Introduction
The challenge in epilepsy surgery is to cure the patient without causing an additional neurological deficit, so the maxim “primum non nocere” should be kept in mind throughout the presurgical investigations, and in particular during SEEG. The goal of the method is to delineate the epileptogenic zone and propose a brain volume for resection without compromising post-operative cognitive, motor, sensory, or socio-emotional brain function (although most emphasis has been on language and motor skills). Functional mapping is therefore, alongside identification of the EZ, a crucial part of the SEEG exploration. The identification of cortical structures essential to cognitive or perceptual function, with great anatomical and physiological precision is challenging. Indeed, it is widely agreed that the human brain is a complex and adaptive system in which a vast range of function arises from coordinated neural activity across diverse spatial and temporal scales. For instance, graph-theoretical investigations have shown that the human brain exhibits a hierarchically modular organization with clusters of nodes (hub, subnetworks) that are densely connected within the cluster, but only sparsely coupled to nodes in other modules. Accordingly, a modern concept of functional mapping needs to take into account this organization and enable the identification of the hub or node essential for function.
The challenge of functional mapping in epilepsy surgery lies in the pathology itself. The lesion underlying drug resistant epilepsy (cortical dysplasia, stroke, cortical malformation, heterotopia) may result in a functional organization that is completely different from that of the healthy subject. In the case of dysplastic tissue, there may be no clear boundaries between the dysplastic and functional networks, as dysplastic cortex is specified prenatally and thus integrated into developing cortical networks. FCD differs from pathologic entities acquired in postnatal life such as tumors, gliosis, and vascular diseases, which often destroy existing networks. Of note, however, research has found surprising consistencies between cortical stimulation mapping results when comparing patients with either early or late onset epilepsy and a third series of patients with new onset, fast growing tumors. Consequently, the clinician assessing language in the epilepsy surgery setting must explore the possibility that language functions may have atypical neural substrates, while being cognizant that for many individuals there may have been no significant reorganization of function (particularly after early epochs of brain development have passed) (see Drane and Pedersen, 2019 for an extended discussion of this topic). Therefore, an individualized and comprehensive assessment should always be done during the presurgical investigation.
Since the early days of presurgical mapping, electrical stimulation was developed to identify essential functional areas. The main principle is simple: electrical stimulation of a specific brain region producing a transient functional impairment might predict which functions will be disturbed if the stimulated cortex were to be removed. For instance, as previously described using ECoG or awake surgery, stimulation is useful to identify essential language cortex in patients undergoing resection of epileptogenic cortex in the specialized language hemisphere.
Despite its proven utility, and the fact that functional mapping is commonly performed during SEEG presurgical investigation, stimulation remains challenging and contains pitfalls because of the technical constraints and the requirement to adapt testing at the individual level. Addressing the question of the sensitivity and specificity of the SEEG-stimulation against reference meta-analytic fMRI studies, authors confirmed that stimulation can reliably identify contacts with/without language function but may under- detect all language sites. Indeed, the major pitfall with potential clinical consequences for the patient, the failure to identify a specific functional region, is a “false negative stimulation,” and has led some to question the validity of this method. Nevertheless, fMRI studies cannot determine which brain areas are essential for function, and it has frequently been shown that such activations may show regional involvement that is not necessarily required for successful task completion.
It should also be noted that “false-positive” errors may also occur in the context of cortical stimulation mapping. The main causes are likely to involve patient fatigue or a lack of appreciation for the effects of after discharges and the extent of electrical spread. These potential problems may be minimized by establishing a baseline performance that is performed at a near perfect level by the patient, making sure that they are not becoming overly fatigued during the mapping session, and carefully attending to the EEG data. In such cases, surgery could be denied, which could have more than likely been safely performed.
The procedure of stimulation during SEEG for eliciting seizures and for functional mapping should be done in conjunction to answer the fundamental question of whether or not there is a spatiotemporal overlap between the epileptogenic and the functional network. Second, to address the question of functional mapping, it is important to know the physiological network organization of the system under exploration. Third, cortical stimulation for functional mapping cannot be unambiguously interpreted in isolation. The integration of physiological rhythms, the presence of a lesion or a cortical malformation must be integrated in the procedure, as well as in the analysis and in the interpretation of the result obtained. Of note, while part and parcel of the traditional French school of SEEG, Americans trained in the “grid and strip” practice of mapping may be less familiar with purposeful seizure induction during intracranial monitoring, and have often been taught that the seizures. occurring through stimulation differ from naturally occurring events (see the following articles for a discussion of eliciting seizures, understanding semiology, and the history of the North American and European schools of stimulation mapping). The rise of minimally invasive surgical tools and procedures has seemingly shifted the American programs to replace the grids and strips approach with SEEG. Of note, however, there has been an uneven adoption of European practices and theory, and many American programs are still finding their way.
What is the role of functional mapping in an individual patient?
What Do We Need to Know Before Performing Stimulation?
Stimulation for functional mapping cannot be done in isolation of the whole surgical epilepsy exploration. Before addressing the question of function, several factors have to be taken into account.
Patient data
A comprehensive clinical assessment should be performed prior to functional mapping by stimulation, in order to determine any pre-existing functional deficits. Some deficits, such as a mild hemiparesis, a proprioceptive deficit, or a quadrantanopia, may be clinically evident, while others, such as a visual agnosia, alexia, prosopagnosia, anomia, or dysexecutive function may only be elicited by detailed neuropsychological assessment, which should be carried out in every patient.
Clinical factors that influence the likelihood of functional reorganization should also be determined, including the age of onset of epilepsy, and the presence and type of any underlying lesion.
