New insights on the ventral attention network: Active suppression and involuntary recruitment during a bimodal task

Abstract Detection of unexpected, yet relevant events is essential in daily life. fMRI studies have revealed the involvement of the ventral attention network (VAN), including the temporo‐parietal junction (TPJ), in such process. In this MEG study with 34 participants (17 women), we used a bimodal (visual/auditory) attention task to determine the neuronal dynamics associated with suppression of the activity of the VAN during top‐down attention and its recruitment when information from the unattended sensory modality is involuntarily integrated. We observed an anticipatory power increase of alpha/beta oscillations (12–20 Hz, previously associated with functional inhibition) in the VAN following a cue indicating the modality to attend. Stronger VAN power increases were associated with better task performance, suggesting that the VAN suppression prevents shifting attention to distractors. Moreover, the TPJ was synchronized with the frontal eye field in that frequency band, indicating that the dorsal attention network (DAN) might participate in such suppression. Furthermore, we found a 12–20 Hz power decrease and enhanced synchronization, in both the VAN and DAN, when information between sensory modalities was congruent, suggesting an involvement of these networks when attention is involuntarily enhanced due to multisensory integration. Our results show that effective multimodal attentional allocation includes the modulation of the VAN and DAN through upper‐alpha/beta oscillations. Altogether these results indicate that the suppressing role of alpha/beta oscillations might operate beyond sensory regions.

In addition, some studies have focused on the involvement of the VAN in supramodal attention tasks (Macaluso, 2010;Macaluso, Frith, & Driver, 2002), as it can be involuntarily activated by irrelevant stimuli coming from sensory modalities not to be attended (e.g., auditory) but spatially congruent to relevant information (e.g., visual) (Santangelo, Olivetti Belardinelli, Spence, & Macaluso, 2009). In the present study, we wanted to determine whether the VAN is also recruited when information coming from an unattended sensory modality (e.g., visual) is congruent with the attended one (e.g., auditory). This process is different from the traditional reorientation of attention as studied in Posner tasks, as it does not involve a full switch of attention from one location (or modality) to another. It is expected rather to involve an involuntary attentional enhancement for the target due to its increased saliency, triggered by congruency between sensory domains (Gau, Bazin, Trampel, Turner, & Noppeney, 2020). Such enhancement should be expressed behaviorally as an improved performance given by congruency across modalities, instead of a cost given by attentional switch from one to the other. With a bimodal attention task, we hypothesized that the VAN would be recruited in the congruent trials of both attention conditions (visual or auditory), as reflected by a decrease of alpha oscillations in the TPJ (indicating a release from suppression, see e.g., Solis-Vivanco, Rodriguez-Violante, Cervantes-Arriaga, Justo-Guillen, and Ricardo-Garcell (2018)) and/or an increase of gamma or theta oscillations (ElShafei et al., 2018;Proskovec et al., 2018;Sauseng et al., 2005).
Such recruitment in congruent trials was further expected to be related to task performance.

| Subjects
We included 36 healthy subjects attending college who were recruited from Radboud University's research participation scheme.
Inclusion criteria for all participants included Dutch as their mother tongue, right-handedness according to the Edinburgh Handedness Inventory (Oldfield, 1971), normal or corrected-to-normal vision, and reported normal audition. Participants with a psychiatric or neurological diagnosis were excluded. Two participants were excluded due to excessive noise or movement artifacts during MEG recordings. The final sample consisted of 17 females and 17 males, with a mean age of 23 ± 2.5 years. The study was conducted at the Donders Institute for Brain, Cognition and Behaviour and fulfilled the Declaration of Helsinki criteria (WMA, 2013).

