Beta band frequency differences between motor and frontal cortices in reaching movements

A. Subjects

Twenty-six right-handed healthy subjects (age range: 20–68; mean ± SD: 40.8 ± 18.2, 16 female) were enrolled for this study. Experiments were conducted with the approval of our Institutional Review Board and each subject expressed written consent to participate to the study.

B. Experimental Design

Subjects performed 15 sets of 56 movements each in a 30-minute session. They were seated in front of a screen with their right upper-limb hidden by a surface and were asked to move a cursor with the right arm on a digitizing tablet to reach for targets appearing on the screen [8], [9]. An array of eight targets (1 cm diameter) equidistant (4 cm) from the center was displayed for entire session (Fig. 1a). Every 1.5s, one target randomly blackened for 400 ms and subjects had to reach it moving as soon, fast and accurately as possible, but without anticipation or guessing. Participants were also instructed to make overlapping out-and back trajectories without any correction or stop. Before the experiment, they were trained to reach a hit rate of at least 95%. High-density EEG (256-channels Electrical Geodesic Inc.) was recorded during the entire session, with the high-impedance amplifier Net Amp 300 and Net Station 4.3 (Electrical Geodesic Inc.). Sampling rate was 250 Hz, impedances were kept below 30 kΩ; the signal was referenced to the vertex Cz.

C. Data Analyses

For each movement we computed (Fig. 1): reaction time, time between target appearance and movement onset; peak velocity; hand path area, area included in the trajectory, normalized by squared path length; total movement time, between onset and end of the out and back movement. For each subject, movements exceeding two SD from the mean in any kinematic indices were excluded from further analysis.

EEG signal was bandpassed and Notch filtered (two-way least-square FIR filter between 1 and 80 Hz; Notch filter 60Hz), before being divided into 4-seconds epochs (−1 to 3 second around the stimulus onset). Data were visually inspected to remove artefactual channels and epochs. Additionally, we used Independent Component Analysis (ICA) with Principal Component Analysis to remove stereotypical artifacts such as muscle-related signal, blinks and saccades [10]. Finally, all the channels previously removed were replaced with spherical spline interpolation and data were averaged-referenced.

After the preprocessing, we centered each trial on the movement onset. Then, we computed Fast Fourier Transformation (Hanning taper, 20 ms time steps, 5 cycles adaptive window) in the beta range, i.e. between 15 and 30 Hz, over three regions of interests (ROIs): a left, right and frontal ROI, expressing consistent movement-related beta modulation across subjects [6], [7]. These ROIs were defined on the grand-average taking the electrode with the highest ERD, ERS, or beta modulation depth, and its six closest neighbors (Fig. 1d). For each ROI, we normalized the power of each trial with the average beta power of the session. Afterward, we computed ERD and ERS magnitude: ERD was defined as the minimum desynchronization occurring in a window between 200 ms before movement onset and 700 ms after it; ERS was the maximum value in a 1-s window after the movement (500 ms to 1500ms after movement onset). These windows were chosen in order to select the peak ERD and ERS (Fig. 1c). We also subtracted the ERD magnitude from the ERS to obtain the beta modulation depth (Fig. 1c). All the analysis and preprocessing of the data have been implemented in Matlab, using EEGLAB [11], Fieldtrip [12] toolboxes, as well as customized code.

D. Statistical Analyses

Changes in kinematic indices were assessed with repeated measure ANOVAs with sets as factors. For EEG measures we used mixed-model ANOVAs with ROIs and sets as within-subject factors. Violation of sphericity was corrected with Greenhouse-Geisser. Significant main effects were followed by Bonferroni-corrected pairwise comparisons; time-frequency differences between ROIs were assessed through Nonparametric Permutation Testing with False Discovery Rate Correction. Correlation between kinematic and EEG data have been assess through Pearson’s coefficients with Bonferroni correction for multiple comparisons. Results were considered significant for p-values <0.05. All statistical analyses were conducted with SPSS v25, Matlab 2017b and Fieldtrip toolbox for Matlab.