SEEG data
Knowledge of the precise anatomical position of each electrode contact is crucial to plan the stimulation procedure. First, the testing procedure and the parameters of stimulation must be adapted according to the different physiological system explored. Second, as bipolar stimulations are performed between two adjacent contacts, the exact position of the contact, such as bank of the sulcus, proximity to white matter, orientation in the cortical strip needs to be checked to interpret the effect of the stimulation. For instance, stimulation between two contacts in gray matter may have a different effect to stimulation between one contact in white matter and one in gray matter. It is possible that stimulation within the gray matter may induce more focal dysfunction whereas stimulation within white matter could produce disconnection phenomena. Third, in addition to the anatomical relations of the electrode contacts, their location in relation to an underlying lesion, and the nature of the lesion should be determined. In addition to correlation with neuroimaging, some lesions, such as cortical dysplasia or ischemia may be associated with characteristic patterns of resting SEEG recordings. It is widely accepted that in several types of cortical malformation such as FCD, , schizencephaly or heterotopia, , functional connectivity between lesion areas, and regions of normal cortex may occur. It is therefore relevant to note whether signal organization within the supposed lesion is physiological or not. The relationship between each contact and the network organization of spikes, high frequency oscillations and, if already available, the network organization of the seizure, must be considered in stimulation planning.
Physiological network
Since the epileptogenic and functional networks may closely overlap, a knowledge of the typical physiological organization of functional networks is essential. Functional connections between two areas, or modular organization of a specific cognitive function are sometimes relevant to the interpretation of a stimulation. We would caution that this is still an evolving area, as the neural circuitry of many functions are complex and not fully understood, and even the cognitive constructs we attempt to measure (e.g., memory, attention, emotion) are undergoing rapid theoretical advancement. Paradigms to employ in stimulation mapping will likewise be expanding over time and a virtuous circle can be created between the research program and clinical activity. , ,
Even if stimulation is performed in each contact exploring gray matter, choosing the most informative pair of contacts may be complex. Electrophysiological data such as evoked potentials in response to auditory stimuli (pure tones), to visual stimuli (checkerboard, picture, face), to somatosensory stimulation or to cognitive tasks (oddball paradigm, naming, etc.) are useful to guide the selection of the relevant contact in the area explored.
Broad-band high gamma activity, extracted from the SEEG signal, has been shown to be a useful general electrophysiological index of cortical processing, and high gamma responses are consistent across a wide variety of experimental tasks and cortical regions, such as language, vision, and eye movement. These data can assist in the selection of testing performed during the procedure. A recent study elegantly demonstrates such an approach before stimulation in SEEG. In addition, Fig. 10.4 illustrates one clinical case in which the recording of gamma activity before stimulation helped to choose the task set, and to understand the underlying processing involved.
Stimulation Mapping in Clinical Practice
We describe here the general procedure for stimulation, as already reported in a paper and in the expert consensus clinical practice guidelines proposed under the auspices of the French Clinical Neurophysiology Society. Aspects relevant to specific functional networks or cortices will be discussed in more detail in the following chapters. As already outlined in the chapter on stimulation to induce seizures, the stimulation protocol must be adapted, according to the different requirements of individual cases, rather than following a rigid protocol. Stimulation planning must take into account all the data described below.
Timing
Stimulation is undertaken during two to five sessions distributed over several days that generally take place after the recording of spontaneous seizures. Distributed sessions should be performed both for patient comfort and for electrophysiological reasons: repetition of stimulation in the same site tends to produce a refractory period and false negative stimulation. The duration of each session should be determined for individual cases, but is usually between 30 and 60 minutes. One important parameter is the time between each stimulation. The physiological state of the tested brain structure before stimulation must be correctly appreciated. It is necessary to have an observation period following each stimulation session, because stimulation may modify the network causing, for example, slowing or after discharges. This is why we propose to space each stimulation by at least 2–3 min and to avoid repeating stimulation of the same structure within a short time frame, to prevent plasticity or summation effects. The EEG should return to its resting state before stimulation is repeated.
Stimulation for functional mapping is often carried out in the context of normal anti-seizure medication (ASM) or very slight reduction. Because it is completely integrated with the SEEG exploration, stimulation can trigger seizures during the procedure.
Stimulation parameters
Electrical stimulation between two adjacent contacts of the same electrode (bipolar stimulation) is typically practiced. Fig. 10.1 depicts the different type of parameters used. The charge density delivered is influenced by the position and orientation of the electrode, as well as the type of cortex stimulated. The parameters used in clinical practice are established for platinum–iridium semiflexible multi-contact intracerebral electrodes 0.8 mm diameter and 2 mm long, with contacts at intervals of 1.5 mm. Stimulation is applied between two adjacent contacts utilizing one or both of the following modalities: low frequency pulse stimulation (1Hz) or sustained train of pulses (50Hz or 60 Hz). Pulse width and train frequency generally are kept constant throughout the whole stimulation procedure (0.3–1 ms). Low frequency pulse stimulation consists of a rectangular biphasic single pulse, 2 ms pulse duration, 0.5 m A to 4 mA delivered every second for a continuous period of up to 40 s. High frequency pulse stimulation consists of a train of rectangular pulses with a 50Hz frequency, 1 ms pulse duration, 0.5–2.5 mA intensity, and 3–5 seconds train duration. These parameters were adjusted to avoid any tissue injury with a charge density per square pulse of <55 μC/cm2. Single pulses have been used for functional mapping, but are less effective in particular to map associative cortex. As a rule, stimulation is begun with low current intensities and gradually increased to reach threshold either for clinical manifestations, such as a positive effect or negative effect of the stimulation or electrophysiological effects such as an afterdischarge or seizure. The initial intensity may vary according to several factors: (i) level of current ASM, (ii) history of generalization (particularly when stimulating lateral premotor or precentral cortex), (iii) whether the area stimulated is proposed to be part of the EZ (as indicated by already recorded spontaneous seizures), (iv) type of cortex stimulated.

