Introduction

About one fifth of patients with epilepsy suffer from medically intractable seizures [29]. These patients often have partial or localization-related epilepsy [14]. A substantial proportion of them may benefit from epilepsy surgery. While many patients with a discrete brain lesion or mesial temporal sclerosis can undergo surgery after non-invasive investigations, other patients need intracranial EEG monitoring to identify the epileptogenic zone and possible eloquent areas that should be spared from resection. Intracranial EEG monitoring can be performed with depth electrodes and subdural electrodes, either alone or combined together. The former allows to explore sulci and internal portions of brain lesions [8], while the latter allows functional mapping and a larger coverage of neocortical structures [24].

Although 25–50 % of patients evaluated for epilepsy surgery undergo invasive EEG monitoring [34], relatively few data are available regarding technique, complications and outcome [1, 6, 9, 15, 16, 18, 22, 23, 25, 26, 33, 43, 45]. The present paper focuses on a series of 58 consecutive patients who underwent epilepsy surgery after EEG monitoring with subdural electrodes and were followed up for at least 2 years.

Methods

Participants

All patients (N = 58) with drug-resistant focal epilepsy who underwent resective surgery after intracranial monitoring with subdural electrodes and were followed up for at least 2 years at the Epilepsy Surgery Centre of the Neuromed-IRCCS Institute between 2003 and 2010 were included in the study. Their mean age was 30.4 ± 3.5 years (range 8–60 years), and 25 (43 %) were female.

Presurgical protocol

All patients underwent a non-invasive diagnostic protocol described in detail elsewhere [28] that included a detailed seizure history, neurological examination, brain MRI scan (1.5 or 3 T, general electric system), neuropsychological assessment, psychiatric evaluation, and video–EEG monitoring. Before electrode implantation, 3-T MRI volumetric sequences (T1-weighted FSPGR, slice thickness 1 mm) were also obtained before and after contrast administration. Neurological examination was abnormal in 17 cases (29 %). Eight patients (14 %) had mental retardation as measured by the Weschler Adult Intelligence Scale-R. In 44 patients (76 %), a structural lesion was preoperatively detected (focal cortical dysplasia in 14 cases, gliosis in 13 cases, low-grade tumor in 7 cases, mesial temporal sclerosis in 2 cases, cortical atrophy in 3 cases, cavernous hemangioma in 2 cases, and other lesions in 3 cases).

Electrode implantation

Subdural electrodes (assembled in grids or strips) were implanted after craniotomy under general anesthesia to further delineate the epileptogenic zone and to perform cortical mapping of eloquent areas if needed. In each patient, the intracranial investigation strategy and electrode positioning were planned on the basis of the “suspected” epileptogenic zone, as defined by non-invasive presurgical evaluation discussed in a multidisciplinary case conference (Fig. 1). Depending on the preoperative planning, a variable number of flexible electrodes (with variable number of electrode contacts) were advanced subdurally in order to cover the exposed cortex. Overall, 59 grids and 122 strips were implanted, with a total of 3,343 and 891 contacts, respectively. Table 1 provides a detailed description of coverage and grid/strip size in each patient. To address the electrode placement over the desired cortical areas, three-dimensional MRI reconstructions were used in each patient (Fig. 2). In one patient, three depth electrodes with ten contacts each were also stereotactically implanted to explore the medial frontal lobe. Avascular trajectories were planned on the basis of preoperative MRI. After positioning of a Radionics CRW stereotactic frame, CT scans were obtained and fused with preoperative MRI to obtain the stereotactic coordinates for each trajectory. Using a tool holder, the depth electrodes were subsequently implanted through a 1-cm burr hole. Before and after subdural electrode positioning, a digital photograph of the operatory field was taken. In some selected cases, contacts were then detected on post-implantation brain CT scans through a semiautomated procedure based on a MATLAB routine described in detail elsewhere [32]. Extraoperative video–ECoG recordings were performed in each patient’s room, with a 128-channel Grass-Telefactor system with a sampling rate of 256 Hz. Bipolar and monopolar recordings of basic rhythms as well as interictal and ictal activity were analyzed. In some cases, adjunctive scalp electrodes were employed for sampling those brain regions that were not explored by intracranial electrodes. In all cases, video–ECoG monitoring was carried on until a number of seizures sufficient to allow the identification of the seizure onset zone was recorded. The mean duration of invasive monitoring was 2.3 days (range 1–14).

