NeuroImage 72 (2013) 83–90

Contents lists available at SciVerse ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img

Coherence between magnetoencephalography and hand-action-related acceleration,force, pressure, and electromyogram

Harri Piitulainen a,⁎,1, Mathieu Bourguignon a,b,1, Xavier De Tiège b, Riitta Hari a,c, Veikko Jousmäki a,b

a Brain Research Unit, O.V. Lounasmaa Laboratory and MEG Core, Aalto NeuroImaging, School of Science, Aalto University, PO BOX 15100, 00076 AALTO, Espoo, Finlandb Laboratoire de Cartographie Fonctionnelle du Cerveau, ULB-Hôpital Erasme, 808 Lennik Street, B-1070 Bruxelles, Belgiumc Advanced Magnetic Imaging (AMI) Centre, School of Science, Aalto University, PO BOX 13000, 00076 AALTO, Espoo, Finland

Abbreviations: CKC, corticokinematic coherence; Fhand action; F1, first harmonic of hand action frequencortex; MNI brain, standard Montreal Neurological Instsensory cortex; SM1, primary sensorimotor cortex.⁎ Corresponding author at: Brain Research Unit, O.V.

of Science, Aalto University, PO Box 15100, 00076 AALT470 22969.

E-mail address: harri.piitulainen@aalto.fi (H. Piitulai1 These authors equally contributed to the article.

1053-8119L © 2013 Elsevier Inc.http://dx.doi.org/10.1016/j.neuroimage.2013.01.029

Open access under CC BY

a b s t r a c t

a r t i c l e i n f o

Article history:Accepted 17 January 2013Available online 26 January 2013

Keywords:Corticokinematic coherenceCorticomuscular coherenceAccelerometerKineticsMagnetoencephalographySensorimotor cortex

Hand velocity and acceleration are coherent with magnetoencephalographic (MEG) signals recorded fromthe contralateral primary sensorimotor (SM1) cortex. To learn more of this interaction, we compared thecoupling of MEG signals with four hand-action-related peripheral signals: acceleration, pressure, force, andelectromyogram (EMG).Fifteen subjects performed self-paced repetitive hand-action tasks for 3.5 min at a rate of about 3 Hz. Eitheracceleration, pressure or force signal was acquired with MEG and EMG signals during (1) flexions–extensionsof right-hand fingers, with thumb touching the other fingers (acceleration; free), (2) dynamic index–thumbpinches against an elastic rubber ball attached to a pressure sensor (pressure and acceleration; squeeze), and(3) brief fixed-finger-position index–thumb pinches against a rigid load cell (force; fixed-pinch).Significant coherence occurred between MEG and all the four peripheral measures at the fundamental fre-quency of the hand action (F0) and its first harmonic (F1). In all tasks, the cortical sources contributing tothe cross-correlograms were located at the contralateral hand SM1 cortex, with average inter-source distance(mean±SEM) of 9.5±0.3 mm. The coherence was stronger with respect to pressure (0.40±0.03 in squeeze)and force (0.38±0.04 in fixed-pinch) than acceleration (0.24±0.03 in free) and EMG (0.25±0.02 in free, and0.29±0.04 in fixed-pinch).The results imply that the SM1 cortex is strongly coherent at F0 and F1 with hand-action-related pressureand force, in addition to the previously demonstrated EMG, velocity, and acceleration. All these measures, es-pecially force and pressure, are potential tools for functional mapping of the SM1 cortex.

© 2013 Elsevier Inc.Open access under CC BY-NC-ND license.

Introduction

Several hand-action-related signals, such as electromyogram (EMG)(Brown et al., 1998; Conway et al., 1995; Hari and Salenius, 1999;Mimaand Hallett, 1999; Salenius et al., 1996, 1997), movement velocity (Jerbiet al., 2007) and acceleration (Bourguignon et al., 2011, 2012) arecoherent with brain signals recorded with magnetoencephalography(MEG) and electroencephalography (EEG).

0, fundamental frequency ofcy; M1 cortex, primary motoritute brain; S1 cortex, primary

Lounasmaa Laboratory, SchoolO, Espoo, Finland. Fax: +358 9

nen).

-NC-ND license.

The corticomuscular coherence between MEG and EMG or EEG andEMG reflects modulation of motor-cortex drive to motor unitpopulations (Conway et al., 1995; Salenius et al., 1997). Corticomuscularcoherence is typically observed during sustained isometric contraction(Conway et al., 1995; Kilner et al., 1999; Salenius et al., 1997), and itpeaks at ~15–40 Hz depending on the exerted force (Hari and Salenius,1999; Mima and Hallett, 1999). Corticomuscular coherence decreasesduring dynamic movements with respect to the steady isometric holdperiod (Hari and Salenius, 1999; Kilner et al., 1999; Salenius and Hari,2003), and its sources display somatotopical order in the primarymotor (M1) cortex contralateral to the contracted muscle (Murayamaet al., 2001; Salenius et al., 1997).