To reduce the risk of a false negative stimulation, both the charge density and the duration of the stimulation must be sufficient. Low intensity and inadequate duration of the stimulation have been reported as a cause of false negative stimulations, while in the associative cortices a long lasting stimulation has been shown to increase the occurrence of positive responses.
This stimulation paradigm using adjacent contacts ensures a good spatial specificity with respect to the structures targeted for stimulation. Our choice of bipolar stimulation presumably produces a more focused electrical field thus leading to more accurate anatomical localization of less than 5 mm spatial accuracy, centered around the stimulated dipole. A recent article found that some of the differences in stimulation intensity can be explained by the observation that responses largely depend on the applied charge per phase taking the variable pulse durations into account. According to electrode size (diameter of 0.8 mm, contacts 2 mm long separated by 1.5 mm from one another), the current charge density/cm 2 /phase, the current density/cm 2 /second and the total charge density/cm 2 can be calculated according to stimulation intensity and train duration ( Fig. 10.1 ). Because the variation of pulse width across center (0.3 or 1 ms), it is preferable to use the total charge density than the intensity in mA to be sure to have an effective stimulation.
Task used
In functional mapping, selection of the task needs to be tailored to the patient level of functioning and to the site of the stimulation. The choice of tasks performed depends on the functional role of the stimulated regions so knowledge of the physiological network is particularly important. A task which is not relevant to the functional system stimulated, and the absence of behavioral assessment of the patient may result in a false negative stimulation. , The analysis of the behavioral response of the patient should take into account both the type of errors and the reaction time. Some recent articles on cognitive and socio-emotional mapping with SEEG have included charts of structure-function relationships observed using cortical stimulation mapping in order to help guide task choice.
Language
Physiological Background
As previously expressed above, stimulation for functional mapping requires knowledge about the physiological network involved in the function. We summarize here some important points about language organization.
It has long been known that the language network is asymmetric between left and right hemisphere: only left hemispheric lesions induce language disturbance in patients with left lateralized language dominance. The idea of a “dominant/major” left hemisphere controlling both language and right hand against a “nondominant/minor” hemisphere was proposed. Thanks to studies of split-brain patients, the specific role of each hemisphere has been described. The left hemisphere hosts linguistic functions, such as phonology and syntax, and the right brain is involved in paralinguistic functions, such as emotional and context processing. The term of “specialized language hemisphere” rather than dominant/nondominant is arguably more appropriate.
Within the left hemisphere the simplistic view of a network between two restricted regions – Broca’s and Wernicke’s area has been updated. Neuroimaging and electrophysiological data have clearly shown that the language network is widely and mainly distributed around the sylvian fissure. Converging data supported the idea of a dual stream model of organization of language processing with a dorsal stream involving in mapping sound to articulation, and a ventral stream in mapping sound to meaning. From the superior temporal gyrus, which is engaged in early cortical stages of speech perception, the system diverges into two processing streams. The dorsal stream, or the “auditory–motor integration,” runs dorso-posteriorly through the inferior parietal region and further to motor and prefrontal areas, Broca’s area in particular. The posterior–anterior “what” pathway or ventral stream projects laterally to the middle and inferior temporal cortices and serves as a sound-to-meaning interface by mapping sound-based representations of speech to widely distributed conceptual representations.
Within the ventral network, a region localized in the basal temporal region has been described to be included in the language network. The basotemporal language area (BTLA) located 2–9 cm from the tip of the temporal lobe, and surrounding the collateral sulcus (CS) and occipitotemporal sulcus has been described for the first time in the context of epilepsy surgery. , Since the first description, it is well known today that BTLA is functionally heterogeneous, involving several sub-regions. Visual perceptual processes predominate in the posterior part of the region, which is located relatively near the striate cortex, whereas the territories involved in language processes contributing to lexical retrieval are more anterior. The anterior part is bilaterally involved in multimodal semantic processing and responds to the retrieval of word meaning. Its activation was previously found to be associated with semantic activities focusing on phonological decisions regarding auditorily presented words. Lesions confined to the anterior part of the BTLA are known to impair patients’ semantic processing performances. Several parts of this region, in particular the fusiform gyrus, contribute to the language network underlying object naming.
In the context of epilepsy surgery, the cortical network underlying visual object naming should be well understood for two reasons – it is the most prevalent deficit present in patient with epilepsy, and the picture naming task is to the most widely used test during presurgical investigations. Verbal picture naming recruits a widely distributed network of cortical areas, predominantly located in the left hemisphere. The network starts with occipital and ventro-temporal structures. From 200 ms onwards, temporal structures are engaged in lexico-semantic processing (activation of the meaning of the picture and its possible names); later, inferior parietal cortex and posterior temporal lobe are associated with phonological encoding. The left inferior frontal gyrus is thought to resolve conflict among alternative representations, as well as syllabification processes. Finally, bilateral pre-motor and motor areas, as well as the inferior frontal gyri, are engaged for articulatory planning and articulation. In addition to this well-defined network some studies have pointed out the specific role of the hippocampus in lexical selection during picture naming. ,
Lesion analysis studies have demonstrated that visual naming can be impaired by surgical lesions occurring nearly anywhere along the ventral visual pathway, with the lateral temporal pole appearing to be more associated with proper noun retrieval. Different types of objects appear to be differentially distributed through this visual stream, and in this age of more focal, minimally invasive epilepsy surgical procedures it is common to see both improvements and declines in visual naming resulting from a single destructive procedure in a given patient. Despite the aforementioned finding of a possible role of the hippocampus in lexical selection during picture naming, more than one study has demonstrated that visual naming does not decline following the ablation of the left amygdylar-hippocampal complex. , Of note, the neural substrates of learning naming associations has been less well studied in the setting of neurosurgery.
