which is usually important in defining seizure onset. Contrarily, using distant electrode contacts helps visualize slow activity and ictal baseline shifts.

There are three common types of referential montage in SEEG. Firstly, there is a system reference montage . These are the signals as recorded against a system reference. This is distinct from a computed referential montage, where each derivation is re-calculated against a chosen, often white matter, contact. Unless there is a failure of the system physical reference, one should never move the recording system reference to another site after initial selection—this creates a discontinuity in the dataset. Instead, the computed reference should be created digitally, typically in the reviewing software. An average reference montage , computed after removing noisy or artifact-containing contacts, is typically only used in research.

In addition to the referential montage mentioned above, interpreting the SEEG signal requires a bipolar montage . The bipolar montage is computed by subtracting each subsequent contact on the electrode shaft, starting from the contact at the end of the depth electrode. A laplacian montage may also be useful in interpreting the SEEG signal, though the loss of the end contacts is a limitation. Screening of SEEG data is generally performed with a bipolar montage, as the derivations of this montage generally help discriminate the generators of faster activity and interictal discharges. It is also possible to overlook background suppression and fast activity on a referential montage, given the potential that higher voltage activity from an active reference electrode may obscure biological low-voltage fast activity.

It is also important to note that referential montage provides complementary and necessary information. Very slow potentials such as an ictal baseline shift, typically occur in a larger tissue area, and therefore may be obscured on a bipolar montage that compares signals at a millimeter scale. A referential montage may also help define the more precise location of a phenomenon noted on bipolar review.

Artifacts

The reader is assumed to be familiar with scalp EEG recording and interpretation. In comparison to scalp or extracranial EEG recordings, intracranial artifacts are much less prevalent. Contacts adjacent to the muscles of mastication will encounter a higher frequency, mostly aperiodic electromyographic signal. Rhythmic muscle artifacts encountered from the neck muscles in scalp EEG may rarely be seen. Pulsatile artifacts from CSF flow or adjacent blood vessels are also rarely present. Common-mode biological artifacts and non-biological noise are typically more prevalent when using an extracranial system reference that is also used for a referential montage.

Filters

Filters should principally be used to exclude recording artifacts. A secondary use of filters is to examine specific frequency components of the signal—this use of filters should only be performed with all necessary caveats in mind. The various types of filters are extensively reviewed elsewhere. ^,

There are three commonly used filters in EEG recording: A high-pass filter (allows higher frequency activity to pass, colloquially called a low-frequency filter), a low-pass filter (that allows low frequencies to pass, sometimes called a high-frequency filter), and a band-stop or notch filter that attenuates the signal at a specific frequency. There are several important things to keep in mind when using filters. Firstly, while the cut frequently represents a “corner” in the fall-off of signal power, this is not a clean cut-off but a roll-off. Secondly, filters can “ring” when a sudden deflection in the signal occurs, resulting in artifactual higher frequency oscillations (Gibb’s phenomenon or filter ringing). This is typically a problem with high-frequency filters such as the line-noise 60 Hz notch filter or other low-pass and band-pass filters. These principles are important to consider when the clinician is looking for superimposed fast activity. This may be mitigated by examining the signal with filters off or by using more advanced approaches (see Signal Processing below). Thirdly, and typically of less consequence for clinical interpretation, filters can slightly shift the time base. This becomes important when using different filter settings for channels that are compared in the same recording. Lastly, given the high noise attenuation by the “driven ground” used in modern clinical EEG systems, the need for a line-noise notch filter should alert the reader and team to technical problems and should not be used routinely.

An anti-aliasing filter is often configured automatically. The Nyquist frequency is the highest frequency that can be sampled at a given sampling rate. If the sampling rate is 2 kS/second, the Nyquist frequency is 1 kHz. Frequencies above the Nyquist frequency will “alias” at a lower frequency. In order to prevent this, an anti-aliasing filter is applied that is half or less of the sampling rate. If there were a signal at the Nyquist frequency, this could only be recorded as a sample that occurs twice each cycle and would therefore appear as a triangular waveform. It is typically desirable to have a higher resolution for a sampled waveform, so the anti-aliasing filter is often below the Nyquist frequency.