| Experimental design
A cross-modal attention task was designed using MATLAB (MathWorks) custom scripts and Psychtoolbox (psychtoolbox.org).
Each trial (5 s duration) began with a black background and a gray central fixation cross that lasted for 1 s and was projected on an acrylic screen by an EIKI LC-XL100L projector with a resolution of 1024 × 768 and a refresh rate of 60 Hz that lasted for 1 s (Figure 1).
Subjects were asked to blink or move their eyes only during this period. Afterwards, the fixation cross turned white and 1,100 ms later an electro-tactile cue (2 ms) was delivered to the left or right thumb.
This cue instructed the participants to allocate attention to the visual (Attend-visual condition; 50% of trials) or auditory (Attend-auditory condition; 50% of trials) stimuli, respectively. The cue was administered with two constant current high voltage stimulators (type DS7A, Digitimer, Hertfordshire, UK; mean current = 3.83 mA). After a postcue interval of 1,150 ms, visual and auditory stimuli were presented simultaneously for 200 ms, and they consisted of three syllables without meaning in Dutch. They were formed by a plosive consonant and the same vowel ("pi," "ti," and "ki"). The timing of the stimuli onset and their duration was carefully controlled. The use of the same vowel ("i") in all stimuli further allowed us to guarantee that the length of the syllables was stable.
Each syllable was delivered with the same probability in both sensory domains. From the total number of trials (798), roughly 75% (599) were incongruent (different syllable between visual and auditory modality) and 25% (199) were congruent (same syllable in both modalities). The higher number of incongruent trials was originally planned to promote anticipatory suppression of distracting information from the irrelevant sensory modality, as this was the main objective of a previous report in which we showed that alpha oscillations can be phase adjusted in anticipation of relevant stimuli at sensory regions (Solis-Vivanco, Jensen, & Bonnefond, 2018). Nevertheless, congruent trials were included in order to explore VAN recruitment, therefore, the focus for this study was beyond sensory areas. Moreover, this proportion resembles oddball tasks used to explore TPJ activation by infrequent stimuli (Corbetta & Shulman, 2002). Visual stimuli were presented at the center of the screen in white. Auditory stimuli were digitally created using a male voice and delivered from the computer controlling the task to plastic ear-tubes adapted to MEG recordings and inserted in participants' ear canals. Each syllable was associated with either one of three buttons in a response pad. Participants were asked to respond as accurate and fast as possible to the syllable in the modality they were instructed to attend in each trial, by pressing the corresponding button using their right index, middle, or ring finger.
The correspondence between the side of the cue and the modality to attend, and the assigned syllables to the buttons were counterbalanced across participants. All trials were randomly distributed across participants. Five breaks were introduced in the experiment, in which participants were informed about their performance.
Reaction times (RT) and response accuracy were recorded along the experiment. Participants were given up to 1,500 ms to respond. During both the training and experimental sessions, clear perception of stimuli was verified in each participant. In addition, although the MEG system we used provides appropriate prescription glasses to be used during acquisition for vision correction, none of our participants needed them.

| Data acquisition
We used a whole-head magnetoencephalography (

| Procedure
The experiment was conducted over three sessions for all participants. During the first session, inclusion criteria were confirmed, general information about the study and informed consent letters were provided, and detailed instructions about the experiment were presented. Participants then performed a practice session with 150 trials inside the MEG room. During the second session, the participants' head shape was digitized, and the actual MEG experiment was conducted. During the third session, the MRI was obtained. All data are available by request to the authors.
Epochs of the MEG recording extending 2 s before and 1 s after the onset of visual and auditory stimuli were extracted. Only epochs containing correct responses were considered for further signal analyses.
Special care was taken to identify and remove artifact activity. Trials containing muscle artifacts, superconducting quantum interference F I G U R E 1 Experimental paradigm. After a lateralized somatosensory cue indicating which sensory domain to attend (visual or auditory), participants were asked to press one of three buttons according to the relevant stimulus in that domain. Baseline, anticipatory, and poststimuli (congruency) time periods are indicated device (SQUID) jumps or eye blinks and saccades (as shown by the eye tracker signal), were rejected using an automatic routine based on mean z-scores across sensors exceeding a threshold given by the data variance within each participant (cut-off of ± 2 SEM). Since visual stimuli were always at the center of the screen and no visual search was needed, long eye movements were highly infrequent. We trained our participants to blink exclusively after giving their response in each trial, so artifacts due to blinking were particularly infrequent as well.
Additional visual inspection was applied to the remaining trials before demeaning and including them in further analyses. The mean number of trials included in the analysis was 558 ± 112, with no significant differences between Attend-visual and Attend-auditory conditions (e.g., across time and frequencies), which was our case (Luck, 2014).
Epochs were analyzed at sensor and source level. The analyses performed at sensor level are essential to identify target time and frequency ranges. Source analyses were then used to determine the source origin of the effects observed at sensor level and to explore the oscillatory activity in specific regions of interest. For the sensor-level analyses, planar gradients of the MEG field distribution were calculated (Bastiaansen & Knosche, 2000). We used a nearest neighbor method where the horizontal and vertical components of the estimated planar gradients were derived, thus approximating the signal measured by MEG systems with planar gradiometers. The planar gradients representation facilitates the interpretation of the sensor-level data, since the largest signal of the planar gradient typically is located above the source (Nolte, 2003).
Time-frequency representations (TFR) for absolute power from 3 to 100 Hz were obtained using a fast Fourier transformation (FFT) approach with an adaptive sliding time window three cycles long (ΔT = 3/f; e.g., ΔT = 300 ms for 10 Hz), similarly to previous studies (Bonnefond & Jensen, 2012;Solis-Vivanco, Jensen, & Bonnefond, 2018). A Hanning taper (also ΔT long) was multiplied by the data prior to the FFT. For the planar gradient, the TFR of power were estimated for the horizontal and vertical components and then summed. The power for the individual trials was averaged over conditions and log-transformed.

