Participants
Eleven patients with drug-resistant epilepsy participated in this study (1 woman and 10 men, mean age = 36.9; SD = 11.5) during their pre-surgical evaluation with stereotactic electroencephalography (sEEG) recording at the Grenoble-Alpes University Hospital (see Table 1 for patients’ information). All patients gave written informed consent to participate in the present study. The study was approved by the National Ethics Advisory Committee (CPP MAPCOG: N° Id RCB: 2017-A03248-45, local CHUGA number: 38RC17.357). All research was performed in accordance with relevant guidelines/regulations and in accordance with the Declaration of Helsinki.
Table 1: Demographic and clinical information of patients included in the study
P
|
A iEEG
|
G
|
H
|
EZ LAT
|
EZ LOC
|
NCS
|
1
|
54
|
M
|
R
|
L
|
PARIETAL
|
17
|
2
|
28
|
M
|
R
|
L
|
TEMP
|
22
|
3
|
26
|
M
|
R
|
L
|
TEMP
|
49
|
4
|
44
|
F
|
R
|
L
|
TEMP
|
31
|
5
|
37
|
M
|
R
|
L
|
FRONTAL
|
9
|
6
|
25
|
M
|
R
|
L
|
TEMP
|
31
|
7
|
21
|
M
|
R
|
R
|
OCCIPITAL
|
3
|
8
|
31
|
M
|
R
|
L
|
MULT
|
24
|
9
|
46
|
M
|
L
|
L
|
TEMP
|
20
|
10
|
53
|
M
|
R
|
R
|
TEMOP
|
21
|
11
|
41
|
M
|
R
|
L
|
INSULA
|
18
|
Abbreviation P: patient; AiEEG: age at the iEEG acquisition; G: Gender; H: handedness; EZ LAT: epileptic zone lateralization; EZ LOC: epileptic zone location; TEMP: temporal lobe; MULT: multilobe; NCS: number of cortical sites included.
Electrode implantation and positioning
Eleven to 15 semi-rigid, multilead electrodes were stereotactically implanted in each patient. Each electrode had a diameter of 0.8 mm and, depending on the target structure, consisted of 8–18 contact leads 2 mm wide and 1.5 mm apart (DIXI Medical, Besançon, France). Electrode implantation were strictly related to individual clinical hypotheses. A preoperative MRI and post-operative MRI or CT scan were co-registered to assess the locations of the electrode contacts for each patient using a coordinate system in relation to the anterior commissure / posterior commissure plane. Electrode contact positions were finally expressed in the Montreal Neurological Institute (MNI) coordinate system to allow group analyses after brain spatial normalization using Statistical Parametric Mapping 12 software (SPM12, Wellcome Department of Imaging Neuroscience, University College London, www.fil.ion.ucl.ac.uk/spm). Visual inspection of the contact locations was also used to check whether each electrode contact was located in grey or white matter. Seven patients were implanted only in the left hemisphere and 4 had a bilateral implantation. For this study we only included cortical sites recorded in the left hemisphere (see Figure 1).
SEEG recordings
SEEG recordings were performed using a video-EEG monitoring system (Micromed, Treviso, Italy) that allowed to simultaneously record up to 256 monopolar contacts, so that a large range of mesial and cortical areas, as well as fissural cortices, was sampled for each patient. Sampling rate was 1024 Hz, with an acquisition band-pass filter between 0.1 and 200 Hz. Data were acquired using a referential montage with reference electrode chosen in the white matter. All other recording sites were chosen in the grey matter. For data analysis, we used a bipolar montage between adjacent contacts of the same electrode to improve sensitivity to local current generators. Coordinates of virtual bipolar contacts that were used to construct images were chosen to be at an equal distance of two successive contacts.
Language comprehension task and stimuli
The task used in the present study was adapted from Pallier et al. 16 see Figure 2. Stimuli were 240 auditory streams including 80 sentences (i.e., with semantic and syntactic information), 80 jabberworckys (i.e., with syntactic information and without semantic information), 40 words lists (i.e., with semantic information and without syntactic information) and 40 list of pseudowords (i.e., without semantic and without syntactic information). All stimuli were controlled in terms of amplitude. The 80 sentences were taken from the study by Pallier, Devauchelle, and Dehaene 16. From these 80 sentences, we generated the 80 jabberwockys by modifying the content words of the sentences with pseudowords while keeping a similar phonology for each last syllable. We then generated 40 sequences of words and 40 sequences of pseudowords by randomly selecting words and pseudowords from the sentences and jabberwockys. In addition, for each sentence, a comprehension question was created that induced a "yes/no" response.
