ILCB Project – The role of the hippocampus in the production of words
Authors: Catherine Liegeois-Chauvel & François-Xavier Alario (LPC)
Experimental support for data analysis: Valérie Chanoine ( CREx, ILCB)
Objective: The study is based on intracranial EEG (iEEG) recording and the objective is to compare the neural bases of memory encoding and retrieval, as well as language production. The role of the hippocampus in language production is particularly focused.
Preprocessing
Preprocessing was performed using Brainstorm software (https://neuroimage.usc.edu/brainstorm/Introduction). The rejection of bad channels was performed visually based on the power spectrum density analysis (Welch).
Epochs were computed off-line between -1.5 and 1.5 s surrounding trials with a correction baseline of 500 ms (-0.5 to 0 s).
5 experimental conditions were defined for epoching:
- Picture Naming (3 blocs called PictureNaming_B1 to B3)
- Target Encoding
- Target Recognition
- Distract Recognition
- Black and White Decision
Trials were visually inspected for artifacts (e.g., epileptiform spikes), and trials containing artifacts were excluded from analysis. Additional trials were also rejected based on errors (false alarms and omissions) from behavioural responses.
Analysis. Local Field Potentials (LFPs)
Cleaned trials per patient and per condition were exported in matlab structures for other analyses. There were downsampled at a lower sampling rate from 1000 to 128 Hz (function resample from matlab) to reduce calculation time of statistical calculation.
For statistical analysis, trials were averaged par contact, per condition and per participant and centered around zero (demeaning method) to obtain a local field potential (LFP).
The contacts of 5 regions of interest (ROIs in left and right side of hemispheres) retained focus.
- anterior Hippocampus (aHippo)
- posterior Hippocampus (pHippo)
- parahippocampal gyrus (PHG)
- temporal Pole (TPole)
- Fusiform (Fusiform)
- Rhinal region (Rhinal)
On the basis of anatomical images of the electrode implantation (CT scan), contacts outside the gray matter were excluded. Each ROI included spatially close individual contacts from several patients (a minimum of six contacts and 2 patients was applied as inclusion criteria). Individual contacts of each ROI are shown in the Table 1.
Table 1. Distribution of contacts per region of interest and per patient
aHippo_L | pHippo_L | PHG_L | TPole_L | Rhinal_L | Fusiform_L | aHippo_R | pHippo_R | TPole_R | |
Pt01 | B’3_6 | C’1_5 | E’1_5 | I’11_13 | |||||
Pt02 | B’1_5 | I’1_7 | E’1_3 | B1_5 | I1_4 | ||||
Pt03 | B2_4 | C4_6 | I1_8 | ||||||
Pt04 | B’1_3 | I’1_8 | E’1_2 | F’1_2 | |||||
Pt05 | B1_4 | C2_6 | |||||||
Pt06 | I’1_4 | B1_4 | I1_5 | ||||||
Pt07 | B’1_5 | C’1_4 | I’1_5 | F’3_6 | |||||
Pt08 | B’2_6 | C’1_5 | I’3_8 | F’3_4 | |||||
Pt09 | B’1_4 | C’3_6 | B1_5 | ||||||
Pt10 | B’1_3 | C’1_2 | F’2_3 | ||||||
Pt11 | B’1_5 | C’6_7 | C’2_5 | I’2_9 | E’1_3 | F’6_8 | |||
Pt12 | B’1_4 | E’1_3 | I’1_5; 7_8 | ||||||
Pt13 | B’1_4 | C’1_3 | E’1_3 | ||||||
Pt14 | B’1_4 | ||||||||
Pt15 | B’1_5 | C’1_5 | E’1_2 | F’4_8 |
Analyses. Data were imported with homemade scripts in MATLAB (MathWorks Corp., Natick, MA, USA) for statistical analyses and data visualization. The statistical significance of the differences between two conditions was evaluated on LPFs (over the time period [-0.5 1.5s] ) by a cluster-based permutation approach (two-sided T-Statistic; test alpha level of .025; 1,000 random permutations; see Maris & Oostenveld, 2007).