Fig. 1
figure 1

Example of planning of subdural electrodes implantation: mesial and dorsolateral frontal structures as well as mesial temporal structures constitute the target regions of this exploration

Full size image
Table 1 Coverage and grid/strip size
Full size table
Fig. 2
figure 2

Three-dimensional preoperative MRI reconstruction are used to address the electrodes placement over the desired cortical areas

Full size image

Brain mapping

In 15 patients in whom the epileptogenic zone was suspected to be close to or overlapping with eloquent area such as primary motor and language areas, functional mapping was also performed. Bipolar electrical stimulation was delivered on the explored cortex through the same subdural contacts used for invasive recordings. To exactly define the boundaries of functional areas, pairs of electrode contacts were progressively tested. Trains of 0.3 ms single pulses (interstimulus interval 16 ms) were delivered at 50 Hz. Intensities and train duration were different among cortical areas, ranging from 0.2 to 0.5 mA and 1 s duration for motor cortex to 5–15 mA and 2–8 s duration for language areas. In all cases, extraoperative cortical stimulation was performed in the patient’s room during neurological evaluation, neuropsychological testing, and ECoG monitoring. Neuropsychological assessment for language required from 2 to 3 h and included naming, comprehension, counting and reading tasks, calculation, sentence completion, and writings tasks. When advised, resective strategies were tailored based on the invasive epileptological and functional mapping. Resections were performed under general anesthesia, with microsurgical technique. All patients underwent postoperative MRI. Neurological, EEG, and neuropsychological evaluations were repeated 1, 2, and 5 years after surgery. Seizure outcome was determined yearly according to Engel’s classification [13].

Results

Invasive video–EEG monitoring

Extraoperative ECoG allowed the identification of the epileptogenic focus in all cases. The temporal lobe was involved in 21 patients (36 %), whereas extratemporal foci were identified in 24 patients (41 %) (frontal in 16 cases, parietal in 4 cases, central in 3 cases, insular in 1 case). Thirteen patients (23 %) had multilobar involvement. The dominant hemisphere was involved in 20 patients (34 %). Among the 15 patients who underwent functional mapping, in 13 cases (87 %) the epileptic focus did not involve eloquent cortex. In one patient with a low-grade tumor arising in Broca’s area, electrical cortical stimulation revealed a re-organization of language area in the adjacent cortex. In another patient with non-lesional left temporal epilepsy, the epileptogenic zone partially involved Wernicke’s area; thus, a subtotal resection was planned based on the functional boundaries.

Complications related to subdural electrode positioning were brain edema in one patient and subdural hemorrhage in two patients. None of them experienced persistent neurological deficits. The patient who was implanted with both subdural and depth electrodes experienced no complications.

Resective surgery

Among the 44 patients with evidence of lesion on preoperative MRI, the lesion was removed alone in 12 cases (27 %), while lesionectomy was combined with tailored cortical resection in 32 cases (73 %). Tailored cortical resection was also performed in 14 patients without evidence of lesion on MRI.

Postoperative transient neurological deficits occurred in five cases (dysphasia in three cases, sensorimotor deficits in the remaining two cases). Another patient experienced permanent lateral homonymous hemanopia. Postoperative MRI showed a gross total lesion removal in 41 of the 44 patients with symptomatic epilepsy (93 %). In the remaining three cases, the lesion was only partially removed. Histopathological findings together with preoperative MRI findings are reported in Table 2.

Table 2 Histophatological findings compared to preoperative MRI
Full size table

The mean duration of follow-up was 48 months (range 24–77). At the last follow-up observation, 34 patients (57 %) were seizure-free, whereas only three patients (5 %) did not experience any worthwhile improvement. Seizure outcome according to Engel’s classification is detailed in Table 3.