In summary, the language network is asymmetric, mainly involving the left hemisphere and widely distributed between several temporal regions and the inferior frontal lobe.
What Can Functional Mapping Tell Us About Language Organization?
So functional mapping in the context of epilepsy surgery must answer two main questions: firstly “what is the language organization in terms of hemispheric specialization?”; secondly “What are the crucial nodes of the network underlying function?”
Stimulation and hemispheric organization
As described above, the left and the right hemisphere have different functions during linguistic processing, but the idea of a simple dichotomy between a homogeneous left hemisphere dedicated to language, and a right hemisphere dedicated to other functions has to be re-examined. First there is a continuum between left and right specialization, and second the hemispheric specialization must be assessed region by region. However three main language representations may be distinguished:
- (1)
a typical representation consists of principal involvement of the left hemisphere
- (2)
an atypical organization including a range of configurations between left and right and
- (3)
an atypical right representation with principal involvement of the right hemisphere.
As we described in the beginning of this chapter, chronic epilepsy or a brain lesion such as stroke or dysplasia may induce a plastic change in language representation, resulting in the higher proportion of atypical language representation that is well recognized in this population.
The issue of hemispheric specialization is usually assessed before SEEG, using several noninvasive tools such as fMRI. , The techniques of fMRI study have previously shown dissociation between left and right hemispheres, for instance with the left IFG and the right STG involved in a language task. , Using ECoG, stimulation has already proven to be useful in this type of interhemispheric dissociation, Fig. 10.2 .

Identification of crucial node for the function
Because the language network is widely distributed in the left (dominant) hemisphere, and often varies among patients, brain mapping should delineate eloquent areas at the individual level. Stimulations have to identify the crucial node for language function. In order to achieve this, as described above, two aspects of the stimulation protocol are critical: (1) to disrupt the language function the stimulation needs to be strong enough to exert a complex effect in a given volume of brain tissue. Stimulation parameters must be carefully chosen to avoid the risk of false negative stimulation. (2) To detect the effect of the stimulation, the task and the analysis of the patient’s behavior must be appropriate to the region stimulated.
Methodological Aspect: An Effective Stimulation
In associative networks and particularly in the language network, the stimulation needs to be sufficient to apply an effect in the network.
We have found that to prevent the occurrence of false negative effects, stimulation duration around 5 s and intensity up to 1.5 mA are required for language functional mapping in SEEG. Bedos Ulvin and colleagues used the same range of intensity but with a longer duration 5–10 s. The length of the stimulation is clearly an important parameter, a stimulation too short (<4s) may not be enough to disorganize the underlying network . This is in keeping with previous SEEG stimulation studies. , ,
The recording of an AD or evoked spike AD helps to establish that the effective stimulation threshold of a given structure has been reached. The spatial extent of the AD is important in interpreting the effect of the stimulation. In contrast to local ADs, a remote or regional AD leads to greater complexity in the interpretation of functional effects. The occurrence of an AD depends on the total charge density delivered, but is also dependent on the type of cortex stimulated. For example, ADs are seen more frequently in mesial temporal lobe structures.
Few studies have addressed the question of the effect of an AD on the clinical response. Some studies considered the AD as “pathological” and stimulations producing an after-discharge were excluded from analysis. , , On the hand, the majority of studies considered stimulation trials with AD in their analysis. , Conceptually, the presence of an AD can reveal a potentially “nonfocal effect” of a stimulation. As is now well recognized, brain regions are highly interconnected and constitute anatomo-functional networks. The behavioral effects of stimulations critically depend on functional connectivity. Electrode sampling of the area explored and stimulated has an important impact on the ability to capture/observe the whole effect of the stimulation, and should not be underestimated in its interpretation. Finally, a short and focal AD confirms that the threshold of the stimulated structure has been reached. Avoiding false negative effects, it increases the specificity of the technique.
Methodological Aspect: The Difficult Task Choice
Within the widespread peri-sylvian language network, different regions have different latencies and linguistic functions. To be effective and avoid false negative stimulation, the patient needs to be engaged in a task adapted to the topography of the testing region.
Practice varies widely across centers as there are no adopted guidelines for cortical stimulation mapping. We list here the tasks usually used in both our centers:
Naming tasks includes a variety of categories of objects, entities (humans and animals), and landmarks, and allows for the evaluation of both common and proper nouns; Automatic speech (counting); repetition of words and sentences; repetition/designation task; tasks include both auditory and visual components, with an effort to examine the same stimuli in different modalities (e.g., the picture of a rooster, the sound of a rooster, and the verbal description of a rooster presented in different modalities); semantic decision tasks; reading aloud of sentences and words; testing in more than one language for bilingual patients; verbal diadochokinesis (the patient is asked to repeat “pataka” 10 times).
The sensitivity of stimulation to produce a functional effect varies according to the task used, as we have previously reported. During automatic speech, stimulation induced fewer positive responses (61% positive/39% negative), compared to reading aloud (78%/22%) or naming (76%/24%), which were more sensitive.
In the posterior part of the left superior temporal gyrus the effect of stimulation is different according to the task and the area stimulated ( Fig. 10.3 ). During a word repetition task hallucinations or illusions are observed when Heschl’s gyrus was stimulated without any language deficit, while the stimulation of the planum temporale (PT) induced auditory symptoms along with comprehension deficit. Articulatory or phonological errors are elicited by the stimulation of the left PT during word or pseudo word repetition, presumably due to a difficulty to maintain task-relevant representations in a phonological loop. Lastly, the posterior part of the left superior temporal sulcus (STS) seems involved in more high-level language processes required in naming and reading tasks, because its stimulation did not induce positive auditory symptoms but naming or reading deficits. The reading deficit included graphene decoding, comprehension deficit and graphene to phoneme deficit.