Types of Intracranial Recording—Cellular and Field Potential

Extracellular intracranial (and intra-parenchymal) electrophysiological recordings in the nervous system are broadly divided into cellular versus field potential recordings. The size of electrode contacts is directly related to the size of the field recorded. Small electrodes can record both cellular activity and highly localized field potentials. When recording voltage fluctuations from a region or field around an electrode, we either remove or do not obtain individual cellular potentials, we speak of a field potential (sometimes local field potential or LFP). Field potential recording measures voltage changes in nervous tissue due to the activity of large numbers of neuronal cell bodies and, more importantly, their dendrites.

Given the size of SEEG electrode contacts, the net activity of millions of neurons are recorded. Such field potential recordings in large “macrocontact” SEEG electrodes arise from the neuropil of gray matter, with individual units’ activity too focal to obtain. This also means that dendritic potentials, accounting for a large component of extracellular space in gray matter, are the source of recordings. Obtaining field potentials is also assisted by the arrangement and morphology of neurons, particularly pyramidal cells, whose cellular long axis is aligned with other pyramidal cells, forming aligned dipoles. Common activity in the dendritic compartments of these neurons enables the recording of higher voltage field potentials. It is important to recognize that when we examine the SEEG signal, we are looking at the activity of dendrites in a large population of pyramidal cells, and these may be sinks or sources of intracerebral currents. This has critical implications for our interpretation of the SEEG signal, since we are recording synchronized population activity in dendrites that may not reflect underlying action potentials in these neurons. If this always held, we would be measuring the inputs to a sampled area, not local neuronal firing. Under normal physiological conditions, high-frequency components of the SEEG signal, particularly high gamma (possibly 50, definitely 80 Hz and above) correspond well to underlying unit activity. ^, In the context of seizures and low-voltage fast activity, gamma activity likely arises principally from local inhibitory neurons and their effects on pyramidal dendrites, but still reflects local neuronal firing.

Scalp EEG Versus the SEEG Signal

While point electric fields decay as an inverse square of the distance, the combined and synchronized action of many dendritic arbors over a larger area of the cortex results in less rapid decay that can even be measured on the scalp. Given the orientation of the dipoles in the cerebral cortex in relation to scalp EEG electrodes, the signal is dominated by dendritic activity in pyramidal cells in the crest of the gyrus. Thus, unlike SEEG, the scalp EEG consists only of summated dendritic potentials of pyramidal cells. As an example of summation, high voltage spindles are relatively simultaneous in frontocentral areas of the adult brain but not completely synchronized in SEEG. Still, on the scalp, they are summated and appear synchronous. Generally, it is held that several centimeters of gyral crest activity are summated in scalp EEG. Thus scalp EEG potentials are recorded when sufficiently synchronous at a particular frequency in a contiguous area of cortical gyri. High gamma activity is typically not measurable in adult scalp EEG (owing to it not being synchronized over a large area and having low voltage). Instead, flattening of the record on scalp EEG likely reflects the involvement of a large cortical area in higher frequency activity, manifesting as suppression.

In contrast, the “listening zone” for SEEG contacts is only up to a few millimeters, providing an accurate review of what is happening locally and is therefore complementary and non-identical to scalp EEG. Furthermore, the ability to record high-frequency activity on SEEG allows one to suspect that there are underlying local action potentials, likely of local inhibitory neurons, ^, measured by their effect on pyramidal cell dendrites, and therefore not just representing inputs from other components of a circuit.