| Source analysis
A frequency-domain beamforming approach based on adaptive spatial filtering techniques (Dynamic imaging of coherent sources; DICS) was used to estimate the absolute power at source level in the entire brain (Gross et al., 2001). We obtained cross-spectral density matrices by applying a multitaper FFT approach (ΔT = 300 ms; 1 Slepian taper resulting in 4 Hz smoothing) on data measured from the axial sensors.
For each participant, a realistically shaped single-shell description of the brain was constructed, based on the individual anatomical MRIs and head shapes (Nolte, 2003). The brain volume of each participant was divided into a grid with a 1 cm resolution and normalized with respect to a template MNI brain (International Consortium for Brain Mapping, Montreal Neurological Institute, Canada) using SPM8 (http://www.fil.ion.ucl.ac.uk/spm). The lead field and the crossspectral density were used to calculate a spatial filter for each grid point (Gross et al., 2001) and the spatial distribution of power was estimated for each sensory condition (Attend-visual/Attend-auditory) and congruency (congruent/incongruent) in each participant. A common filter was used whenever two conditions were compared (based on the cross-spectral density matrices of the combined conditions). As for the sensor level analyses, the estimated power was averaged over trials and log-transformed. The power difference between sensory conditions (visual/auditory) and congruency or time periods was calculated and averaged across participants. For the source reconstruction 33 subjects were included as the MRI of 1 subject was missing.
All source data were estimated around 15 Hz according to the peak frequency effect observed in sensor level analyses (see Results section). The source estimates were plotted on a standard MNI brain found in SPM8.
In order to explore the oscillatory dynamics within regions of interest (ROI) of the VAN (see Results section), we used a linearly constrained minimum variance (LCMV) scalar beamformer spatial filter algorithm to generate maps of source activity on a 1 cm grid (Van Veen, van Drongelen, Yuchtman, & Suzuki, 1997). The beamformer source reconstruction calculates a set of weights that maps the sensor data to time-series of single trials at the source locations, allowing to reconstruct the signal at source level. In addition to TFR of power, we explored the functional connectivity across these reconstructed time series by means of TFR of coherence. In accordance to Nolte et al. (2004), we used the imaginary part of the coherence value, since it is less biased by power. All of our analyses were focused on the time period before the onset of stimuli (i.e., the anticipatory period, during which we expected a suppression of the VAN activity due to topdown attentional orientation compared with baseline) and the time period after (during which we explored VAN modulations due to a congruency effect between sensory modalities). A 500 ms time window from −700 to −200 ms with respect to the onset of the somatosensory cue was used as baseline ( Figure 1). This time window was an appropriate baseline measure as the activity in the frequency range of interest was not modulated during this time, that is, it did not exhibit any anticipatory modulation.