In the medial and basal part of the temporal lobe, the choice of task also produced different results during stimulation. Stimulations performed during naming or reading were more likely to induce a language deficit than during a repetition or automatic speech task. As noted, there can also be differences based upon object or face type (e.g., naming animals vs. objects vs. famous persons or landmarks), and based upon the level of naming specificity (e.g., common vs. proper nouns). In the inferior frontal gyrus or in mesial temporal anterior part (hippocampus, anterior part of the parahippocampal gyrus), the naming task appears to be the most sensitive. In the insula, even with a small sample, we found more elementary speech arrest or slowing in automatic speech than in the inferior frontal gyrus or in anterior temporal lobe.
In the basal temporal region Bedos-Ulvin et al. have demonstrated a disruption during stimulation of a naming task, whereas with the same parameters of stimulation, at the same site, a sematic task such as the semantic picture matching task (picture version of the Pyramids and Palm Trees Test) was perfectly performed. The low sensitivity of certain tasks has also been reported with cortical stimulation performed with grids or during awake craniotomy, in particular the low sensitivity of the automatic speech task.
The importance of task selection should be emphasized because of its clinical and scientific implications. In line with previous reports with subdural grids during presurgical testing, our experience highlights that it is vital to appropriately select the task according to the area explored by the electrode. Fig. 10.3 summarizes the tasks that we propose to be appropriate to different stimulated areas.
How to Interpret the Effect of the Stimulation
Outside the methodological pitfalls, language stimulation for functional mapping may be challenging due to other issues inherent to the patient behavior itself. On occasion, the same stimulation with the same parameters does not induce the same effect; sometimes testing must be adjusted to the level of the patient’s language ability.
Variability of the responses
Empirically, a language response seems to represent an inhibitory effect of the stimulation due to the temporary inactivation of a local population of neurons (pyramidal cells or interneurons). However, we are far from a perfect understanding of the functional effects of stimulation. When we apply a train of stimulation, we do not know precisely the physiological effect of the stimulation. We regularly observe variability of effect when stimulating the same region, with the same parameters of stimulation. One explanation could be that the “inhibitory effect” of the stimulation induces a rapid plasticity of the system during the minutes following the trial. This is why it is recommended to: (1) verify that the SEEG signal returns to its resting state between stimulations, (2) to space stimulations, and (3) to avoid subsequent stimulation of the same site.
Indeed, since stimulation is disrupting a network, can the effect of stimulation at any given site really be considered as focal? While precise analysis of local effects of stimulation is important, it seems very likely that other effects distant to the stimulation site are involved. A systemic effect of stimulation is an important question to consider for interpretation.
How to manage the task according to the level of patient
In the naming task, another parameter needs to be taken into account, which is the linguistic property of the stimulus used. It is well known that low frequency items such as “penguin” or “hippopotamus” are more difficult to name in term of latencies than the frequently used items “dog” and “cat.”
Ideally, the patient should initially be familiarized with all the pictures, to ensure that they can accurately name them. But on the other hand, if we use a limited number of items several times, another issue arises: the fact that the network involved in naming changes according to the number of times that the item has been named. , Indeed, in a block naming task, in which the same picture is named six times, we have observed a change in the event related potential response after only the second presentation of the same picture. Therefore, it is likely that the network involved in a task repeated two or three times involves the network differently. On the other hand, in patients pre-existing naming difficulties or in left temporal lobe epilepsy, where such difficulties are common, we find a more reliable interpretation is possible when a set of 60 items well known to the patient is used.
Values of different types of errors?
The definition of a positive response to stimulation can be difficult. When the patient stops a naming task and then recovers the name a few seconds later, the result is quite obvious. Other types of errors may be more subtle, e.g., a slightly prolonged latency. In our practice we take into account effects not only occurring immediately during the stimulation, but also 10–20 s later, due to the complex effect of the stimulation, and possible summation of the effect. As we illustrated in Figs. 10.3 and 10.4 the type of errors is important depending which structure is involved. For instance, a naming deficit due to lack of lexical access with preservation of semantic access ( Fig. 10.4 , stimulation of aBTLA) allows us to identify “positive” areas as a specific functional sub-unit underlying lexical access.

Conversely, we must keep in mind that the process underling visual naming cannot be treated as a homogenous, unitary phenomenon, but rather a complex, multistep process that involves multiple brain regions. , This point is particularly well illustrated in the case of the patient in Fig. 10.4 , in which we were able to tease apart several steps of the process during a naming task.
