A perceptual decision making EEG/fMRI data set

We present a neuroimaging data set comprising behavioural, electroencephalographic (EEG), and functional magnetic resonance imaging (fMRI) data that were acquired from human subjects performing a perceptual decision making task. EEG data were acquired both independently and simultaneously with fMRI data. Potential data usages include the validation of biocomputational accounts of human perceptual decision making or the empirical validation of simultaneous EEG/fMRI data processing algorithms. The dataset is available from the Open Science Framework and organized according to the Brain Imaging Data Structure standard.

. Subject data set inventory. Green ticks represent successfully recorded data set components, blue crosses represent absent data set components. Overall, the dataset includes behavioural data acquired simultaneously with EEG data outside of the MR environment from 16 subjects (BEH outside), behavioural data acquired simultaneously with EEG and fMRI data from 16 subjects (BEH inside), EEG data recorded outside the MR environment from 16 subjects (EEG outside), EEG data recorded simultaneously with fMRI data from 17 subjects (EEG inside), and fMRI data from 16 subjects (fMRI).
sign with factors visual stimulus coherence and spatial prioritization (attention) as detailed below ( Figure 1a and b). Subjects performed the task in two separate experiments. Specically, after preparation of the EEG cap, subjects rst performed two runs of the task outside the MR scanner while behavioural and EEG data were recorded. We refer to the data of this experiment as EEG only and outside MRT data. After a short break, subjects were situated in the MR scanner and, simultaneously with the acquisition of behavioural and fMRI data, EEG data was acquired in ve runs of the task. We refer to the data of this experiment as simultaneous EEG/fMRI and inside MRT data. Table 1 provides an overview of the components of the data set that could successfully be recorded from each subject.

Participants
Seventeen participants ( Figure 1. Study design, perceptual decision task, and dataset overview. a. 2×2 factorial experimental design with factors stimulus coherence (low, high) and spatial prioritization (yes, no). On each trial of the experiment, the subject was presented with a face or car stimulus, which had been manipulated according to its spatial phase coherence. Simultaneously, the subject was prompted to either spatially prioritize the stimulus display or not. The stimulus category (face or car) that the subject was asked to discriminate, was manipulated orthogonally to the other factors. b. Single experimental trial. Prior to the presentation of the stimulus, a one-headed arrow could indicate the hemield of the subsequent stimulus presentation for spatial prioritization. Alternatively, a two-headed arrow was uninformative with respect to the location of the upcoming stimulus in the no spatial prioritization condition. The cueing arrow was shown continuously for 1 s pre-stimulus, the stimulus itself was shown for 200 ms. The subject was asked to respond as quickly and as accurately as possible with no restrictions on the response window.
The inter-trial interval was 0300 ms for the EEG only and 10-12 s for the simultaneous EEG/fMRI recordings. c.
Schematic overview of the study and dataset. EEG and simultaneous EEG/fMRI data were acquired over two data acquisition sessions. The resulting imaging and behavioural data were standardized into BIDS format. The gure depicts the folder structure as available from the Open Science Framework. and contrast as assessed by a one-way ANOVA with factor image category and levels face and car (mean driving luminance: F (1,34) = 0.08, p = 0.78, contrast: F (1,34) = 0.23, p = 0.64).

Phase-scrambled stimulus set
To manipulate the informativeness of the images, the images' spatial phase spectra were linearly weighted with a phase spectrum of a uniform noise image using the weighted mean phase technique described in Dakin et al. 22 . With the original phase spectrum of an image given by φ o , the scrambled phase spectrum φ s was computed as φ n is the phase spectrum of a random noise image, created by sampling each pixel's value uniformly from the interval [0, 1] using Matlab's rand.m function, and w ∈ [0, 1] is the signal-to-noise weighting coecient. Stimuli with weighting coecients w 1 = 0.9 (high coherence) and w 2 = 0.5 (low coherence) were chosen for the experiment in order to elicit reliable dierences in the response times for either stimulus class, while still allowing accurate performance of the task. The stimulus set generation is implemented in the Matlab (The MathWorks, NA) function pdm_experiment _stim-ulus_creation.m, which is available with the dataset along with all custom generated programming code.