| Statistical analysis
Since RT showed normal distributions (Kolmogorov-Smirnov Z for both sensory conditions and congruency modalities ≥0.55, p ≥ .41), they were analyzed using repeated measures analysis of variance (ANOVA) (RM-ANOVA) with condition (Attend-visual and Attendauditory) and congruency (congruent and incongruent) as withinsubject factors. For all described RM-ANOVA, a Greenhouse-Geisser correction was used in case of violation of sphericity assumption and the Bonferroni test was used for post hoc comparisons.
Significant differences of power due to top-down modulation (i.e., anticipatory vs. baseline time periods in both sensory conditions) or congruency (congruent vs. incongruent) at both sensor and source levels were assessed using a cluster-based nonparametric randomization test (Maris & Oostenveld, 2007). This test controls for the Type I error rate in situations involving multiple comparisons over sensors, frequencies and times by clustering neighboring sensors, time points and frequency points that show the same effect. For this analysis we included frequencies from 3 to 40 Hz (using 1 Hz increments) with an adaptive time window long enough to include at least 3 cycles in each frequency. We explored from −600 ms to the onset of stimuli for the anticipatory period, and from 200 to 500 ms after for the ( (g) Higher increase of 12-20 Hz power for both conditions at source level was associated with less total number of interference errors along the task (p = .004) reference distribution to evaluate the statistics significance of a given effect (Monte Carlo estimation). Additionally, for all source level analyses we also conducted a false discovery rate (FDR) correction. This correction allowed us to overcome some limitations of the cluster correction approach such as considering a set of connected smaller cluster (by chance) as one big cluster. Only clusters surviving both the cluster correction and the FDR were reported.

| RESULTS
We used a bimodal (visual/auditory) attentional task that included cueing for relevant stimuli (Figure 1) to quantify the neurophysiological activity associated with active suppression of the VAN activity during top-down guided attentional allocation. By controlling congruency between the sensory modalities, we also explored the recruitment of the VAN when unexpected relevant (congruent) information arising from the unattended modality was presented, and whether this improved task performance.

| Congruent stimuli enhance task performance
Behavioral results were reported before in Solis-Vivanco, Rodriguez-   Figure 3b). In order to explore the causal relationship between DAN and VAN nodes, we calculated TFR of the phase-slope index between them for both conditions. Nevertheless, these analyses did not provide reliable results, probably due to small signal-to-noise ratios. No further causal analyses were carried out.
In summary, a power increase of 12-20 Hz was observed in right cortical regions before the onset of relevant stimuli, including areas from the VAN and DAN. In addition, higher increase of power in these regions predicted better task performance. Moreover, we observed increased functional connectivity between VAN (especially TPJ) and DAN nodes (SPL and FEF), during this period. Both increases in power and connectivity were stronger in those participants with better ability to filter out distracting visual information.

| Alpha power decrease in visual cortex in anticipation of stimuli
Alpha/Beta modulation has been reported more often in sensory networks than in other networks. As reported in Solis-Vivanco, Since this result has been extensively described in this previous paper, it was not further analyzed here.

| The VAN is recruited after detection of congruency across sensory modalities
We explored whether the regions that showed 12-20 Hz power increase during the anticipatory period (VAN and DAN nodes) were also modulated by enhanced attention, that is, elicited by congruency between attended and unattended stimuli. We selected grid points with maximal power differences between the anticipatory period and baseline including both conditions together (although this grid points were also significant for each condition separately) and reconstructed the signal at those points during the congruency period (from stimuli onset to 600 ms afterwards) by means of an LCMV filter. These grid points included the right TPJ, IFG, FEF, SPL, and MFG (Figure 4a).
TFRs of these ROIs revealed a clear decrease for congruent compared with incongruent trials in the 12-20 Hz band starting around 150 ms after stimuli onset (Figure 4b). Interestingly, this congruency effect was earlier for Attend-visual than for Attend-auditory trials. When 25%), we explored whether congruent trials evoked the P300 eventrelated field (P300m) at parietal regions due to an oddball effect (Polich, 2007). Nevertheless, this ERF was not observed for any condition (data not shown).
When exploring with a RM-ANOVA the functional connectivity across relevant ROI (TPJ, IFG congruency effect was observed at this frequency range ( Figure S3).
In summary, we found a power decrease, earlier in the visual condition, in the same frequency range as the anticipatory period in the VAN (TPJ and IFG), but also in the DAN (FEF and SPL), for congruent compared with incongruent trials. Such decrease predicted performance speed. In addition, increased connectivity between VAN and DAN nodes was observed for congruent trials. These results suggest an involvement of the two networks reflecting effective attentional enhancement after unexpected detection of relevant information in previously unattended sensory modalities.