The corticokinematic coherence (CKC), on the other hand, reflectscoupling betweenprimary sensorimotor (SM1) cortexMEGandkinemat-ic (e.g. acceleration) signals of executed or observed hand movements(Bourguignon et al., 2011, 2013a; Jerbi et al., 2007). CKC is observed dur-ing fast repetitive dynamic executed or observed handmovements at thesame frequency as the movements are performed, typically at 2–5 Hz

84 H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

(Bourguignon et al., 2011, 2012, 2013a; Jerbi et al., 2007). In addition tothe hand kinematics, rhythmic modulation of EMG during repetitive vol-untary hand movements is coherent with MEG at SM1 cortex at themovement frequency (~1.4–3.9 Hz) (Pollok et al., 2004, 2005a,b).

We have recently proposed CKC as a tool for functional mapping ofthe SM1 cortex (Bourguignon et al., 2011), but it is still unclear towhat extent hand-action-related kinetic signals such as force andpressure are coupled with the cortical MEG signals.

The purpose of the current study was to compare coherence betweenMEG and four hand-action-related peripheral signals (acceleration, pres-sure, force, and EMG) during repetitive self-paced continuous dynamic orfixed hand-actions. The tasks were selected to vary in their range ofmotion to clarify whether the degree of hand movement is crucial forthe coherence. We expected all these signals to be phase-synchronizedwith MEG signals as they all carry information about the fundamentalfrequency of the repetitive hand-actions. The coherence strengths and re-spective source locations were compared between the peripheral signals.

Materials and methods

Subjects

Fifteen healthy subjects (mean age 29.4 yrs, range 21–38 yrs;8 males, 7 females) without any history of neuropsychiatric disease ormovement disorderswere studied. According to Edinburgh handednessinventory (Oldfield, 1971), 14 subjects were right-handed (mean score92, range 67–100 on the scale from –100 to 100) and one subject wasambidextrous (–20).

The study had prior approval by the ethics committee of theHelsinki and Uusimaa hospital district. The subjects gave informedconsent before participation. Subjects were compensated monetarilyfor the lost working hours and travel expenses.

Experimental protocol

During MEG recordings, the subjects were sitting with their lefthand on the thigh, the right hand on a table in front of them. Earplugswere used to minimize concomitant auditory noise. A white papersheet was taped vertically on the MEG gantry to prevent the subjectsfrom seeing their moving right hand. Subjects were instructed to fixate

Fig. 1. Representative signals of one subject in free (left), squeeze (middle) and fixed-pin1–10 Hz) over the SM1 cortex. Acceleration signal is from one of the 3-axis accelerometer chnorm of the three accelerations (norm). Pressure and force signals were recorded during squmeasured during all three tasks.

a self-chosen detail in a picture (21×30 cm2) on the wall of the mag-netically shielded room, positioned 2.8 m in front of them, 11° to theleft from the midline.

Subjects performed three hand-action tasks: (1) free: dynamicflexions–extensions of right-hand fingers, with thumb touching theother fingers (unrestricted ~10-cm range of motion between thethumb and other fingers); the acceleration of the right index fingerwas monitored with a 3-axis accelerometer (ADXL335 iMEMS Acceler-ometer, Analog Devices Inc., Norwood,MA, USA) attached on the nail ofthe index finger (Fig. 1), (2) squeeze: dynamic index–thumb flexions(pinches) against an elastic rubber ball (~1-cm range ofmotion) attachedto a pneumatic pressure sensor (MPX5050DP, Motorola Inc., Denver,Colorado, USA), and (3) fixed-pinch: brief fixed-finger-position index–thumb pinches (minimal movement, fingers fixed to noncompliantforce sensor) against a rigid load cell (1042, Vishay Precision Group,Malvern, PA, USA). During squeeze task, acceleration was also recordedcorrespondingly to the free task.

The subjects were instructed to perform repetitive, self-paced handactions continuously for 3.5 min at comfortable but rather fast rate oftheir own preference (which turned out to be about 3 Hz) and at lowintensity (to avoid muscle fatigue). The order of the four tasks was ran-domized for each subject.

Measurements

MEGThe measurements were carried out at the MEG Core of Aalto

University. MEG signals recorded in a magnetically shielded room(Imedco AG, Hägendorf, Switzerland) with a 306-channel whole-scalpneuromagnetometer (Elekta Neuromag™, Elekta Oy, Helsinki, Finland).The recording passband was 0.1–330 Hz and the signals were sampledat 1 kHz. The subject's head position inside theMEG helmet was contin-uously monitored by feeding current to four head-tracking coils locatedon the scalp; the locations of the coilswith respect to anatomicalfiducialswere determined with an electromagnetic tracker (Fastrak, Polhemus,Colchester, VT, USA).

Peripheral signalsThe four peripheral signals (acceleration, pressure, force, and EMG)

were low-pass filtered at 330 Hz and sampled at 1 kHz, time-locked to

ch (right) hand-action tasks. MEG signal from a single gradiometer channel (filteredannels (raw), measured from the tip of right index finger during free task, and Euclidianeeze and fixed-pinch tasks respectively. EMG signals from flexor carpi radialismuscle was

85H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

MEG signals. Surface EMG was measured during all hand-action taskswith two EMG electrodes placed in bipolar configuration (20 mminter-electrode distance) over flexor carpi radialis muscle (impedanceb10 kΩ).

MRI3D-T1 magnetic resonance images (MRIs) were acquired with

whole-body General Electric Signa® 3.0TMRI scanner (Signa VH/i, Gen-eral Electric, Milwaukee, WI) at AMI Centre, Aalto University.