Experimental procedure
Subjects performed a perceptual decision task in a 2 × 2 factorial within-subject design with experimental factors stimulus coherence (with levels low and high) and spatial prioritization (with level yes and no) (Figure 1a). On each trial, a visual stimulus depicting either a face or a car was presented in one visual hemield with a left/right eccentricity of the stimulus center of 11 degrees of visual angle and a stimulus extension of 9 degrees of visual angle. Individual stimuli were presented for 200 ms and the subject was asked to indicate via a button press whether the stimulus depicted a face or a car. For the button presses, subjects used their right index and middle nger for the two categories, and the mapping from stimulus category to response button was counterbalanced across subjects. As described above, the informativeness of the visual stimulus was manipulated by altering the phase coherence of its spatial frequency spectrum resulting in low and high stimulus coherence trials. On half of the trials, a cueing arrow shown continuously for 1 s prior to the stimulus indicated in which visual hemield the stimulus would be presented ( Figure   1b). Subjects were asked to allocate their spatial attention to the respective visual hemield, while maintaining steady central xation (spatial prioritization condition). On the other half of the trials, the two-headed cuing arrow was uninformative and the stimulus was presented randomly in either visual hemield (no spatial prioritization condition). We refer to the attention factor as spatial prioritization rather than spatial attention, because in the no spatial prioritization condition, spatial attention is equally directed to both potential stimulus displays. This diers from classical spatial attention experiments, in which evoked responses for spatially attended stimuli are typically contrasted with spatially inhibited stimuli 23 . Face and car stimuli were equally distributed across the four experimental conditions. The stimulus presentation order was randomized. Subjects were asked to respond as quickly and as accurately as possible with an emphasis on responding as quickly as possible and to maintain stable xation on the central xation cross throughout the experiment. For the EEG recordings outside of the scanner, data from 72 trials for each of the four conditions (half of them face stimuli) were recorded with an inter-trial interval randomized between 0 ms and 300 ms. Here, the data acquisition was split into two experimental runs of approximately 10 minutes each. For the simultaneous EEG/fMRI recordings data from 90 trials for each of the four conditions (half of them face stimuli) were recorded with an inter-trial interval discretely randomized between 10 and 12 s. This long inter-trial interval was chosen to obtain reliable recordings of single-trial haemodynamic responses.

MRI data acquisition
The simultaneous EEG-fMRI experiment was conducted at the Birmingham University Imaging Centre using a 3T Philips Achieva MRI scanner. An initial T1-weighted anatomical scan (1 mm isotropic voxels) and T2*-weighted functional data were collected with an eight-channel phasedarray SENSE head coil. EPI data (gradient echo-pulse sequence) were acquired from 32 slices (3x3x4 mm resolution, TR 2,000 ms, TE 35 ms, SENSE factor 2, ip angle 80 deg). Slices were oriented parallel to the AC-PC axis of the subject's brain and positioned to cover the entire brain space.

Data Records
The data set is available from the Open Science Framework via the private project https: //osf.io/q4t8k/. It is organized according to the Brain Imaging Data Structure (BIDS) specication 19 . Extensive documentation of this neuroimaging data standard, including its metadata specications for behavioural, EEG, and fMRI data, is available from the BIDS website (http://bids.neuroimaging.io/). Table 2 provides an overview of the data set organization.
We here limit the description of the data records to the top level in the rst column. At this level, the data set contains metadata les, a code/ folder hosting the custom-written Matlab code for the study and the technical validation analyses reported herein, and a stimuli/ folder hosting the visual stimulus set. The behavioural and neuroimaging data itself is organized in a subject-wise manner in the sub<subject id>/ folders, and substructured into further folders as documented in Table 2. Finally, the folder derivatives/ contains EEG data les from inside the scanner that underwent run-wise segmentation and gradient and ballistocardiogram artefact removal. sub-<subject_id>/sourcedata-eeg_inside-MRT/ A folder containing EEG and behavioural data acquired from inside the scanner. eeg/ EEG data les.
derivatives/ A folder containing artefact-corrected EEG les acquired from inside the scanner.