| DISCUSSION
In the present study, we aimed to determine (a) the oscillatory profile of the suppression of the VAN during top-down oriented attention processes and (b) whether this network was recruited when the information presented in an unattended sensory modality was congruent with the information presented in the attended modality, a special case of attention enhancement, in order to improve task performance.
We found a power increase of alpha/beta (12-20 Hz) oscillations in the VAN during top-down attentional orientation. This increase was papers that investigated this question (Shulman et al., 2003;Shulman et al., 2007;Todd et al., 2005) and is in line with the hypothesis that this suppression would reflect a mechanism allowing to protect goaldriven behavior from distractors. We show here that such suppression is expressed in a broad frequency band related to high alpha/low beta oscillations, that is, in a higher frequency band compared with the frequencies observed in sensory networks. Importantly, the same frequency range was modulated during sensory processing (see below).
However, modulations in low beta over sensory regions have also been reported during attentional tasks (e.g., van Ede, Koster, and Maris (2012) and Siegel et al. (2008)). Moreover, a modulation in a similar frequency range (10-20 Hz) in and between nodes of the VAN and DAN has been reported during a visual search task (Spaak, Fonken, Jensen, & de Lange, 2016) as well as in anticipation of or during the processing of matching stimuli in multisensory paradigms (Goschl, Friese, Daume, Konig, & Engel, 2015;Misselhorn, Friese, & Engel, 2019;Wang, Goschl, Friese, Konig, & Engel, 2019). Interestingly, a comprehensive study has demonstrated that alpha-beta frequency peaks differ across regions and experimental designs (Haegens, Cousijn, Wallis, Harrison, & Nobre, 2014) (see also ElShafei et al. (2018)). Altogether, these results indicate that the frequency range of alpha (and beta) oscillations might diverge between brain regions or between tasks although they might still be associated with a similar mechanism, for example, functional inhibition. Further studies are required to understand whether these differences result for example, from the anatomical connectivity of the networks involved and/or from the requirement for multi-timescale processing according to the cognitive process involved.
Furthermore, we observed a stronger coherence between the nodes of the VAN network as well as between FEF (part of the DAN) and TPJ during the anticipatory period. Interestingly, participants with stronger ability to suppress visual distractors showed higher connectivity across these nodes. This latest result could provide evidence in favor of the idea that the suppression of the VAN is driven by the DAN (Shulman et al., 2003). Nevertheless, since we did not find a clear direction of such connectivity between VAN and DAN (possibly due to a reduced signal-to-noise ratio), this hypothesis should be taken with caution. On the other hand, we found a decrease in connectivity between MFG and DAN (SPL) during top-down attention.
This was unexpected, since the right posterior MFG has been discussed as another candidate region for linking the dorsal with the ventral system (Corbetta et al., 2008). Further research is needed to confirm and understand our result.
We further observed that the VAN was recruited, as indexed as well by a power decrease in the 12-20 Hz band, when a congruent stimulus was presented in the unattended sensory domain. Importantly, during stimuli processing and when comparing congruent versus incongruent trials there was no evidence of power modulation on left regions as observed during the anticipatory period, which supports the attentional role of this network, rather than a motor one.
The congruency effect is a special case of attentional enhancement, as relevant information is still present in the attended dimension though its saliency is increased due to multisensory integration, and includes a consequent behavioral benefit (Gau et al., 2020;Van der Burg, Olivers, Bronkhorst, & Theeuwes, 2008). Importantly, this effect remains even when congruent trials are rare ( Van der Burg et al., 2008), resembling oddball paradigms under which the VAN is usually activated (Kim, 2014). Furthermore, the VAN alpha/beta decrease predicted response speed in both conditions and was observed earlier in the Attend-visual condition, that is, when the unattended stimulus was presented in the auditory domain, compared with the Attend-auditory condition. The processing of auditory stimuli has been shown to be faster than processing of visual ones, which could explain the earlier activation of the VAN in the Attend-visual condition (Pain & Hibbs, 2007). However, it should be noted that reaction times were faster in the Attend-visual condition, congruency effects were stronger for the Attend-auditory condition at sensor level, and VAN suppression was better in the strong visual suppressors during the anticipatory period. These results may reflect the sensory dominance of the visual domain, and hence the need for more effective modulation (suppression/recruitment) of the VAN for this type of information.