Data processing

PreprocessingContinuous MEG data were first preprocessed off-line using the

signal-space-separation (SSS) method to suppress external interfer-ences and to correct for head movements (Taulu et al., 2004). TheMEG data were band-pass filtered offline at 1–195 Hz. Offline filterswere 1–195 for acceleration, 1–45 Hz for pressure and force, and20–195 Hz for EMG signals. Pressure and force signals were filteredwith lower low-pass cut-off frequency due to their low-frequencycontent. High-pass cut-off frequency of 20 Hz was applied to EMGsignals to remove potential movement artifacts inherent to thehand-actions.

Coherence analysisTo perform frequency and coherence analyses, continuous data were

split into 2048 ms epochs with 1638-ms epoch overlap, leading to afrequency resolution of ~0.5 Hz (Bortel and Sovka, 2007). MEG epochswith magnetometer signals >3 pT or gradiometer signals >0.7 pT/cmwere excluded to avoid contamination by eye movements and blinks,muscle activity, or external MEG artifacts. Coherence analysis (Hallidayet al., 1995), yielding the cross-, power-, and coherence spectra, as wellas the cross-correlogram, was performed between MEG signals andEMG signal for all tasks, and between MEG and acceleration, pressure,or force signals depending of the task (free: accelerationfree and EMGfree,squeeze: pressure, accelerationsqueeze and EMGsqueeze and fixed-pinch:force and EMGfixed-pinch). Acceleration applied in free and squeeze tasks,was computed at every time step as the Euclidian norm of the threeorthogonal accelerometer channels. Rectified pressure and force signalswere applied in squeeze and fixed-pinch tasks, respectively. EMG wasalso rectified prior to coherence analysis. Before the coherence analysis,each epoch of acceleration, pressure, force, and EMG was normalizedby its Euclidian norm (Bourguignon et al., 2011; Brown et al., 1998;Pohja et al., 2005).

Source locationsCross-correlograms were band-pass filtered at 1–45 Hz. For one

subject, a strong artifact present in the MEG–force cross-correlogramwas removed by raising the low-pass filter limit to 2.5 Hz. Source anal-ysiswas performed in the timedomain, on the spatial distribution of thefiltered cross-correlogram, as previously done in corticomuscularcoherence and CKC studies (Bourguignon et al., 2011; Brown et al.,1998; Pohja et al., 2005). Individual MRIs were used to fit a sphericalheadmodel to the centroparietal brain region. Then, equivalent currentdipoles (ECDs) were estimated within the spherical head model at themain peak of the filtered cross-correlogram, using a selection of atleast 100 sensors that comprised all the most responsive sensors overthe hemisphere contralateral to the acting hand. Sources were consid-ered valid when the goodness-of-fit value exceeded 75% and the confi-dence volumewas below 500 mm3. The sources were visualized on thecoregistered individual MRIs.

To compare source coordinates across subjects, a non-linear trans-formation from individual MRIs to the standard Montreal NeurologicalInstitute (MNI) brain was first computed using the spatial normaliza-tion algorithm implemented in Statistical Parametric Mapping (SPM8,

Wellcome Department of Cognitive Neurology, London, UK) and thenapplied to source coordinates.

Statistical analysis

Hand-action characteristicsHand action regularity in time and amplitude domain, rate of the

hand actions, and the number of accepted trials used in the coherenceanalysis were compared between the tasks with a one-way, three-levels (tasks: free, squeeze, fixed-pinch) repeated-measures analysis ofvariance (ANOVA). Post-hoc paired t-tests with Holm–Bonferroni correc-tion for multiple comparisons were performed for all the significantANOVA effects. The regularity of the hand actions in time domain wasassessed with coefficient of variation of intervals between the successivehand actions. The regularity of hand action amplitude was assessed foracceleration (free), pressure (squeeze) and force (fixed-pinch) signalswith the coefficient of variation of the absolute peak values during thesuccessive hand actions.

Statistical significance of coherenceThe statistical significances of individual coherence levels were

assessed with surrogate data to overcome the multiple-comparisonissue. First, 1000 surrogate coherence spectra per each individual wereobtained by computing coherence between realMEG signals and Fouriertransform surrogate peripheral signals (acceleration, pressure, force,EMG); the Fourier transform surrogate imposes power spectrum to re-main the same as in the original signal but it replaces the phase of Fouriercoefficients by random numbers in the range [−π; π] (Faes et al., 2004).Then, a single maximum coherence value across a pre-selection of 18gradiometers covering the left rolandic area in the 1–10 Hz frequencyrange was extracted for each surrogate coherence spectrum; similararea selection has been previously used by Kim and Chung (2007) andwas chosen here as the maximum coherence was expected to occur atthis location (e.g. Bourguignon et al., 2011). Finally, the 0.95-percentileof this maximum coherence value yielded the coherence threshold ofpb0.05.

Strength of coherenceFrequencies of interest, showing consistent coherence across sub-

jects, were first identified (Bourguignon et al., 2011). Then, group-levelcomparison of the coherence strengths between MEG and each periph-eral signal was performed with a two-way 7 signals (accelerationfree,accelerationsqueeze, pressure, force, EMGsqueeze, EMGfree, EMGfixed-pinch)×number of frequencies of interest repeated-measures ANOVA, withHolm–Bonferroni correction for multiple comparisons. The dependentvariable was the maximal coherence value across the pre-selection of18 gradiometers covering the left rolandic area, independently for eachperipheral signal and frequency.