A. Subjects reduce their total movement time with practice

All subjects completed the tasks without difficulty; movements had bell-shaped velocity profiles and were mainly straight. Kinematic results are summarized in Fig. 2. Briefly, we found that reaction time and peak velocity increased across sets, however they do not reach significance (F(4.86,121.40)=1.53, p=0.188; F(1.83,45.73)=1.29, p=0.285, respectively). Hand path area, an inverse index of trajectory accuracy, showed a borderline significant decrease across sets (F(6.47, 161.83)=1.96, p=0.069). Total movement time showed a significant effect of practice (F(3.24, 80.97)=8.01, p<0.001).

B. Beta modulation depth in the right sensory-motor area is lower compared to the other ROIs

We then analyzed ERD, ERS and beta modulation depth differences in the three ROIs, as well as their magnitude changes with practice. In agreement with previous results, we found that the right sensory-motor ROI, ipsilaterally to the movement, showed lower ERD, ERS and beta modulation depth than the other two ROIs (Fig. 3). Specifically, the right ROI ERD was significantly lower than the left (post hoc test p=0.006) and frontal (p=0.002) ones. Interestingly, the frontal ROI had borderline significant greater ERD than the left ROI (p=0.055). ERS and beta modulation depth were significantly lower in the right ROI (ERS: right vs. left p=0.009; right vs. frontal p=0.004; left vs frontal p=0.936 Modulation Depth: right vs. left p=0.006; right vs. frontal p=0.003; left vs. frontal p=0.743). Repeated measure ANOVAs also indicated a significant increase across sets for ERD, ERS and modulation depth (Tab. I); no interactions were found between ROIs and sets. We then correlated practice-related kinematic changes with ERD, ERS and beta modulation increases; however, we did not find any direct relationship between performance and beta modulation changes.

C. Timing of ERD and ERS is similar in the three ROIs

We investigated whether ERD and ERS timings at their peak occurrence were similar in the three ROIs (Fig. 4). We found no difference between ROIs (ERD: F(1.17, 29.16)=0.01 p=0.956; ERS: F(1.14, 28.60)=0.27, p=0.637); the results of the ANOVA also revealed that timing of peak ERS decreased in the course of practice (F(6.21, 155.15) = 3.74, p=0.001), while the ERD timing did not change (F(14,350)=1.22, p=0.258) and no interactions were found (ERD: F(11.99, 299.78)=0.70, p=0.751; ERS: F(11.57, 289.36)=0.74, p=0.703). Correlations between timings and kinematic parameters did not revealed any relationship between ERD and ERS time changes and performance.

D. ERS of the frontal ROI occurs in a different range of the beta band compared to the sensory-motor cortices

We finally verified whether the movement-related beta oscillatory changes occurred in the same frequency range in the three ROIs. Thus, we isolated the frequency of the peak ERD and ERS. As a first step, we checked whether ERD and ERS peak frequency changed across sets and ROIs. ANOVAs results on peak ERD did not reveal any effect of set (F(14,350)=0.47, p=0.946), ROI (F(2,50)=0.59, p=0.561) and interaction (F(28,700)=0.84, p=0.705). On the other hand, analysis on ERS peak showed a main effect of ROI (Fig. 5; F(2,50)= 7.93, p=0.001). Pairwise comparison revealed that the ERS frequency of the frontal ROI was higher than those of the sensory-motor ones (frontal vs left p=0.001; frontal vs right p=0.007) that did not differ from each other (p=1.000, Fig.4b). No effect of sets (F(7.1, 177.50)=1.28, p=0219) nor interaction were found (F(10.59, 264.81)=1.12, p=0.348). This difference of ERS frequency between the frontal and the sensory-motor regions was even more visible when we computed time-frequency representations of the three ROIs (Fig. 6) on all the trials. In fact, Nonparametric Permutation Testing with False Discovery Rate correction revealed that ERS occurred in the range between 15 and 18 Hz over the left sensory-motor ROI, while the frontal ROI ERS between 23 and 29 Hz (Fig. 7). We did not find any frequency difference between ROIs when we analyzed ERD, which mainly occurred around 19 Hz. Of note, we only compared frequency differences between the left and frontal ROIs, because the significant lower magnitude values of the right ROI could have biased the results. Finally, we did not find any correlation between practice-related peak frequency of both ERD and ERS and kinematic changes.