Interestingly, the DAN was also more activated during congruent than incongruent trials after an early decrease in IFG and TPJ, although only FEF reached the significant level after multiple comparison corrections in the Attend-auditory condition. In line with this, previous fMRI work has also reported higher FEF activity during reorienting of attention (e.g., Corbetta & Shulman, 2011;Vossel, Thiel, & Fink, 2006). Although the timing of activation of the different networks would need to be further investigated, it seems that the reflecting a suppression of this node after its recruitment. The time window we could analyze did not allow to determine whether a similar increase was later observed in the Attend-auditory condition.
We also found a congruency effect in the visual cortex for the Attend-visual condition, although it did not survive multiple comparisons corrections. The congruency effect only in the Attend-visual condition, not in the Attend-auditory condition, over the visual cortex (i.e., the relevant area) suggests that congruency further enhances the processing of the attended stimulus and not necessarily of the unattended stimulus. Interestingly, such enhancement operates in the same frequency range as the one in VAN and DAN nodes, possibly facilitating communication between sensory and attentional networks.
Further investigation will be required to test this hypothesis.
In addition, we observed a coherence increase for congruent trials within and between VAN and DAN (TPJ, IFG, FEF, and SPL). This alpha/beta synchrony could reveal the mechanism allowing the interaction within and between these networks during involuntary attentional enhancement (Vossel et al., 2014), though we did not find a direct association with task performance. It should be noted that We did not find any significant difference between congruent and incongruent trials in other frequency bands, neither in the theta band as reported by Proskovec et al. (2018)  reported that irrelevant, novel auditory stimuli generate a reduction of power in the alpha/beta band at parietal regions (Solis-Vivanco, Rodriguez-Violante, et al., 2018). Nevertheless, whether the source of this decrease includes the VAN (and DAN) remains to be explored. In addition, future studies might explore the role of disengagement (i.e., full switch from attended to unattended domain) over DAN and VAN activation, which was explored in the Posner task used by Proskovec et al. (2018), but not necessarily present in our study.
As a final remark, we hypothesized that the decrease of the BOLD signal observed in the VAN network during similar tasks in fMRI (Shulman et al., 2003;Shulman et al., 2007;Todd et al., 2005) would be reflected in an increase of alpha/beta oscillations as observed over sensory areas (Haegens et al., 2011;Sadaghiani et al., 2012). We therefore considered the observed increase of alpha/beta power in the VAN as potential evidence of inhibition of the activity of this network. While the top-down role of alpha/beta activity, mainly in the DAN and sensory hierarchy, has been reported in the literature (Bastos et al., 2015;Michalareas et al., 2016), and without necessarily alluding a potential inhibitory role (Lobier, Palva, & Palva, 2018), we suggest that both possibilities (i.e., top-down regulation and functional inhibition) are not mutually exclusive. This is particularly true when we consider the activity of the VAN in addition to, or more specifically in interaction with, the activity of the DAN. We propose that changes in alpha/beta power and synchronization indicates the involvement of the VAN in both reducing interference from distractions (alpha/beta power increase during delay) and extracting relevant information from unattended channels (alpha/beta decrease in the congruent condition), in both cases in interaction with the DAN. We therefore consider the changes in alpha/beta power in the VAN as well as the increased connectivity with the DAN as correlates of attentional top-down guidance through functional inhibition.
Among the limitations of our study, we did not explore the VAN and DAN effects at sensory regions. A recent fMRI study by Rossi, Huang, Furtak, Belliveau, and Ahveninen (2014) showed increased connectivity between auditory cortex and different nodes of the DAN and VAN during cued voluntary and novelty-driven auditory orienting, respectively. In addition, our sample included only young adults.
Future research might explore the VAN and DAN modulation during attentional orientation and involuntary enhancement along development, including children and older adults. In addition, how these networks can be compromised in patients with neurologic and psychiatric disorders with attention impairment remains to be explored.
In conclusion, our results show that effective attentional allocation, regardless of sensory modality, includes the modulation and cooperation between ventral and dorsal attention networks through upper-alpha/beta oscillations.

ACKNOWLEDGMENTS
We thank Rocio Silva-Zunino, Jessica Askamp, and Paul Gaalman for their technical assistance. We also thank Emiliano Macaluso for his helpful discussions regarding the theoretical framework and references.

CONFLICT OF INTERESTS
The authors declare no competing financial interests.

ETHICS STATEMENT
This study fulfilled the Declaration of Helsinki criteria and was approved by the local board ethics committee at the Donders Institute for Brain, Cognition, and Behaviour, Radboud University.

DATA AVAILABILITY STATEMENT
All data are available by request to the authors.