Source locationsThe differences between source locations (the ECDs) of the 7 hand-

action-related peripheral signals (accelerationfree, accelerationsqueeze,pressure, force, EMGfree, EMGsqueeze, EMGfixed-pinch) were assessed withnon-parametric permutation test (Nichols and Holmes, 2002). Thisapproach was preferred over parametric assessment for its ability todeal with the multiple comparison issue (i.e. one test for all 3 coordi-nates) and for its rotational invariance (i.e. it does not depend on the ar-bitrary coordinate system orientation). First, the meanMNI coordinates(across subjects) were computed for the sources of each peripheral sig-nal, yielding 7 source coordinates. Then, a χ parameter to quantifysource dispersion was computed as the root-mean-square of the dis-tances between all the possible 21 pairs of sources. The χ value is lowfor clustered sources, whereas a high χ value suggests differencebetween source locations. Under the null hypothesis that the sourcelocation is the same regardless of the task or peripheral signal, their la-beling is exchangeable prior χ computation (Nichols and Holmes,

86 H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

2002). To reject this null hypothesis and compute a significance thresh-old for the correctly labeled sources, the sample distribution of theχ value was computed from 100 000 different permutations. Each per-mutation consisted of shuffling all the 7 labels (accelerationfree,accelerationsqueeze, pressure, force, EMGfree, EMGsqueeze, EMGfixed-pinch) in-dependently for each subject— leading to (71)15 possible permutation(number of possible pairs of sources to the power of number of sub-jects). The χ threshold at pb0.05 was defined as the 95 percentile ofthe sample distribution (Nichols and Holmes, 2002).

Results

Hand-action-related peripheral signals showed synchronous oscil-lations with the hand actions. Fig. 1 shows 2-s epochs of the MEG, ac-celeration, pressure, force, and EMG signals from a representativesubject during each task. Clear sinusoidal oscillation of pressure andforce signals can be seen, but the oscillation pattern is more complexfor EMG and acceleration signals.

Hand action characteristicsAll 15 subjects performed the tasks without difficulties. The tasks

were performed at low intensity (mean±standard error of mean(SEM), peak magnitude (Euclidian norm) of acceleration during freetask 5.5±1.4 G, peak force during fixed-pinch task 0.51±0.09 N, andpeak pressure during squeeze task 0.05±0.03 Bar, corresponding toforce of 0.70±0.42 N applied on area of 1.5 cm2, comparable to thecontact of the finger tip to the elastic rubber ball).

The subjects were able tomaintain regular rate throughout each task,without clear breaks during the tasks. However, the regularity of boththe rate and amplitude of the hand-actions differed between the tasks(F2,28=7.82, p=0.001 for rate; F2,28=6.71, p=0.008 for amplitude).Coefficient of variation of the inter-action interval was higher duringfixed-pinch task (0.17±0.19) than free task (0.10±0.13, p=0.003),with no significant difference compared with the squeeze task (0.14±0.18). Coefficient of variation of the inter-action absolute peak amplitudewas lower for the pressure signal of the squeeze task (0.18±0.09) thanfor the acceleration signal of the free task (0.34±0.11, p=0.002) andthe force signal of the fixed-pinch task (0.31±0.16, p=0.024).Moreover,the tasks were performed at different rates (F2,28=20.36, pb0.001). Freetask was performed at faster rate (mean±SEM, 3.9±0.15 Hz) thansqueeze (3.3±0.15 Hz, p=0.006) and fixed-pinch (2.7±0.20 Hz,pb0.001) tasks, and the squeeze task was performed at faster rate thanthe fixed-pinch task (p=0.01). The number of accepted trials for coher-ence analysis did not differ (mean±SEM, free: 489±5.6 squeeze:484±14.9, fixed-pinch: 464±14.5).

CoherenceAll subjects displayed significant coherence (pb0.05) between MEG

and each peripheral signal, peaking either at the fundamental frequencyof hand action (F0), ranging from1.8 to 5.0 Hz, at itsfirst harmonic (F1),or both. Fig. 2 shows individual coherence spectra, superimposed for all

Fig. 2. Individual coherence spectra for each reference signal and all 15 subjects. Each tracefor a single subject. For each frequency bin, the coherence value displayed is the maximumhorizontal axis is the frequency in F0 units (i.e. 1 corresponds to F0 and 2 corresponds to F

subjects, between MEG and the peripheral signals. Although some sub-jects showed coherence at several harmonics, only F0 and F1 were in-cluded in the further analysis. The coherence reached significancelevel (pb0.05) in 201/210 cases; 15/15 cases for MEG-accelerationfree,15/15 cases for MEG–accelerationsqueeze, 15/15 cases for MEG–pressure,14/15 cases for MEG–force, 14/15 cases for MEG–EMGfree, 15/15 casesfor MEG–EMGsqueeze and 14/15 cases for MEG–EMGfixed-pinch, at F0, andin 14/15 cases for each MEG–peripheral signal pair at F1, except in 15/15 cases for MEG–accelerationsqueeze.