Another key distinction between scalp EEG and SEEG is that the scalp EEG electrodes, as with subdural recording, are always above and orthogonal to the apical dendrites of gyral crest pyramidal cells, as described above. This means that the orientation and location of EEG contacts are not variable in relation to the generators of scalp EEG potentials. This means that much of the normal background in scalp EEG has a typical voltage and polarity, encouraging the use of a standard calibration (7 μV/mm). In contrast, SEEG electrode contacts may have almost any orientation and location in relation to generators. Polarity flips as the depth electrode advances through generators, and there is no standard calibration for all electrodes. However, a starting point is often 100–150 μV/mm, often less, and modified further depending on the structure sampled.

The SEEG Signal Varies Between Regions of the Cerebral Cortex

There are three further implications for the practice of SEEG. Firstly, cellular morphology and architecture determine how much we can obtain high-voltage potentials. Some regions of the cerebral cortex show less cellular alignment or columnar organization producing lower voltage signals and different morphology for epileptiform and ictal discharges. The best example is the amygdala, which has a heterogeneous organization in its various nuclei. From a practical standpoint, this means that a higher gain is needed to see discharges and low-voltage activity. Therefore, the amygdala’s high amplitude spikes and sharp waves are especially salient given that it requires marked synchrony to provide these field potentials. It should also encourage the reader to consider that high-voltage potentials recorded in the amygdala may originate from the adjacent and medial cerebral cortex or the head of the hippocampus. Conversely, the high cell density and arrangement of pyramidal neurons in cornu ammonis of the hippocampus result in high voltage potentials—the gain for these channels can be reduced.

Secondly, electrode orientation matters and will change the orientation of the recording contacts to cortical dipoles. Therefore orthogonal implantation of gray matter tends to provide better recordings given alignment with pyramidal cell dipoles and the bipolar montage that is typical for reading SEEG. Overall, these observations prevent standard voltage calibration in SEEG; not all channels should necessarily have the same gain setting.

Lastly, the frequency of recorded rhythms varies by the network or region sampled, as is discussed below (Section The normal intracranial EEG background ).

Near and Far Fields in SEEG

We have briefly mentioned far-field effects, which are important to consider when reading SEEG electrophysiological studies. The alternative is a near-field potential that arises locally. While direct cortical recordings have been conceptualized as near-fields, in contrast to MEG and EEG far-fields, it seems likely that some far-field phenomena can be recorded on SEEG contacts, particularly in a referential montage. There are two situations in which far fields should be considered: the first is obvious and relates to using a distant reference electrode. The signal can be contaminated with an active reference. This is due to the montage and is not due to far field contamination at the recording contact. More importantly, however, is that large portions of synchronous adjacent cortex show a reduced spatial decay of the electric field. This means that the listening zone of an SEEG contact may be greater for large, synchronized, and more planar generators. Clinical interpretation of SEEG may not need to take this into account, and there is no ready way to do so at present. Theoretically, the large generators associated with seizures may lead to electric fields having less spatial decay than some normal brain activity. This also applies to spatial large ictal baseline shifts. The term “volume conduction” should be avoided and should not be confused with far fields—volume conduction instead refers to tissue conduction (e.g., recording electrocardiogram signal in the feet and hands).

Mesoscale Potentials or Field Potentials?

An alternative set of descriptors can help draw parallels to work in other domains. Arising from studies of brain connections, connectomics , the terms micro-, meso-, and macro-scale are used. Respectively, these terms refer to cellular connections, connections of local processing units (sometimes cortical columns vs. functional units), and large-scale connections between brain areas. This could correspond to microelectrode recording of brain cells (see below), versus local field potentials, versus the activity of large cortical regions, as in EEG. If we make a parallel schema from electrophysiology, SEEG is within the domain of mesoscale potentials, but the difficulty of this being somewhat ill-defined persists, with the size of the local field varying depending on electrode type. Generally, it might be reasonable to use a broad term for the SEEG signal—field potentials—and appreciate distinctions from scalp EEG. In fact, the term SEEG is inaccurate: Intracerebral field potentials are distinct from EEG for the above reasons.