Memory
It is well known that successful memory relies on a series of distinct cognitive functions that may be carried out in a distributed manner throughout the brain, in which temporal lobe structures and the hippocampus play a central role. Theta frequency fluctuations of the local field potential recording in the hippocampus have long been implicated in learning and memory. Successful memory is associated both with increased narrow-band theta oscillations and a broad-band tilt of the power spectrum. Theta oscillations specifically support associative memory, whereas the spectral tilt reflects a general index of activation. Hippocampal stimulation may induce memory deficits through disruption of physiological oscillations. Although memory functions are critical in temporal lobe epilepsy surgery, and systematic stimulation studies of the hippocampus pertinent, only a few studies have been performed using depth electrodes. The methodology used in the different studies are relatively varied, and raise the questions of what tasks should be used, when should stimulation be applied, and what type of stimulation is appropriate? Moreover, recent questions have arisen more broadly about how memory should be studied in general, and whether current behavioral testing paradigms actually measure the most meaningful aspects of learning and memory. ,
The verbal memory task used varied from simple paired associative learning task or short word list encoding , to encoding a list of 16 items or words. , The interference procedure varied from 10 seconds to 10 minutes. The recollection period was free recall , , or a recognition task. , , Some studies included a nonverbal memory paradigm. , Testing must be precisely time-locked to the time of the stimulation and most of the studies used a computer-controlled task. The stimulation used was single pulse or a successive train of pulses. , , The majority used AD subthreshold. Verbal memory tasks consist of 3 phases: encoding phase, distraction phase and recollection phase. Several studies have compared stimulation during the encoding phase and during the recollection phase. According to the results of these studies, the best time to induce a memory deficit is by stimulation in the encoding phase. ,
Recently a study addressed this specific question of the most effective time to stimulate during a verbal memory task. They applied stimulation of the medial temporal lobe during various phases of the task: encoding, distractor, interval, and recall. They found that the disruptive effect of the stimulation of the dominant mesial temporal lobe was timing dependent. Stimulation during the interval between learning and recall most strongly disrupted recall.
As with stimulation of the language functions, stimulation must be of sufficient intensity to exert a disruptive effect on the memory network. Following initial studies which raised the possibility that eliciting an AD might specifically cause a memory deficit, subsequent studies have aimed to avoid such a global effect, by using subthreshold stimulation.
Several studies have shown hemisphere-specific memory deficits induced by stimulation. Left hippocampal stimulation produced word recognition memory deficits, whereas right hippocampal stimulation produced face recognition memory deficits. Conversely, another study suggests that recognition can be processed independently by the hippocampus of either hemisphere. Using the same single pulse stimulation parameters authors described an induction of memory deficits by bilateral hippocampi stimulation only whereas unilateral stimulation did not. Another study showed that stimulation of “pathological” hippocampus, (i.e., included in the epileptogenic zone) failed to induce a verbal memory deficit whereas electrical stimulation of the left hippocampus outside the EZ (right temporal lobe seizure onset) interferes with verbal learning performance.
The paucity of studies of memory mapping reflects the difficulty of functional mapping within the region. Despite the great clinical interest of these procedures, they are rarely used in clinical practice. We routinely perform stimulation of the hippocampus during an associative memory task. However, the results are often difficult to interpret in terms of significance at the individual level, particularly given the range of baseline memory function in this population.
The question of the intensity of the stimulation has also not been resolved. If we consider other studies describing an enhancement of memory function during hippocampal stimulation, we have a long way to go in our understanding of how best to explore memory functions though stimulation. Further studies are needed to reproduce and confirm these results.
Most existing memory stimulation studies were experimental in nature, and did not use the data to predict memory outcome. We recently proposed an “electric Wada,” using stimulation from a depth electrode implanted in the left hippocampus in an attempt to determine if a patient could potentially undergo stereotactic laser amygdalohippocampotomy (SLAH) despite failing an intracarotid amobarbital procedure (IAP). The reasoning was that the stimulation of the hippocampus would more accurately simulate the procedure rather than an IAP which would affect the anterior temporal lobe region and not necessarily the entire hippocampus. Stimulation was applied at various intensities and durations while the patient underwent a simulated Wada procedure and a variety of working memory and episodic memory tasks. The patient initially underwent radiofrequency (RF) ablation using the single depth. As there was a brief period of seizure freedom and no apparent cognitive decline on testing, the patient eventually underwent left SLAH, achieved seizure freedom at 1 year, and had a generally positive cognitive and function outcome with no memory decline. Much more work is needed to establish the safety and methodological parameters of such paradigms. It would have been ideal to have stimulated each hippocampus independently, and to sample broader TL regions as well. At present, with stimulation from SEEG, we will not always have the spatial coverage that we may need, and placing extra electrodes would not be without risk to the patient. There are some promising noninvasive techniques that may allow us to conduct such studies in the near future. ,
Other functions
Selimbeyoglu and Parivizi have nicely summarized results based on electrocorticography (ECoG) and SEEG stimulation. Here, we report the main findings published and observed routinely during SEEG.
Vestibular Symptoms
Two studies have well described vestibular symptoms induced by stimulation of parietal areas. The study by Kahane detailed precisely all the experiences reported. Patients experienced illusions of rotation, translations, or indefinable feelings of body motion. In this study, authors identified a lateral cortical temporo-parietal area called the temporo–peri-sylvian vestibular cortex, from which vestibular symptoms, and above all rotatory sensations, were particularly easily elicited. This area extended above and below the sylvian fissure, mainly inside Brodmann areas 40, 21, and 22. It included the parietal operculum which was particularly sensitive for eliciting pitch plane illusions, and the mid and posterior part of the first and second temporal gyri which preferentially caused yaw plane illusions. In a more recent study, Balestrini and colleagues reviewed stimulation effects in the parietal lobe. The occurrence of vertigo was significantly associated with high frequency electrical stimulation (50 vs. 1Hz) of the posterior cingulum.
Face Recognition
Stimulation of the right inferior occipital gyrus has induced a transient selective impairment in face recognition (prosopagnosia). , The region corresponds to the right occipital face area well defined by fMRI. Additionally, stimulation of the ventral visual processing stream of the right TL region has also been associated with deficits in familiar face recognition in the setting of both tumor surgery and epilepsy.