The strength of coherence varied between the peripheral signals(F6,84=7.73, pb0.001), but not between the F0 and F1 frequencies(F1,14=2.00, p=0.18) nor interaction (F6,84=1.45, p=0.25). Fig. 3 illus-trates the coherence level betweenMEG and each peripheral signal aver-aged over F0 and F1 frequencies. MEG coherence was stronger withpressure and force signals than with accelerationfree (pressure: p=0.0014, force: p=0.02), EMGfree (pressure: p=0.0013, force: p=0.036), and EMGfixed-pinch (pressure: p=0.012, force: p=0.011) signals.MEG coherence with accelerationfree signal was also stronger than withaccelerationfree (p=0.04). The coherence levels for MEG–pressure,MEG–force, MEG–EMGsqueeze and MEG–accelerationsqueeze did not differfrom each other, nor did the MEG–accelerationfree and the three MEG–EMGs. Table 1 shows strength of coherence for all sevenMEG–peripheralsignal pairs at F0 and F1.

Source locationsFig. 4 shows location of the subjects' individual coherent sources

for accelerationfree, pressure, force and EMGfree signals, and the respectivegroup means. For all subjects, hand actions and peripheral signals,the individual cross-correlograms showed a clear oscillatory patternin the contralateral rolandic sensors. Source analysis of the cross-correlograms identified sources in the contralateral hand area of theSM1 cortex, either in posterior bank of precentral gyrus or in anteriorbank of postcentral gyrus, with average inter-source distance (mean±SEM) of 9.4±0.8 mm (range 0.3–28.7 mm) between the individualsources. Table 2 shows the goodness-of-fit and the volume of confidenceof the ECDs.

At group level (averaged across subjects), the 7 sources were lo-cated close to each other, within 6 mm patch, with no significant dif-ference in the source locations (p=0.18). The respective MNIcoordinates were: MEG–accelerationfree [–37.8 –20.0 –51.2], MEG–accelerationfree [–45.0 –22.1 52.5], MEG–pressure [–41.8 –19.4 55.5],MEG–force [–41.4 –18.1 54.3], MEG–EMGfree [–37.5 –19.1 54.7], MEG–EMGsqueeze [–39.6 –18.9 55.2], MEG–EMGfixed-pinch [–40.2 –20.9 54.1].

Discussion

We found that pressure and force signals, in addition to previouslydescribed velocity, acceleration, and EMG signals (Bourguignon et al.,2011, 2013b; Jerbi et al., 2007; Pollok et al., 2004), are strongly coher-ent with MEG signals both during repetitive dynamic and fixed con-tractions at hand-action frequency. The coherent cortical sources for

represents the coherence between MEG and the hand-action-related peripheral signalcoherence across the pre-selected 18 gradiometers covering the left rolandic area. The1). Gray horizontal line shows the threshold for statistical significance (pb0.05).

Fig. 3. Mean±SEM coherence levels between MEG and acceleration, pressure, force,and EMG signals in the three tasks. The coherence levels are averaged across F0 andF1, and are displayed for each signal separately. **pb0.01, *pb0.05 significant differ-ence with respect to MEG–pressure coherence. †pb0.05 significant difference with re-spect to MEG–force coherence. #pb0.05 significant difference with respect to MEG–accelerationsqueeze coherence.

87H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

each task and the respective hand-action-related peripheral signalswere located close to each other at the contralateral hand area ofthe SM1 cortex, with no significant differences between the sources.The source locations agree with the source locations reported in studieson CKC betweenMEG and acceleration (Bourguignon et al., 2011, 2012)or velocity (Jerbi et al., 2007), as well as on corticomuscular coherencebetweenMEGand EMG (Salenius et al., 1997) and corticomuscular cou-pling at tremor frequency (Pollok et al., 2004). Furthermore, the coher-ence with respect to MEG was stronger with pressure and force signalsthan with acceleration and EMG signals in most of the hand-actiontasks.

Coupling between MEG and hand-action-related peripheral signals

In each subject, all measured hand-action-related peripheral signalswere coupled with MEG signals either at F0, F1, or both. The coherencebetween MEG and the peripheral signals could reflect (1) coupling be-tween generation of motor command at the M1 cortex and the conse-quent motor output and/or (2) coupling between somatosensoryinput to the SM1 cortex and the hand action.

In both cases, subtle variability in the performance of the handactions (e.g. variation in rate and amplitude) could reduce the coher-ence level. Indeed, the peak amplitude of the peripheral signal wasabout 30–40% more stable in the squeeze task (pressure) than in thefree (acceleration) and fixed-pinch (force) tasks. This finding could part-ly explain the weaker coherence between MEG and acceleration than

Table 1Strength of coherence at F0 and F1.