Terminology

Epileptogenic Network —There are three definitions of epileptogenic tissue: (1.) The earliest definition is operational, not theoretical, and refers the network involved in the principal organization of the seizure. ^, (2.) Later, from Lüders and colleagues, the tissue that needs to be removed to cure the epilepsy, and is a theoretical region (styled as “epileptogenic zone”). (3.) And, while it makes the most sense based on the word’s etymology, the most recent and least common usage is: The tissue that gives rise to the patient’s epilepsy. An example of this latter definition might be Bruton’s “dual” or “double” pathology, where one discharging lesion kindles the hippocampus over time (e.g., a tumor or cavernoma). It is important to understand how the term is used in the book or article being read, and the term should be clearly defined whenever it is used. The first definition is most used in the context of the SEEG and, along with the lesional and irritative zones or networks should be clearly documented at the end of an SEEG exploration. As will be further described below, in SEEG these are not theoretical concepts but are based on the SEEG signal and should be described topographically and in relation to networks when possible.

Ictal Transition versus Ictal (Seizure) Onset —It is often difficult to identify exactly where a seizure starts. For example, increasingly prominent bursts of fast activity may appear riding upon discharges which evolve before low-voltage fast activity appears. These discharges are commonly similar to interictal discharges. The question about whether a seizure has occurred is more important than the time of precise onset. It typically makes more sense to identify a period of ictal transition—the time over which findings that may be similar to interictal abnormalities evolve into a seizure, and where these appear causally related to the following seizure. Generally, in SEEG, it is more helpful to think about the ictal transition and commonly preceding ictal preparation. Seizure onset can be appropriate when the ictal transition is sharp and clear.

Ictal Preparation (“pre-ictal” phase) is the period preceding or overlapping with the ictal transition, which is typically similar to interictal epileptiform abnormalities but is reproducibly and convincingly causally related to the ictal transition. For example, repetitive spiking might always be seen before the ictal transition in a particular electrode.

Anatomo-Electroclinical Correlation —this concept is at the heart of SEEG methodology and refers to the interpretation method of SEEG. ^{, ,} One must correlate the anatomical substrates of the seizure with the electrophysiological pattern and clinical manifestations throughout each part of the seizure. The missing yet critical piece of information from this triumvirate of anatomy, EEG and clinical semiology is the pathophysiology. Electrophysiology patterns and clinical manifestations are determined by the precise large-scale, local and microscopic circuits involved in the seizure. The involvement of these circuits is in turn determined by the underlying pathology and connectivity. In many SEEG patients, the pathology is unclear. Even when the etiology is known, the way that this gives rise to seizures and affects cortical microcircuits is often unclear. Similarly, the connectivity of the human forebrain is incompletely understood, as is how pathology modifies this connectivity. Modern clinical and fundamental research in epilepsy should address these questions to allow clearer anatomopathological electroclinical correlations in the future. This incomplete knowledge leaves some unaccounted variability in seizure onsets, as will become clear below where we discuss the electrographic patterns of seizures.

Subdural Versus SEEG Concepts of the Symptomatogenic Zone, Irritative Zone, Lesional Zone, Ictal Onset Zone —these concepts have also been defined in a more theoretical way in the context of subdural recording, as developed and popularized by Rosenow and Luders. While each has a clear definition, the concepts of symptomatogenic and ictal onset zones do not fit as well with the SEEG method and approach. One of the main reasons in each case is that symptoms/signs, seizure triggering, and seizure onset are all conceived as network phenomena in SEEG, rather than being foci. In Rosenow and Luder’s schema, the symptomatogenic zone is conceived of as a localized area that gives rise to the symptom(s) characteristic of the patient’s seizure. This has some limitations: In SEEG, the clinical correlation unfolds over time, so the area of the symptom generation migrates. Secondly, in SEEG, the presumption is that network activity (i.e., across a group of connected structures) is modified, not a single focus. Additionally, the seizure may be asymptomatic at onset. Finally, clinical and subjective manifestations differ in correlation with ictal electrophysiology according to not only spatial (anatomic) but also temporal features of epileptic discharge (see below). A common example of all three concerns: while the symptoms of a somatosensory aura may be localized to a somatosensory “symptomatic zone”, it can be that the high-frequency activity in the motor cortex does not result in a movement but drives the somatosensory cortex to produce a somatosensory symptom.

