Cross-frequency phase-amplitude coupling in repetitive movements in patients with Parkinson's disease.

Experimental design. A: schematic diagram of the experimental setup. The device for recording repetitive movements consisted of a transducer (green plate), recording the press force and two photoelectric sensors (white) reporting the height of the extended index finger. B: experimental protocol for the three movement tasks. Gray areas represent the short periods when subjects prepared to perform the task. Blue areas represent the active movement periods. C: experimental design of the tapping tasks. There were two conditions for each tapping task. Left: subjects performed repetitive tapping while looking at the fixation cross. Right: after each tapping cycle, subjects received color feedback indicating whether the extended index finger had reached the upper level. A green square indicated that the level of the preceding tapping movement had been at or above the upper photoelectric sensor. A red square indicated that the tapping height had been below the upper photoelectric sensor. Subjects were instructed to maximize green feedback by performing sufficiently extended tapping movements. D: schematic representation of the force and amplitude trajectories during pressing (top curve) and slow tapping (bottom curve). In pressing, the peripheral movement onset was defined as the moment when the force signal exceeded the lower force threshold (1.3 N). No force signals were resolved above the upper threshold of 4.4 N. During tapping, the peripheral movement onset was defined as the moment when the index finger was extended above the light beam of the lower photoelectric sensor. An omission was recorded if the index finger was not extended high enough to reach or cross the light beam of the upper sensor. E: estimation of the EMG slope. This was defined as the slope of the line (blue) connecting the 25% and 75% percentile (red dots) of the normalized EMG signal closest to the movement onset of tapping.

Selected performance parameters of the repetitive movement tasks. A: pressing: i) rate of repetitive pressing actions, ii) maximum amplitude of EMG recorded from FDI muscle during pressing, iii) slope of EMG activity in FDI muscle during the build-up of press force. Inset: averaged EMG curves of patients (red) and controls (blue). Group differences were estimated by Wilcoxon rank-sum tests. All P values were not significant even before correction. B: slow tapping: i) rate of repetitive slow tapping actions, ii) completion ratio of index finger extensions meeting the upper amplitude criterion (dotted: without feedback, hatched: with feedback), iii) slope of EMG activity in FDI muscle upon index finger extension. Inset: averaged EMG signals of patients (red) and controls (blue). Note the reduced EMG slope in patients. P values were adjusted by FDR correction in the respective post hoc tests. Only the adjusted P values below 0.05 are shown in the figure. C: fast tapping: i) rate of repetitive fast tapping actions, ii) rate of index finger extensions meeting the upper amplitude criterion (no visual feedback about the level of index finger extension), iii) box plot combined with mean lines showed decline of fraction of extension movements meeting the upper amplitude criterion across sequential 1-s time bins in fast tapping task without feedback. P values were adjusted by FDR correction in the respective post hoc tests. Only the adjusted P values below 0.05 are shown in the figure. The tests included 19 patients (6 females, mean age: 60.9 ± 10.8 yr) with and 20 healthy controls (8 females, mean age: 62.6 ± 7.9 yr). FDI, first dorsal interosseous; FDR, false discovery rate.

State-related PAC for the different tasks, averaged across four regions. We averaged zMVL values across the n × n pairwise PAC matrix and over the whole 3-s time series. A three-way nonparametric mixed ANOVA (PAC ∼ group × task × region) showed significant interaction effects between group and tasks. Post hoc tests of the group × task interaction effect showed significant differences between the resting state and all three movement tasks in the patients. Note that state-related PAC differed between patients and controls in resting state and during pressing. The tests included 19 patients (6 females, mean age: 60.9 ± 10.8 yr) with and 20 healthy controls (8 females, mean age: 62.6 ± 7.9 yr). P values were adjusted by FDR correction across 10 post hoc comparisons (4 between-group tests and 6 within-group tests). The adjusted P values below 0.1 are shown in the figure. FDR, false discovery rate; PAC, phase-amplitude coupling; zMVL, z score of the normalized mean vector length.