Table 3 Seizure outcome according to Engel’s classification
Full size table

Discussion

Intracranial electrode implantation represents an important assessment tool in candidates for epilepsy surgery, as non-invasive EEG recordings cannot adequately localize the epileptogenic zone in approximately 25 % of patients with refractory seizures [35]. Subdural electrodes have a relatively wide coverage with fairly uniform spacing and can be implanted over the cortical convexity, the basal neocortex, the mesial neocortex, or the interhemispheric neocortex [21]. On the other hand, depth electrodes may have a better view of some deep-seated dipoles, despite their markedly smaller sampling. Thus, they are used primarily to study deep structures such as the mesial temporal lobe [10, 36, 37, 47]. The decision whether to use subdural electrodes alone, depth electrodes alone, or a combination of both types of electrodes depends mainly on the underlying pathology and suspected epilepsy type. According to current literature [2, 5, 19, 39], we implanted subdural grid electrodes alone in neocortical epilepsy or when the suspected epileptogenic zone was close to eloquent cortex. Combined subdural plus depth electrodes implantation was performed only in one patient, in order to assess the earlier involvement of medial frontal structures in the epileptogenic process as compared with neocortical regions.

Careful mapping of the irritative zone and ictal onset zone followed by complete surgical resection of the mapped epileptogenic zone was found to be associated with relatively good postoperative outcome [1, 17, 38]. In all our patients, the use of intracranial electrodes for extraoperative video–EEG monitoring allowed us to record one or more seizures. Even though intraoperative ECoG has been widely employed in tailoring cortical resections for epilepsy surgery [3, 4], this technique usually does not enable to detect the ictal onset zone because the amount of time for recording is relatively short.

While they allow functional mapping and provide large coverage of neocortical structures, subdural electrodes have also disadvantages, as they may exert a mass effect on the brain and there can be cerebrospinal fluid leakage through the openings in the dura. These openings may also favor the development of infection [20], especially when a high number of contacts are implanted [45]. In our series, no patient experienced permanent neurological sequelae, while three patients (5.2 %) had transient complications not requiring treatment. This relatively low proportion of complications related to electrode implantation is comparable with those reported by most other similar studies [1, 6, 15, 18, 22, 23, 25, 26, 33, 43].

Our seizure outcome was slightly less favorable (57 % in Engel’s class I) as compared with those reported in the literature after standard temporal resections (60–90 % in Engel’s class I), usually without invasive investigations [13]. This is likely to ascribe to the high prevalence (64 %) of extratemporal lobe involvement in our series. In most studies, the percentage of success after extratemporal resections is 32–51 %, with better outcome in patients with detectable lesions [12, 40, 41, 44].

In our series, 12 of the 44 patients with symptomatic epilepsy underwent lesionectomy alone, while in the other patients, lesion removal was combined with tailored cortical resections, since invasive recordings demonstrated that the epileptogenic zone exceeded the boundaries of the epileptogenic lesion. In five of the 14 patients without detectable lesions on preoperative MRI, histophatological examination revealed malformations of cortical development. This finding confirmed the reported association between cortical dysplasia and epilepsy [27, 39].

In patients with epileptic focus suspected to be close to eloquent brain areas, good functional outcome was reported after intraoperative mapping under awake craniotomy [7, 11]. During intraoperative mapping, language and/or motor function is tested intermittently during the actual process of tumor resection under local anesthesia. Although it is widely considered to be the gold standard, particularly for language evaluation [7], awake craniotomy carries several disadvantages, including delayed operative time, patient discomfort, and less accuracy in neuropsychological assessment. In addition, cortical stimulation often elicits epileptiform afterdischarges, which can sometimes progress into seizures. This can delay or prevent functional mapping at some sites, particularly those closest to the seizure focus. In our series, extraoperative cortical stimulation allowed to identify the exact spatial relationship between epileptogenic focus and eloquent cortex in all cases. Also, continuous ECoG monitoring during cortical stimulation was useful in detecting afterdischarges, thus preventing stimulation-related seizures.

The role of fMRI as a possible substitute for neurophysiological methods in testing eloquent areas is still debated. There is a fairly high percentage (9 %) of false positive/negative results regarding language lateralization [46]. Also, the accurate cortical localization of higher functions such as writing or calculation is still a major problem [42]. Therefore, critical surgical decisions should not be based on this technique alone [30, 31].

In conclusion, our findings suggest that invasive EEG recording with subdural electrodes allows the identification of the epileptogenic zone and adequate functional brain mapping in most candidates for epilepsy surgery, with a relatively low incidence of complications.