F0 F1

Signal Mean±SEM Range Mean±SEM Range

Accelerationfree 0.28±0.04 0.06–0.61 0.21±0.04 0.04–0.47Accelerationsqueeze 0.41±0.04 0.13–0.64 0.29±0.04 0.14–0.69Pressuresqueeze 0.45±0.04 0.09–0.64 0.36±0.05 0.04–0.63Forcefixed-pinch 0.38±0.04 0.07–0.66 0.38±0.06 0.06–0.73EMGfree 0.23±0.03 0.07–0.46 0.28±0.05 0.07–0.67EMGsqueeze 0.31±0.04 0.02–0.58 0.33±0.03 0.05–0.56EMGfixed-pinch 0.31±0.04 0.05–0.58 0.27±0.04 0.04–0.54

betweenMEG and pressure. However, all the applied peripheral signalswere normalizedwith their Euclidian norm in the coherence analysis todecrease effect of inter-epoch amplitude variation. This procedurereduces the amplitude variation, but some intra-epoch variation willremain. Nevertheless, the clearly reduced stability in hand-action am-plitude of the free and fixed-pinch tasks very likely caused variation inthe cortical motor output and sensory afferent input and thus, couldhave reduced the physiological coupling between periphery and thebrain.

The rate stability, on the contrary, was the highest in the free task,which on the other hand resulted in the weakest coherence. Moreover,the mean coherence between MEG and acceleration was about 19%(p=0.002) stronger during the squeeze task than during the free task.These results suggest that both the task and the stability of thehand-action amplitude affect the coherence strength, whereas theeffect of stability of the hand-action rate is minor. Our subjects, whowere instructed to perform the self-paced hand actions fast but at com-fortable rate, reported the squeeze task as the most comfortable to per-form. Therefore, comfortable tasks with stable movement amplitudemay enhance the brain–hand synchronization. However, further stud-ies are needed to fully clarify the effect of the stability of hand-actionrate and amplitude on CKC.

Largemovements in the joints of the handwere not crucial for the co-herence with MEG as the repetitive fixed pinches, with only small inev-itable cyclic movements in the joints of index and thumb, resulted insignificant coherence. The coherence was even stronger during therepetitive fixed pinches (MEG–force) than during the free dynamichand movements (MEG–acceleration). During the repetitive fixed-pinches, in addition to some inevitable joint movements, the musclelength is also changing due to cyclic oscillation in contraction forcebecause of elongation of tendon and connective tissue (Hodgson et al.,2006). Such subtle changes in muscle length are sufficient to activate ine.g. muscle spindles receptors that are extremely sensitive to smalllength changes (as low as 5 μm during vibration) of their parent muscle(Brown et al., 1967). Mechanoreceptors of the skin were also activatedduring the repetitive fixed-pinches as e.g. Pacinian corpuscles are capa-ble to detect tiny 10 nm skinmotions (Brisben et al., 1999) andMeissnercorpuscles respond to sudden forces as low as 0.5 N acting on the skin ofthe fingers (Macefield et al., 1996). Therefore, if the coherence is drivenby somatosensory input to the SM1 cortex, coherence between MEGand hand-action-related signals can be expected during all kinds of re-petitive hand actions, including repetitive isometric contractions.

The differences in coherence level could be explained by propertiesof each peripheral signal. At hand-action frequencies, the measured ac-celeration, pressure, force, and EMG are closely related to each other,but with some differences. For the fixed-pinch and squeeze tasks therewas a single well–defined sinusoidal oscillation visible in the forceand pressure signals for each contraction cycle. In contrast, for EMGand acceleration, the hand action related features exhibited a morecomplex oscillatory pattern, although distinguishable repetitive fea-tures at hand action frequencies were present. Therefore, the forceand pressure signals may provide better estimation of the time coursesof the motor and sensory events responsible for the MEG coherencethan acceleration and EMG signals. In addition to the properties of thehand-action-related signals, the task affected the coherence strength,as indicated by (1) the stronger MEG–acceleration coherence in thesqueeze task than in the free task, and (2) the comparable strengthsfor MEG–pressure, MEG–EMG and MEG–acceleration coherences inthe squeeze task.

It is important to note thatwe did not observe significantMEG–EMGcoherence at 15–40 Hz typical for the well-known corticomuscular co-herence (Conway et al., 1995; Mima and Hallett, 1999; Salenius et al.,1997). This result is in line with the observation that corticomuscularcoherence vanishes during movements with respect to the steady iso-metric hold period (Kilner et al., 1999); indeed, all current tasks in-volved dynamic elongations and shortenings of muscles and tendons

Fig. 4. Source locations based on the cross-correlograms for acceleration and EMG in free task, pressure in squeeze task, and force in fixed-pinch task for each subject (three upperrows) superimposed on individual MRI in transverse plane. Mean source locations (averaged across the 15 subjects, lowest row) were within 6-mm patch at the “hand knob”, andare superimposed on transverse, sagittal, and coronal planes of the MNI brain.

Table 2Source parameters based on the cross-correlograms in each experimental condition.

Signal Goodness-of-fit (%) Confidence volume(mm3)

Mean Range Mean Range

Accelerationfree 87.1 78.8–97.4 47 5–149Accelerationsqueeze 92.2 86.8–97.6 11 1–41Pressuresqueeze 91.1 75.9–97.5 8 1–39Forcefixed-pinch 90.6 83.9–95.4 12 1–54EMGfree 89.0 76.2–95.6 37 2–136EMGsqueeze 91.3 82.4–96.7 16 2–48EMGfixed-pinch 91.5 85.0–95.5 60 1–399

88 H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

along with movements and the changes of the contraction force whichexplains the absence of corticomuscular coherence at 15–40 Hz.