Fluctuation of PAC in the 3-s time series. A: PAC dynamics in the four conditions. Top: rectified z-score normalized EMG (means ± SE), recorded from first dorsal interosseus muscle, averaged after alignment to arbitrary time points (rest) or to peripheral movement onset (pressing, slow and fast tapping). Bottom: PAC (means ± SE) averaged after alignment to arbitrary time points (rest) or to peripheral movement onset (pressing, slow and fast tapping). B: box plot shows coefficient of variance of PAC across the four conditions. The tests included 19 patients (6 females, mean age: 60.9 ± 10.8 yr) with and 20 healthy controls (8 females, mean age: 62.6 ± 7.9 yr). P values were adjusted by FDR correction across 10 post hoc comparisons (4 between-group tests and 6 within-group tests). The adjusted P values below 0.1 are shown in the figure. dynPAC, dynamic PAC; FDI, first dorsal interosseous; FDR, false discovery rate; PAC, phase-amplitude coupling.

PAC dynamics across movement transitions. A: definition of five periods of movement based on four movement transition points in the pressing task (left) and the slow tapping task (right). We considered four kinetic (force threshold for pressing) or kinematic (amplitude criterion for tapping) events along the movement cycle as the peripheral movement transition points. The movement transition points (displayed in B and D) at the level of the cortex were determined by taking into account the mechanical delays, electromechanical delays and the corticomuscular conduction time to estimate the timing of the transition times. B: pressing task. Dynamic PAC resulting from averaging after alignment with the transition points on the (left) cortical movement onset (#1) and (right) cortical movement offset (#4). The timing of the other two cortical movement transition points (left: transition point #2, right: transition point #3) was determined based on the averaged durations of periods (left: P2 and P3, right: P3 and P4) across subjects. Note similar PAC modulation patterns for press onset and offset. C: box graph combined with mean lines of movement-related dynamic PAC in the 5 periods of a single pressing cycle in patients and controls. Two-way mixed ANOVA (dynPAC ∼ group × period) revealed significant main effects of group and periods, but no significant interaction effects. Post hoc comparisons were performed between neighboring periods (4 tests). P values were adjusted by FDR correction across four post hoc tests. Adjusted P values below 0.1 are shown in the figure. D: slow tapping task. Dynamic PAC resulting from averaging across subjects after alignment with the cortical movement transition point #1. The timing of the other three cortical movement transition points was based on the average duration of the following three periods (T2, T3, and T4) across subjects. E: box graph combined with mean lines of movement-related dynamic PAC in the five periods of a single tapping cycle in patients and controls. Two-way mixed ANOVA (dynPAC ∼ group × period) revealed significant interaction effects between group and period, suggesting that dynPAC of patients was modulated differently than that of controls. One-way ANOVA tests for the effect of period were applied to patients and controls separately, both revealed significant effects. Post hoc comparisons were then performed between neighboring periods (4 tests) in patients and controls, respectively. Patients displayed markedly less modulation of PAC than controls. P values were then adjusted by FDR correction across 4 post hoc tests. Adjusted P values below 0.1 are shown in the figure. The tests in the figure included 19 patients (6 females, mean age: 60.9 ± 10.8 yr) with and 20 healthy controls (8 females, mean age: 62.6 ± 7.9 yr). dynPAC, dynamic PAC; FDI, first dorsal interosseous; FDR, false discovery rate; PAC, phase-amplitude coupling.

Relationship between movement-related dynPAC and β power. A: dynamics of β power in the pressing task. Left: time-frequency spectrogram. Note the reduction of β power after the onset of the pressing in both patients and controls. Right: box plot combined with mean line graphs show dynamics of β power across the five periods of the pressing movement cycle. Significant modulations (FDR corrected values < 0.05, 0.01, 0.001) are marked by asterisks in the figure. B: dynamics of β power in the slow tapping task. Left: time-frequency metric plots. Right: box plot combined with mean line graphs show the dynamics of β power across the five periods of the slow tapping movement cycle. Significant modulations (FDR corrected values < 0.05, 0.01, 0.001) are marked by asterisks in the figure. C: scatter plots of the relationship between the absolute strength of PAC and β power in the pre-onset phase. D: scatter plots of the relationship between PAC change and power change from the first to the second period of the movement cycle. Left: pressing task. Right: slow tapping task. The tests in the figure included 19 patients (6 females, mean age: 60.9 ± 10.8 yr) with and 20 healthy controls (8 females, mean age: 62.6 ± 7. 9 yr). dynPAC, dynamic PAC; FDI, first dorsal interosseous; FDR, false discovery rate; PAC, phase-amplitude coupling.