Behavioural data
To validate the behavioural data quality, response times and response accuracy were evaluated for both the EEG only and EEG/fMRI experiment (Figures 2a and b). In both experiments, faster median response times were observed for the high stimulus coherence compared to the low stimulus coherence condition and for the spatial prioritization compared to the no spatial prioritization condition (Figure 2a). Equivalently, response accuracy increased with stimulus coherence and spatial prioritization (Figure 2b)    times across subjects, error bars ± standard error of the mean (SEM). Light grey bars depict behavioural data from the EEG data set recorded outside the MR scanner (EEG), dark grey bars depict behavioural data from the simultaneous EEG/fMRI recordings. b. Response accuracy. Bars depict the response accuracy in percent correct across subjects, error bars ± standard error of the mean (SEM). As for panel a, light grey bars depict behavioural data from the EEG data set recorded outside the MR scanner (EEG), dark grey bars depict behavioural data from the simultaneous EEG/fMRI recordings. c and d. Condition-specic group average ERPs for contralateral trials, pooled over electrodes O2, PO4, PO8 for left hemield trials and O1, PO3, PO7, for right hemield trials for the EEG data recorded outside the MR scanner (c) and simultaneously with the fMRI data (d). e and f. Topography plots of the ERP data averaged over conditions for four selected post-stimulus time points for both EEG data sets. The main positive deections at all time points were observed in the set of parieto-occipital electrodes (O2, O1, PO8, PO7, PO4, PO3) evaluated in the ERP analyses.
Overall, the amplitude variations in the EEG/fMRI data are less pronounced and slightly delayed when compared to the EEG only data. Finally, to asses the topographic expression of the eventrelated responses, the ERP data were averaged over experimental conditions and projected onto a scalp representation at selected time-points using Fieldtrip's topoplot.m function 28 . Here, for both data sets the strongest positive deections for the time points of interest were observed for the set of posterior parieto-occipital electrodes selected for the reported ERPs (Figures 2e and 2f ). In summary, the EEG data from both recordings yield standard ERP waveforms for visuomotor reaction tasks and reproduce known eects of stimulus coherence and spatial prioritization 23,27,29 . The EEG data analyses reported here are implemented in the functions pdm_erp_analysis_eeg_only.m and pdm_erp_analysis_eeg_fmri.m.

fMRI data
To validate the fMRI data quality, a mass-univariate summary-statistics GLM analysis was performed that assessed condition-induced eects at the group-level. SPM12 (V6906) was used for both fMRI data preprocessing and statistical modelling. Prior to GLM parameter estimation at the subject-level, fMRI data were motion-corrected by realigning EPI volumes to the rst volume of the rst run of a given subject, normalized to MNI spaced using the SPM MNI-EPI template, re-interpolated to 2 mm isotropic voxel size, and smoothed using an 8 mm isotropic Gaussian kernel. The rst-level GLM design matrix for each subject was then specied in run-wise, blockdiagonal form. Here, each block comprised the four condition-specic stimulus onset functions, convolved with the canonical haemodynamic response function, in the column-wise order: high stimulus coherence/spatial prioritization, high stimulus coherence/no spatial prioritization, low stimulus coherence/spatial prioritization, low stimulus coherence/no spatial prioritization ( Figure   3a)  Table 3). Testing for higher BOLD activity for the reverse direction of low coherence as compared to high coherence stimuli yielded statistically signicant activation clusters in the right anterior cingulate gyrus and the right inferior frontal gyrus. Activations in the superior parietal lobule, as well as the left and right frontal eye-elds were marginally statistically suggestive (Figure 3d, Table 3). For the directed main eects of spatial prioritization, a cluster in the left superior parietal lobule survived a cluster-dening threshold of p < 0.001. However, this cluster was not signicant at the family-wise error corrected cluster level (Figure 3e, Table 3). Finally, no activity clusters were detected for the reverse main eect of no spatial prioritization as compared to spatial prioritization trials at a cluster dening threshold of p < 0.001. In summary, evaluation of the fMRI data resulted in activations of areas known to be involved in the processing of visual perceptual decisions 6 Table 3. Experimental main eects group-level fMRI results. The table lists a cluster's anatomical label according to the WFU PickAtlas 34 atlas, its center of gravity in MNI coordinates, the T-value of its peak voxel, and the cluster level family-wise corrected p-value based on a cluster-dening threshold of p < 0.001 with an cluster extent threshold of 0 voxels 35 .

Conclusion
In summary, the data set presented here provides a comprehensive representation of the neural processes underlying perceptual decision making as aorded by non-invasive neuroimaging techniques.
Basic technical validation analyses suggest the presence of commonly observed experimental effects of stimulus coherence and spatial prioritization at the behavioural, EEG, and fMRI data level.
The data set may thus be suited for both the development and validation of novel data analytical techniques, as well as for providing new insights into the neural mechanisms of perceptual decision making.

Usage Notes
Because the data were donated by human participants, ethical considerations require some limitations on the access and reuse of the data. To obtain data access, potential data users should contact the corresponding author with a brief statement to which purpose the data will be used.
Upon positive review of their request, data users are required to sign a data use agreement, which is available from https://osf.io/hkevu/. Data users will then be added as collaborator to the Open Science Framework project hosting the data set and can download the data. We kindly request data users to follow the ODC Attribution/Share-Alike Community Norms (https: //opendatacommons.org/norms/odc-by-sa/), and acknowledge Yasmin K. Georgie, Camillo Porcaro, Stephen D. Mayhew, Andrew P. Bagshaw, and Dirk Ostwald in any publication derived from these data, citing this paper and Data Citation 1 as the source of the data.