During the language comprehension task patients were instructed to carefully listen to the auditory stimuli and to answer the question presented at the end of some trials. Typical trial started with a fixation cross during 500ms, followed by an auditory stimulus (a sentence, a jabberwocky, a list of words or a list of pseudowords), and then a grey background (interstimulus interval) for 1500ms. On 25% of sentence trials, following the auditory stimulus, a written verification question was presented on the screen, in order to evaluate if the patients were preforming the task correctly. Participants were instructed to give yes or no responses to the comprehension questions using two buttons. The question remained on the screen until the patient responded and was followed by a 500ms inter-stimulus interval. A total of 240 trials were randomly presented and a pause was proposed every 50 trials.
Preprocessing sEEG signal
All preprocesses and data analyses were performed using FieldTrip toolbox 55 and custom scripts on Matlab ® (version 9.7.0, Matlab R2019b). Raw data were transformed into four time-series corresponding to alpha, beta, low-Gamma and high-Gamma frequency band amplitudes with the following procedure 40,56: Step 1:continuous sEEG signals were first bandpass-filtered in successive 2, 4 or 10 Hz wide frequency bands from 8Hz to 12Hz, 16Hz to 24Hz, 30Hz to 50Hz or 50Hz to 150Hz (for alpha, beta, low-gamma and high-gamma frequency bands respectively) using a zero-phase forward and reverse filter. Step 2: for each bandpass-filtered signal we computed the envelope using standard Hilbert transform 57. Step 3: for each band this envelope signal was divided by its means across the entire recording session and multiplied by 100. This procedure yields instantaneous envelope values expressed in percentage (%) of the mean. Step 4: the envelope signals for each consecutive frequency bands were averaged together to provide one single time series (alpha activity, beta activity, low-gamma activity or high-gamma activity) across the entire session. Step 5: The obtained envelopes were epoched into data segments centered around each stimulus and baseline corrected (the baseline corresponded to the 450ms preceding the beginning of audio presentation).
Regions of interest (ROI)
We computed 14 regions of interest (ROI) allowing to reduce between patient implantation variability. ROIs were defined using Harvard-Oxford atlas 58. The included ROI were: the anterior superior temporal gyrus (Ant STG), the superior posterior temporal gyrus (Post STG), the anterior middle temporal gyrus (Ant MTG), the posterior middle temporal gyrus (Post MTG), the supramarginal gyrus (SMG), the temporal pole, the pars opercularis (IFG Oper), the pars triangularis (IFG Tri), and the pars orbitalis (IFG Orbi). The spatial distribution and the number of cortical sites included in each ROI in relation with the number of patients are shown in Figure 3. See below for the specific computation performed in order to avoid over representation of the activity of one patient who would have more contacts in an ROI
Amplitude Analysis
First, we checked whether all patients performed correctly the task. Average percentage of correct answer to verification questions were well above chance level (M = 83.2%, SD = 15.1%) meaning that patients performed the task correctly. For each trial and in each frequency band, we computed the mean amplitude during the auditory stimuli from 1000ms to 2000ms after the beginning of the presentation. We selected this time windows in order to avoid perception effects after the onset of the audio stream targeting deep stimulus analysis responses. Mean amplitude was then Z-Scored for each channel using the pre-stimulus baseline as reference ([-450ms-0ms]). Z-scores allowed us to average single trials of each channels that were in the same ROI for each patient. This normalization procedure was performed in order to avoid over representation of the activity of one patient who would have more contacts in an ROI. This step was necessary to perform group analysis because patients presented an important variability in implantation anatomical coverage and in number of contact sites per ROI. Consequently, for each ROI, we were able to obtain trial Z-score means for each patient. If one patient presented more than one channel in a specific ROI an average of patient’s channels was also computed.
We compared each language (sentences, jabberwockys, word-lists and pseudoword- lists) condition against the baseline condition using a t-test. Finally, we evaluated Semantic effects by comparing semantic (i.e., sentences and word-lists) to non-semantic conditions to (i.e., jabberwockys and pseudoword-lists) and Syntactic effects by comparing syntactic conditions (i.e., sentences and jabberwockys) to non-syntactic ones (i.e., word and pseudoword lists) using t-tests. The p value was corrected for multiple comparisons using Bonferroni procedure: as we had nine ROIs we divided the critical p value by nine obtaining p = 0.0056.