A. ERD and ERS magnitude changes with practice

During a 30-minute reaching task we found a progressive increase of ERD, ERS and beta modulation depth over all the ROIs. A possible explanation of such ERD, ERS increase might be fatigue: with time subjects may get tired and thus require a higher involvement of the sensory-motor cortex represented by a lower ERD and a higher ERS. However, behavioral data do not reveal any fatigue. Indeed, neither reaction time nor peak velocity decrease with practice. We thus speculated that the lack of correlation between practice- related magnitude changes and kinematic data can depend on the fact that beta modulation does not directly reflect movement characteristics, but instead it can be related to movement planning and execution of a movement through sensory-motor integration processes [13]–[15]. Also, practice-related beta modulation may represent plasticity and thus LTP-like phenomena in the sensorimotor cortex [5]–[7].

B. Beta modulation depth is reduced in the right sensory-motor area

In agreement with previous results [6], [7], we observed greater ERD, ERS and beta modulation over the sensory- motor cortex contralaterally to the moving arm compared to the ipsilateral ROI. In fact, this area has been reported to be active during voluntary movements [8], [16]. However, beta modulatory activity over the ipsilateral motor cortex showed, albeit to a lesser extent, a similar increase across sets, suggesting a direct involvement of such area in movement production. Previous works have speculated on the role of the ipsilateral sensory-motor cortex on movement planning and execution; its activation may be related to movement organization and mirror movement suppression [17]–[21]. Moreover, animal studies revealed that projections between proximal sensory-motor areas exist and thus the activation of the contralaterally motor areas could spread to the ipsilateral through callosal connections [22]–[24]. Also, ERS has been related to increased cortical inhibition in the contralateral but not in the ipsilateral motor cortex [25]–[27]. The different role of the motor cortices during voluntary movement may explain the magnitude differences of ERD, ERS and beta modulation. Finally, the involvement of the frontal cortex, which shows beta modulatory activity of similar magnitude to that of the contralateral motor cortex, could be explained by the different functions related to motor control that occur in the frontal area, such as attentional processes related to motor planning and performance as well as to the updating of the internal models that are necessary for efficient planning and execution [28].

C. ERS of the frontal ROI shifts toward higher frequencies within the beta band

One of the most innovative findings is that, compared to the sensory-motor areas, the frontal ERS is overall shifted toward the high beta range. Intriguingly, ERS peak occurred simultaneously over the three ROIs and does not present magnitude differences with the left ROI. Time-frequency representations revealed that the ERS in the left sensory-motor cortex occurs between 15 and 18 Hz, the frontal ROI shows an ERS in the higher beta range, between 23 and 29 Hz. The frontal cortex has several functions that require the integration of information from multiple regions of the brain. In fact, it is responsible for learning, memory, attention, and motivation, in order to plan the most appropriate behavior in response to internal and external stimuli. During a movement, the frontal cortex combines information from sensory and motor cortices; in particular, studies on reaching and catching movements underlined the importance of the frontal region during monitoring and correction of an ongoing performance and, most importantly, during updating of internal models [29]–[31]. Some studies hypothesized that beta activity originates from multiple sources and thus reflects different functional processes [16], [32]–[34]. Our result perfectly fits this idea that beta can have many roles that can be distinguished by differences in the predominant frequency. Also, our finding is further supported by previous studies showing that the ERS frequency following finger dorsal flexions is lower in the hand motor area, compared to the supplementary motor cortex [13], [14]. A possible explanation of this phenomenon may be that, after the movement ends, the sensory motor region and the frontal area show different oscillations that may be related to different processes.