We suggest that the coupling between MEG and EMG at the funda-mental frequency (1.8–5.0 Hz) of the hand actions has different mech-anisms than the corticomuscular coherence observed during steadyisometric contraction, although in both cases the hand area of theSM1 cortex is involved. CKC may reflect primarily the afferent sensoryinput from moving hand and its muscles to the cortex, as is suggestedby our recent observation of strong CKC during passive finger move-ments (Piitulainen et al., 2013).

Corticomuscular coherence is typically strongest during sustained iso-metric contractions (Conway et al., 1995; Kilner et al., 1999; Salenius

89H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

et al., 1997), and is observed between ~15 and 40 Hz depending on theexerted force level (Hari and Salenius, 1999; Mima and Hallett, 1999).Corticomuscular coherence has been suggested tomainly reflectmodula-tion of efferent population-level firing from the M1 cortex to the motorunits of the muscle(s) (Baker et al., 1997; Salenius and Hari, 2003), al-though afferent sensory feedback from themuscle to the central nervoussystem may also contribute to the corticomuscular coherence (Baker,2007). MEG–EMG coherence at hand-action frequency and its first har-monic likely have somewhat different neuronal origin. In line with thecurrent results, cerebromuscular coupling has been detected at tremorfrequency (~3.9 Hz) and its first harmonic (~8.1 Hz) during imitationof parkinsonian tremor in healthy subjects (Pollok et al., 2004), at move-ment frequency (~1.4 Hz) and its first harmonic (~2.5 Hz) during audi-torily paced repetitive finger movements (Pollok et al., 2005a,b),at movement frequency (~0.5 Hz) during repetitive wrist flexion-extensions, and at low frequencies (~5 Hz) during phasic dynamicmovements (inter-movement time 8–25 s) (Feige et al., 2000). These ob-servationsmay have similar sensorimotor origin as suggested for the cur-rent coherence between MEG and the hand-action-related peripheralsignals. Thus, the EMG signal seems to provide a good estimation of thetime courses of themotor and sensory events responsible for theMEG co-herence as is the case for acceleration, force, and pressure signals.

For the current fast repetitive hand actions, we were not able to de-termine consistent delays between the hand-action-related peripheralsignals and the MEG signals and thus could not estimate consistent la-tencies for the sensory input to cortex and/or the cortical motor outputto the muscle. Thus we were not able to pinpoint the relative contribu-tions of afferent and efferent pathways to the observed coherence.However, our recent study showed strong CKC during passive fingermovements (Piitulainen et al., 2013) suggesting that CKC may reflectprimarily the afferent sensory input from the periphery to the cortex.

Locations of cortical sources of coherent activity

The magnetic field patterns were adequately explained by singlecurrent dipoles for all subjects, tasks, and peripheral signals (seeTable 2). At the level of individual subjects, the cortical sources of coher-ent activity were typically located at or close to the “hand knob” of theM1 cortex (Yousry et al., 1997) or slightly posterior to central sulcus atthe primary sensory (S1) cortex. At group level, the sources were locat-ed at the same regions in theMNI brain. TheM1 cortex seems to have animportant role in the generation of the coherent activity. In monkeys,neuronal spiking and low-frequency (b4 Hz) local field potentialsrecorded from the M1 and premotor cortices are synchronized withthe reach and grasping kinematics (Bansal et al., 2011). Therefore, thelow-frequency cyclic M1 cortex activity could potentially drive the co-herence during repetitive hand-actions.

The M1 cortex is important in the generation of the motor com-mands, but it also receives proprioceptive input with similar short laten-cies as the S1 cortex does (Devanandan and Heath, 1975; Lucier et al.,1975). Therefore, although the coherent sources were partly located atM1 cortex, they can also reflect sensory feedback from themoving hand.

Implications for functional mapping of SM1 cortex

To date, functional magnetic resonance imaging (fMRI) has been themain tool in the non-invasive presurgical evaluation of the SM1 cortex(Bartsch et al., 2006;De Tiège et al., 2009). Unfortunately, interpretationof fMRI maps is challenging in patients with altered neurovascular cou-pling caused by various brain disorders (Bartsch et al., 2006; D'Espositoet al., 2003; Korvenoja et al., 2006; Krings et al., 2001). In such patients,MEGmay represent an alternative to fMRI as it provides direct informa-tion about neuronal activity. MEG may be superior to fMRI in some pa-tients with space-occupying lesions for central sulcus identification(Korvenoja et al., 2006; Mäkelä et al., 2006).

The strong coherence and nearly 100% success rate of the SM1 cor-tex localization among the subjects in the current and preceding CKCstudies (Bourguignon et al., 2011, 2012) indicate that coherence analy-sis between MEG and various hand-action-related signals has a strongpotential in functionalmapping of the SM1 cortex. For instance, the suc-cess rate of significant corticomuscular coherence during stationary iso-metric contractions is less than 80% (Pohja et al., 2005).