The irritative zone term is used on both SEEG and in subdural recording and is defined as the area that demonstrates interictal abnormalities. This may be linked to the regions that produce the seizure or give rise to the process of epileptogenesis itself. Further complicating matters, while the term “irritable cortex” is circular (on scalp EEG the cortex can be said to be irritable because of interictal discharges, and that it has interictal discharges because it is irritable without any further useful conceptual information being supplied), it is often used to refer to any cortex where interictal discharges are noted. In SEEG, this is operationalized by documentation of the regions or networks in which interictal discharges occur. The primary irritative zone defines the spikes located in the epileptogenic network, and the secondary are those outside—these can only be defined once seizures are recorded.

The Lesional Zone is electrophysiologically determined in SEEG, particularly by the presence of continuous or sub-continuous slowing, not to be confused with a radiologic lesion. This term was first defined in SEEG in the setting of tumors. In contrast, Rosenow and Lüders defined the “epileptogenic lesion” as hypothesized or actual tissue pathology and best identified by radiologic findings. This would include areas or networks that show pathologic slowing or interictal discharges. This can, again, often be a network rather than a discrete zone. It is important to note that the lesional zone often is distinct from, or extends beyond, an area of imaging abnormality.

Finally, the Ictal Onset Zone can often be more than one discrete area of cortex and refers to where one can see the point at which clear evolution of the seizure commences. The terms ictal onset as a point in time, and either early ictal network or ictal onset network may be preferred. In SEEG it is important to describe the electrophysiology (with topography and clinical accompaniments) of seizure onset, rather than describing this as an anatomical zone alone.

The Normal Intracranial EEG Background

The normal intracranial EEG background was first described in the era of Jasper and Penfield from intraoperative recordings from the pial surface. The anterior-posterior gradient evident on scalp EEG recording was redemonstrated in these studies: they described a preponderance of beta activity anterior to the central sulcus, with posterior alpha activity that increased in voltage as electrode recording approached the occipital cortex. Though these studies were performed solely by visual analysis, more recent intracranial data, which allows for frequency analysis and improved anatomical resolution, has generally substantiated the anterior-posterior gradient of frequency and amplitude.

While there is at least one comprehensive atlas of subdural EEG electrophysiological findings, there is not yet a comprehensive atlas of intracranial findings on SEEG. There is ongoing work in this area and there are descriptions of background frequencies by brain region. Frauscher and colleagues collected data from patients undergoing intracranial EEG across three surgical centers, most of which had depth electrodes. In wakefulness, the posterior dominant rhythm correlated with a robust alpha peak in the occipital lobe; a less prominent alpha peak was seen in the parietal lobe, which also exhibited an additional beta band. Higher beta and gamma bands were more anteriorly found in the frontal lobe ( Fig. 5.2 ). Wang and colleagues found an additional tendency toward higher frequencies in the medial cortices.

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

Creation and Evolution of SEEG The Departmental Model for Practice 9 Large Aneurysm of the Internal Carotid Artery Obliterated with Seven Fenestrated Clips under Intraoperative Monitoring of Anterior Choroidal Arterial Blood Flow Insufficiency Lumbopelvic Parameters Nerve Root Anomalies with Application to the Lumbar Plexus 22 Radiation Therapy and Stereotactic Radiosurgery for Large Vestibular Schwannoma

Stay updated, free articles. Join our Telegram channel

Join