The principle of this approach is to first perform a statistical t-test between the 2 experimental conditions on each sample pair (channel -time) to prepare the selection of clusters. All samples are selected whose t-value is larger than some threshold (here, for a two-sided t-test, the t-values are thresholded at the 2.5-th and the 97.5-th quantiles). Selected samples are then clustered in connected sets on the basis of temporal and spatial adjacency. Cluster-level statistics with 1000 repetitions are calculated by taking the sum of the t-values within every cluster and the maximum of the cluster-level statistics was taken.
We used this statistical approach mainly because of its elegant handling of the multiple-comparisons problem on spatial and time domain. In the case of iEEG, spatial adjacency was defined in terms of neighbourhood between contacts (here, 1 contact is defined as a neighbour when it belongs to the same electrode and is the closest contact).
We have distinguished 2 types of analysis: individual-level and group-level analyses.
Individual-level ROI analysis.
At the individual-level, independent samples (time points) were based on trials. The permutation tests statistic between 2 experimental conditions were computed for each contact pair (contact-sample time) or for each ROI pair (ROI-time points).
In the case of individual-level ROI analysis, the trials (LFPs) from the contacts belonging to the same ROI were considered per contact (matrices [nContacts x nTimes] per subject, per ROI and per condition) or averaged across contacts (matrices [1 x nTimes] per subject and par ROI). In the latter case, there was no sense in clustering across space (loss of the spatial adjacency). The Figure 1 illustrated the result for a given comparison of 2 experimental tasks (here, Target-Recognition versus Distract-Recognition), a given subject (here, Patient Pt 01) and a given ROI (here, left anterior Hippocampal Region).
Figure 1. ROI Results on individual LFPs. Comparison between Target-Recognition and Distract-Recognition for the patient Pt 01 with ROI= left anterior Hippocampal region. On the right side, the LFPs timeseries are presented par contact belonging to this ROI and the statistical tests (cluster-based permutation tests) between the 2 conditions of Recognition are illustrated by p-Values corrected on the spatial and time dimensions at each timepoint. The colobar actually corresponds to -log(pValue), which means that the significance threshold is greater than or equal to about 1.3. In other words, when the black color represents non-significant values, the gray value represents trending values and all other colors, from blue to yellow, indicate gradual significant values from
minus to plus.
Sanity Check
Some patients did not complete the memory and language tasks on the same day. To address a possible session problem, we compared the categorization tasks between the language and the memory session. With this in mind, we excluded patients who showed a significant difference between the memory and the language categorization tasks for a given ROI and for a critical window time of [0-0.8s].
Table 2. Distribution of contacts per region of interest and per patient after sanity check
aHippo_L | pHippo_L | PHG_L | TPole_L | Rhinal_L | Fusiform_L | aHippo_R | pHippo_R | TPole_R | |
Pt01 | B’3_6 | C’1_5 | E’1_5 | I’11_13 | |||||
Pt02 | I’1_7 | B1_5 | I1_4 | ||||||
Pt03 | B2_4 | C4_6 | I1_8 | ||||||
Pt04 | B’1_3 | I’1_8 | E’1_2 | F’1_2 | |||||
Pt05 | C2_6 | ||||||||
Pt06 | I’1_4 | B1_4 | I1_5 | ||||||
Pt07 | B’1_5 | C’1_4 | I’1_5 | ||||||
Pt08 | B’2_6 | C’1_5 | I’3_8 | F’3_4 | |||||
Pt09 | B’1_4 | C’3_6 | B1_5 | ||||||
Pt11 | B’1_5 | C’6_7 | C’2_5 | I’2_9 | E’1_3 | ||||
Pt12 | B’1_4 | E’1_3 | I’1_5; 7_8 | ||||||
Pt13 | B’1_4 | E’1_3 | |||||||
Pt14 | B’1_4 | ||||||||
Pt15 | B’1_5 | C’1_5 | E’1_2 | F’4_8 |
Group-level ROI visualization.
In group-level ROI visualization, the trials from each contact belonging to the same ROI were averaged per condition for each subject and presented on the same figure (see Figure 2).
page presented by Valérie Chanoine – 11th, March 2021