Visual-spatial processing, construction, navigation
There has been little systematic effort to study these constructs, although the occasional case study or experiment has been completed. Nevertheless, deficits in some of these functions can lead to varying degrees of disability for a given patient. Unilateral spatial neglect tends to occur more often following right-sided lesions, and this function can be mapped with line bisection and cancellation tasks. Limited data have related these functions of spatial neglect and discrimination to the inferior and superior parietal lobules and also portions of the posteriori temporal lobe. Mental rotation tasks have also been used to assess the nondominant parietal lobe in the setting of stimulation mapping. , Spatial judgment tasks and basic face processing functions have been associated with nondominant parietal lobe stimulation (specifically the parietal-occipital junction) and the posteroinferior frontal lobe. Navigation tasks have also been used in humans, at least for research purposes, but have not routinely made it into clinical use. Navigation paradigms have been used extensively in rodents, which has contributed to a discovery of hippocampal place cells, yet there appear to be substantial differences between the rodent and human in this regard.
Executive control processes
Monitoring of executive functions is particularly important during frontal resections, but it is important to note that executive functions involve more distributed cortico-cortico and cortico-subcortical networks, and deficits in this domain can develop with damage outside the frontal lobes. There has not been a great deal of research completed in this area with stimulation mapping, but there are many tasks which could be explored. For example, Puglisi et al. (2018) employed a Stroop paradigm during stimulation mapping, and noted that sparing the identified subcortical sites in the nondominant frontal lobe region led to preserved executive functions as compared to those who did not receive this evaluation.
Socio-emotional function
Within the limbic system, the amygdala has been the subject of a large number of studies about its role in emotions. Interestingly, electrical stimulations of right or left amygdala did not induce the same valence in the emotional effect: stimulation of the right amygdala induced negative emotions, especially fear and sadness, whereas stimulations of the left amygdala were able to induce either pleasant (happiness) or unpleasant (fear, anxiety, sadness) emotions. In this study only high frequency stimulation has been reported, and the question of task is not clearly relevant. Well connected to amygdala, stimulation of the insula can induce an anxiety attack.
There are clearly a number of structures implicated in socio-emotional processing, many of which are often involved in the onset zones of various seizure types. These include the amygdalohippocampal complex, the insula, the cingulate cortex, the anterior temporal lobe (e.g., superior temporal pole), select regions of the broader temporal lobe (e.g., the right temporo-parietal junction), select regions of the broader temporal lobe (e.g., the right temporo-parietal junction), the right dorso-medial prefrontal cortex, the bilateral inferior parietal lobules, and the “default mode network.” We have argued that socio-emotional deficits are more commonplace post-surgical deficits than ever recognized, and have suggested a range of tasks that could be studied as part of a presurgical evaluation. These include functions such as recognizing emotion from facial expression or tone of voice, experiencing emotional valence associated with daily events, aspects of “theory of mind,” alterations in level of emotional experience or emotional control, alterations in habits/preferences driven by changes in emotional experience, and secondary effects upon memory (e.g., lack of emotional valence may alter what one attends to in the environment in a detrimental manner, such as diminished fear learning). All of these tasks can be adapted to the SEEG mapping setting, and some work has been completed in this area over the years. As an example, a number of studies have demonstrated that the motor component of an emotional response (e.g., crying, laughing) can be elicited by electrical stimulation without creating the associated emotional valence. , Our own group found that we could evoke positive affect and anxiolysis from the stimulation of the left dorsal anterior cingulum bundle in four patients, and were able to use this technique to conduct intraoperative cognitive mapping without the sedation of anesthesia. In some cases, inclusion of electrode sites in a surgical procedure which were involved in emotional circuity have resulted in post-surgical deficits. For example, Marincovik et al. (2000) found that destruction of cortical areas that proved sensitive to faces led to post-surgical deficits in emotional face recognition (especially fear). Overall, mapping socio-emotional function may lead to improved functional outcome, but remains in the early stages of development.
Motor Behaviors
In the central region, shock (pulse) stimulations or a train of short duration is used to induce focal clonic movement. Negative motor responses are seen in the pre-SMA region during testing, whereas SMA proper stimulation induced more elementary positive motor behavior such tonic posturing and or palilalia.
In a recent study, Caruana et al. report a clear functional difference between the subparts of the cingulate gyrus. The pregenual part of the cingulate cortex hosted the majority of emotional, interoceptive and autonomic responses. The anterior midcingulate sector controlled the majority of complex motor behaviors along a ventro-dorsal axis: the whole-body behaviors directed to the extra-personal space were elicited ventrally, close to the corpus callosum, hand actions in the peri-personal space were evoked by the stimulation of the intermediate area, and body-directed actions were induced by the stimulation of the dorsal area of the cingulate sulcus. The caudal part of the midcingulate cortex and the posterior cingulate cortex are mainly devoted to sensory modalities. In addition, the caudal part of the midcingulate cortex hosts the majority of vestibular responses, while the posterior cingulate cortex was the principal recipient of visual effects.
Sensorimotor System
Somato-sensory sensations and pain
Somato-sensory sensations like anesthesia, paresthesia or thermic dispersion, have been reported mainly during stimulation of the parietal lobe with low and high frequency stimulation. , The main effect was found in the post central gyrus, but all other parts of the parietal lobe such as the precuneus and posterior cingulum were also involved in the somato-sensory system. Insular stimulation can induce paresthesias and localized warm sensations. Somato-sensory symptoms have been also reported during stimulation of the premotor frontal region especially in the posterior part of the cingulate motor area.
Insular stimulations are well known to elicit painful sensations with high frequency stimulation. Studies suggest that pain is induced preferentially by stimulation of the posterior two thirds of the insula. Painful and nonpainful somesthetic representation in the human insula overlap and both types of responses showed a trend toward a somatotopic organization. Painful sensation has been more frequently described in the nondominant hemisphere.
