Computational protocol: Dissociable endogenous and exogenous attention in disorders of consciousness☆

Similar protocols

Protocol publication

[…] During the experiment, 128-channel high-density EEG data in microvolts (μV), sampled at 250 Hz and referenced to the vertex, were collected using the Net Amps 300 amplifier (Electrical Geodesics Inc., Oregon, USA). Data from 91 channels over the scalp surface (at locations shown in B, top) were retained for further analysis. Channels on the neck, cheeks and forehead, which mostly contributed more movement-related noise than signal in patients, were excluded. The retained continuous data were low-pass filtered at 20 Hz, high-pass filtered at 0.5 Hz, and epoched between − 300 and 800 ms relative to the start of the presentation of each word. The epochs generated were baseline-corrected relative to the mean activity during the − 300−0 ms window.Data containing excessive eye movement or muscular artefact were rejected by a quasi-automated procedure: noisy channels and epochs were identified by calculating their normalised variance and then manually rejected or retained by visual confirmation. Independent Components Analysis (ICA) based on the Infomax ICA algorithm () was used to visually identify and reject noisy components. Finally, previously rejected channels were interpolated using spherical spline interpolation, and data were re-referenced to the average of all channels. These processing steps were implemented using custom MATLAB scripts based on EEGLAB (). The number of channels interpolated, epochs and ICA components rejected in healthy volunteer and patient datasets discussed in the section are listed in . Also specified therein are the numbers of explicit target, implicit target and distractor trials available for the statistical analysis procedure described next.Data containing excessive eye movement or muscular artefact were rejected by a quasi-automated procedure: noisy channels and epochs were identified by calculating their normalised variance and then manually rejected or retained by visual confirmation. Independent Components Analysis (ICA) based on the Infomax ICA algorithm () was used to visually identify and reject noisy components. Finally, previously rejected channels were interpolated using spherical spline interpolation, and data were re-referenced to the average of all channels. These processing steps were implemented using custom MATLAB scripts based on EEGLAB (). The number of channels interpolated, epochs and ICA components rejected in healthy volunteer and patient datasets discussed in the section are listed in Inline Supplementary Table S2. Also specified therein are the numbers of explicit target, implicit target and distractor trials available for the statistical analysis procedure described next. Inline Supplementary Table S2 Inline Supplementary Table S2 can be found online at http://dx.doi.org/10.1016/j.nicl.2013.10.008.Epochs from an experimental condition and its own baseline period, or pairs of conditions of interest, were compared using a non-parametric t-test based on that employed in the FieldTrip toolbox (). This test identified temporal clusters of statistically significant differences between the Global Field Power (GFP) () of the ERPs in the two conditions using a Monte Carlo procedure for estimating p-values. To elaborate, we first calculated ERPs by separately averaging epochs (for single-subject analysis) or subject-wise averages (for group analysis) included in each condition. The difference between the GFP time courses of the two ERPs was then tested for statistical significance using a randomisation testing procedure. To do this, the original epochs/subject-wise averages were mixed together and separated into two new sets that contained random samples from the original conditions. These sets were again separately averaged to calculate new ERPs and GFP difference time course. This randomised resampling step was repeated 1000 times, to generate as many GFP difference time courses. The original GFP difference at each time point within a time window of interest was then compared to the maximum GFP differences obtained within that time window over the randomisation iterations, to calculate a time point-wise t-value and p-value. Significant time points with p-values < 0.05 were clustered together based on temporal contiguity, and the cluster with the largest sum of constituent t-values, the cluster-level t-value, was retained. This procedure was then repeated for the GFP difference generated in each randomisation iteration, to identify the largest such cluster generated in each iteration. Finally, the cluster-level t-value generated with the original GFP difference was compared to the distribution of cluster-level t-values generated by the randomisation iterations, to calculate a non-parametric p-value. This represented the Monte Carlo estimate of the level of statistical significance of the cluster identified in the original GFP. As shown by and , this comparison of the original GFP difference at each time point to the maximal GFP difference obtained in each iteration, followed by temporal clustering of time points, effectively and sensitively controls for familywise error (FWE) and multiple comparisons. Cluster-level t-values and p-values calculated as above are reported in the text and figures.In addition, we tested for the statistical consistency of topographical structure within a time window of interest across the individual epochs/subject-wise averages comprising an ERP, using the Topographic Consistency Test (TCT) (). This test employed a non-parametric, GFP-based approach to estimate the significance of a single ERP topography, complementary to the clustering analysis described above. Briefly, the GFP of an ERP at each time point within a time window of interest was compared to the distribution of GFPs at the same time point, calculated over 1000 randomisation iterations in each of which the scalp topography of individual epochs was repeatedly randomised. This generated a Monte Carlo p-value that represented the probability with which the original GFP topography could have been generated just by chance (see for details). These p-values generated by the TCT are reported in the figures, alongside results from the clustering analysis. […]

Pipeline specifications

Software tools EEGLAB, FieldTrip
Application Clinical electrophysiology
Organisms Homo sapiens