Post-stimulus beta responses are modulated by task duration

Abstract

Modulation of beta-band neural oscillations during and following movement is a robust marker of brain function. In particular, the post-movement beta rebound (PMBR), which occurs on movement cessation, has been related to inhibition and connectivity in the healthy brain, and is perturbed in disease. However, to realise the potential of the PMBR as a biomarker, its modulation by task parameters must be characterised and its functional role determined. Here, we used MEG to image brain electrophysiology during and after a grip-force task, with the aim to characterise how task duration, in the form of an isometric contraction, modulates beta responses. Fourteen participants exerted a 30% maximum voluntary grip-force for 2, 5 and 10 s. Our results showed that the amplitude of the PMBR is modulated by task duration, with increasing duration significantly reducing PMBR amplitude and increasing its time-to-peak. No variation in the amplitude of the movement related beta decrease (MRBD) with task duration was observed. To gain insight into what may underlie these trial-averaged results, we used a Hidden Markov Model to identify the individual trial dynamics of a brain network encompassing bilateral sensorimotor areas. The rapidly evolving dynamics of this network demonstrated similar variation with task parameters to the 'classical' rebound, and we show that the modulation of the PMBR can be well-described in terms of increased frequency of beta events on a millisecond timescale rather than modulation of beta amplitude during this time period. Our results add to the emerging picture that, in the case of a carefully controlled paradigm, beta modulation can be systematically controlled by task parameters and such control can reveal new information as to the processes that generate the average beta timecourse. These findings will support design of clinically relevant paradigms and analysis pipelines in future use of the PMBR as a marker of neuropathology.

Fig. 1

Overview of the experiment. (A) The grip-force bar (Current Designs, Philadelphia, USA) attached to a fingerless glove to allow relaxation of the hand. (B) Example single trial. The target force profile is shown (red) with real-time force output from a single trial overlaid (black). The visual stimulus appeared 2 s before the force output period, which was sustained for 2, 5 or 10 s [shown here for 2 s]. The profile remained on the screen for a further 0.5 s after the end of the force output period and was followed by a fixation cross for 27.5 s. (C) Schematic diagram of one run. Single trials were repeated 15 times for each duration in a pseudo-random order, totalling 45 trials within one run. This was followed by a second run after an approximately 15-min break. Two 3-s target profiles of 100%MVF were presented before and after each run to monitor fatigue.

Fig. 2

Behavioural results. (A–C) Example of output for one subject of (A) grip-force, (B) forearm extensor bundle EMG trace, (C) forearm flexor bundle EMG trace. (D–F) Average 2-s, 5-s and 10-s responses across subjects and runs for (D) grip-force, (E) EMG amplitude in the forearm extensor bundle, (F) EMG amplitude in the forearm flexor bundle.

Fig. 3

(A) Pseudo-t-statistical map showing the PMBR localised to motor cortex for one run of a single representative subject (radiological view). (B) Average time frequency spectrograms extracted from individual subject PMBR location for the three contraction durations (top panel shows 2-s task duration, middle 5-s task duration and the bottom panel shows 10-s task duration). Spectrograms show the relative change in power for each frequency band where baseline was 16–24 s (blue box). Time zero is cessation of the contraction. (C) Average time courses of beta band (15–30 Hz) amplitude for the three task durations from the peak location of the PMBR across 14 subjects. Responses are aligned to contraction offset (time = 0 s). Red line shows the response to 2-s task duration, green line response to the 5-s task duration and blue line to the 10-s task duration. Error bars show the standard error across subjects. (D–F) Measures from Weibull curves fitted to the PMBR showing effects of task duration. All times reported on y-axes are measured relative to contraction offset. (D) The amplitude of the PMBR peak (R-square 0.98), (E) the time at which peak of PMBR occurs (R-square 0.92) and (F) the time taken for rebound to return to baseline (R-square 0.01). Error bars show the standard error. Blue dashed line shows linear fit of the data.

Fig. 4

(A) Pseudo-t-statistical map showing MRBD localised to motor cortex for one run of a single representative subject (radiological view). (B) Average time frequency spectrograms extracted from individual subject MRBD location for the three contraction durations. Spectrograms show the relative change in power for each frequency band where baseline was 2–10 s prior to the end of the trial (blue box). Time zero is contraction onset. (C) Average time courses of beta band (15–30 Hz) amplitude for the three task durations from the peak location of the MRBD across 14 subjects. Responses are aligned to contraction onset (time = 0 s). Red line shows the responses to 2-s task duration, green line response to the 5-s task duration and blue line to the 10-s task duration. Error bars show the standard error across subjects. (D–E) Measures from a trapezoid fitted to the MRBD showing effects of task duration. (D) Amplitude of MRBD and (E) integral of MRBD plotted against task duration. Error bars show the standard error across subjects. Blue dashed line shows linear fit of the data.

Fig. 5

Pseudo-t-statistical map of the group average location of the MRBD (blue, peak (−40, −20, 58) mm) overlaid with the PMBR (red, peak (−36, −12, 62) mm) with the cross hairs at the PMBR peak (radiological view). Ŧ-stat maps were created in individual subject space before normalising to MNI space and averaging over subjects.

Fig. 6

Results from the sensorimotor state of the HMM. (A) State map of state 7 (see Fig. 7) showing a sensorimotor state. (B) Plots of the binary time course for a subset of the first 100 trials for the three conditions (2-s, 5-s and 10-s task duration) against time where dark grey is 1 (in state 7) and white is 0 (not in state 7). (C) Probabilistic time course derived from (B) showing probability of being in state 7 at any given time, for the three conditions. Responses are aligned to contraction offset (time = 0 s), akin to Fig. 3. Red line shows the 2-s task duration, green line is 5-s task duration and blue line is 10-s task duration. (D) Summary metrics for state 7, averaged over all conditions for each subject and then averaged over subjects, separated into three epochs relative to movement offset: post-movement (1–5s), movement (-4 – 0s) and rest (20–24s). Error bars show the standard error over subjects. Additional analyses are shown in Fig. 8.

Fig. 7

State maps from a HMM inferred with 8 states where red shows brain regions with increased power relative to average and blue decreased power, and associated state probability time courses for the three conditions, where red is 2-s, green is 5-s and blue is 10-s contraction.

Fig. 8

Summary metrics for state 7 (see Fig. 6) during the PMBR period (1–5s post movement offset) for the three contraction durations. Data are averaged over trials within each condition (2s, 5s and 10s contraction duration) and then over subjects. Error bars show the standard error over subjects.

Fig. A1

The effect of ISI on the PMBR and MRBD characterisation: (A) Trial averaged time frequency spectra and (B) beta band envelopes for three separate ISIs (5 s (red); 10s (green); 30 s (blue)). Baselines were determined as a single value taken from the long ISI condition. (C&D) Equivalent to (A&B) but where baselines were determined independently from each ISI.

Figure S1