Autonomic Nervous System
Digestive sensations
Mulak et al. reported high frequency stimulations eliciting digestive sensation. The temporal pole (BA 38), hippocampus, amygdala and anterior cingulate cortex (ACC; BA 24/BA 32) are the typical anatomical locations connected with epigastric sensations. Retrosternal sensations are preferentially related to the anterior cingulate cortex, while oro-pharyngeal sensations are mostly related to the suprasylvian opercular cortex and the insula. Authors found a great variability of induced digestive and associated symptoms corresponding to a widely distributed region.
Cardiorespiratory response
Recently, we have gained a deeper understanding of autonomic semiology of seizures using stimulation of limbic system. First stimulation of subcallosal cortex (BA 25) induced a significant systolic hypotensive change because of a reduction in sympathetic drive and a probable reduction in cardiac output rather than bradycardia or peripheral vasodilation–induced hypotension. This effect was not observed after stimulation of the amygdala, hippocampal, insular, orbitofrontal, or temporal cortices.
Secondly, right and left insular stimulations can induce bradycardia or tachycardia. Tachycardia was accompanied by an increase in LF/HF ratio, suggesting an increase in sympathetic tone; while bradycardia tended to be accompanied by an increase in parasympathetic tone reflected by an increase in HF. Authors found mild left/right asymmetry in insular subregions where increased or decreased heart rates were produced after stimulation. Spatial distribution of tachycardia responses predominated in the posterior insula, whereas bradycardia sites were more anterior in the median part of the insula.
Third, stimulation of the amygdala induced a central apnea and oxygen desaturation. This was most frequently observed following stimulation of the medial-most amygdalar contacts located in the central nucleus. These studies highlight the role of the amygdala in voluntary respiratory control.
Visual
One well conducted study reports many visual phenomena evoked electrical stimulations performed with depth electrodes implanted in the occipito-parieto-temporal cortex of 22 epileptic patients. Eighty-five percent of the elementary visual hallucination (spot or a blob) and intermediary hallucinations such as geometric forms (e.g., kaleidoscope, square, triangle, star, diamond) were induced by stimulating the calcarine sulcus, the lingual gyrus, the lateral occipital cortex, the fusiform gyrus and the cuneus/parieto-occipital sulcus. Intermediary hallucinations were evoked at more anterior sites in infra-calcarine occipital structures and in the ventral temporal region than were elementary hallucinations. Most elementary and intermediary hallucinations were colored and moving. Elementary and intermediary hallucinations were all located in the contralateral visual hemifield of the stimulation side, except those elicited by stimulations of the calcarine sulcus, which were in the center of the visual field.
Complex hallucinations defined as meaningful visual hallucinations (e.g., face, landscape, animal, body parts) are less frequently reported, and were induced in more associative cortices. For instance, hallucinations of faces were induced by stimulating the cuneus/parieto-occipital sulcus, the precuneus, and the fusiform gyrus. Visual illusions such as a change of color or spatial modification of the visual background were all elicited by stimulating regions of the right cerebral hemisphere, including the rhinal cortex, parahippocampal gyrus, collateral sulcus, and fusiform gyrus. With the exception of the most posterior cortical sites, the probability of evoking a visual phenomenon was significantly higher in the right than the left hemisphere. Visual perceptive impairments (face recognition) were observed only in two patients. Prosopagnosia was evoked by the stimulation of the right inferior occipital gyrus in one patient and by the stimulation of the right middle fusiform gyrus in the second patient.
The authors observed that the probability of evoking a visual phenomenon decreased substantially from the occipital pole to the most anterior sites of the temporal lobe, and this decrease was more pronounced in the left hemisphere. They hypothesize that greater sensitivity of the right occipito-parieto-temporal regions to intracerebral electrical stimulation to evoke visual phenomena supports the idea of a predominant role of right hemispheric visual areas from perception to recognition of visual forms, regardless of visuospatial and attentional factors.
Future Directions
As noted throughout, future directions in stimulation mapping with SEEG should involve the continued development of noninvasive methods for mapping that can supplement the SEEG implant scheme, , the development of methodology to assess memory at the individual subject level, continued development of both active and passive mapping paradigms, and the upgrading of tasks and paradigms to match advances in the theoretical underpinnings of socio-emotional and cognitive processing. The need for more complex test paradigms that mirror the complexity of brain processes and human behavior (e.g., using memory to prospectively plan for the future) are being increasingly recognized. These trends should be supported by the ever-developing nature of technology to allow for paradigmatic studies that were not possible during a prior era (e.g., virtual and augmented reality, machine learning approaches to big data, use of videography/graphic arts).
Summary
Functional mapping in SEEG is performed to determine whether or not there are overlapping spatio-temporal dynamics between the epileptogenic network and functional networks. Functional mapping procedures cannot be dissociated from the anatomo-electro-clinical correlations seen while carefully analyzing the recorded seizures. The observation of disruptions within functional networks induced by electrical cortical stimulation permits the localization of specific functional sub-units involved in such networks. Functional deficits can be observed during or shortly after seizures. The analysis of these electro-clinical correlations can provide additional information about the functional organization in particular of language and memory. In our practice we combine this electro-clinical data obtained from recorded seizures and stimulation studies with information gathered through different tools such as fMRI, and high gamma activity to support the clinical decision process.
The drive to treat epilepsy through the use of surgery while preserving cognitive function has contributed a substantial wealth of knowledge in our understanding of cognitive function. Stimulation contributes to better understanding of brain functioning in general, in neuroscientific studies, but also on an individual level to enhance the exploration of our patient. What we have learned about brain function representation over the last few years is impressive. However, we have to admit that we have a long way to go in our understanding of the mapping of brain functions. At the bedside, it is sometimes difficult to be confident about the contribution of the area explored to the functional network. Thanks to the increasing body of neuroscience data collected, in future years, the way in which we approach functional mapping may well be very different.
Table 10.1 summarizes the positive stimulation sites according to the cognitive function.