The currently studied tasks and peripheral signalswere all highly cor-related with the hand actions, and thus can serve as potential tools forfunctionalmapping of the SM1 cortexwith EEG/MEG.However, their ap-plicability may vary. The squeeze task with pressure monitoring provid-ed the most straightforward and comfortable protocol for hand-areamapping. The pressure sensors can be readily made MEG-compatible,and the same is true for the force transducers used in the fixed-pinchtask; however, the subjects considered the fixed-pinch task awkward toperform. Acceleration sensors are easy to use and they do not restrictthe choice of dynamic movements. However, the task and reference sig-nal need to be selected based on the special characteristics of the patient,and further studies are needed to confirm the applicability of CKC forfunctionalmapping of the SM1 cortex in different patient groups. Finally,for a consistent identification of the SM1 cortex with MEG in individualpatients, a multimodal approach relying on corticokinematic coherence,corticomuscular coherence, and somatosensory evoked fields would bepreferred.

Conclusions

MEG signals are strongly coupled with hand-action-related accel-eration, force, pressure, and surface EMG during dynamic and repeti-tive fixed contractions. Coherence and source analysis based on thesesignals proved them all to have similar cortical origin at the SM1 cor-tex. All these signals can potentially be used in functional mapping ofthe human SM1 cortex.

Acknowledgments

This study has been supported by the Academy of Finland (NationalCenters of Excellence Program 2006–2011 #129678, #131483), by theSalWe Research Program for Mind and Body (Tekes — the FinnishFunding Agency for Technology and Innovation grant 1104/10), theERC Advanced Grant (#232946 to Riitta Hari) and the “Brains Back toBrussels” grant to Veikko Jousmäki from the Institut d'Encouragementde la Recherche Scientifique et de l'Innovation de Bruxelles (Brussels,Belgium), the Fonds de la Recherche Scientifique (FRS-FNRS, Belgium,Research Convention 3.4611.08). Mathieu Bourguignon benefits of aresearch grant from the FRIA (FRS-FNRS, Belgium). Xavier De Tiège isClinicien-Chercheur Spécialiste at the FRS-FNRS, Belgium. We thankHelge Kainulainen and Ronny Schreiber at the Brain Research Unit(Aalto University School of Science, Espoo, Finland) for technicalsupport.

References

Baker, S.N., 2007. Oscillatory interactions between sensorimotor cortex and the periph-ery. Curr. Opin. Neurobiol. 17, 649–655.

Baker, S.N., Olivier, E., Lemon, R.N., 1997. Coherent oscillations in monkey motor cortexand hand muscle EMG show task-dependent modulation. J. Physiol. 501, 225–241.

Bansal, A.K., Vargas-Irwin, C.E., Truccolo,W., Donoghue, J.P., 2011. Relationships among low-frequency local field potentials, spiking activity, and three-dimensional reach and graspkinematics in primary motor and ventral premotor cortices. J. Neurophysiol. 105,1603–1619.

Bartsch, A.J., Homola, G., Biller, A., Solymosi, L., Bendszus, M., 2006. Diagnostic functionalMRI: illustrated clinical applications and decision-making. J. Magn. Reson. Imaging23, 921–932.

Bortel, R., Sovka, P., 2007. Approximation of statistical distribution of magnitude squaredcoherence estimated with segment overlapping. Signal Process. 87, 1100–1117.

Bourguignon, M., De Tiège, X., Op de Beeck, M., Pirotte, B., Van Bogaert, P., Goldman, S.,Hari, R., Jousmäki, V., 2011. Functional motor-cortex mapping using corticokinematiccoherence. NeuroImage 55, 1475–1479.

90 H. Piitulainen et al. / NeuroImage 72 (2013) 83–90

Bourguignon, M., Jousmäki, V., Op de Beeck, M., Van Bogaert, P., Goldman, S., De Tiège,X., 2012. Neuronal network coherent with hand kinematics during fast repetitivehand movements. NeuroImage 59, 1684–1691.

Bourguignon, M., De Tiège, X., Op de Beeck, M., Van Bogaert, P., Goldman, S., Jousmäki,V., Hari, R., 2013a. Primary motor cortex and cerebellum are coupled with the kine-matics of observed hand movements. NeuroImage 66, 500–507.

Bourguignon, M., De Tiège, X., Op de Beeck, M., Ligot, N., Paquier, P., Van Bogaert, P.,Goldman, S., Hari, R., Jousmäki, V., 2013b. The pace of prosodic phrasing couplesthe listener's cortex to the reader's voice. Hum. Brain Mapp. 34, 314–326.

Brisben, A.J., Hsiao, S.S., Johnson, K.O., 1999. Detection of vibration transmitted throughan object grasped in the hand. J. Neurophysiol. 81, 1548–1558.

Brown, M.C., Engberg, I., Matthews, P.B., 1967. The relative sensitivity to vibration ofmuscle receptors of the cat. J. Physiol. 192, 773–800.

Brown, P., Salenius, S., Rothwell, J.C., Hari, R., 1998. Cortical correlate of the piperrhythm in humans. J. Neurophysiol. 80, 2911–2917.

Conway, B.A., Halliday, D.M., Farmer, S.F., Shahani, U., Maas, P., Weir, A.I., Rosenberg, J.R.,1995. Synchronization between motor cortex and spinal motoneuronal pool duringthe performance of a maintained motor task in man. J. Physiol. 489, 917–924.

De Tiège, X., Connelly, A., Liegeois, F., Harkness, W., Clark, C.A., Chong, W.K., Gadian,D.G., Cross, J.H., 2009. Influence of motor functional magnetic resonance imagingon the surgical management of children and adolescents with symptomatic focalepilepsy. Neurosurgery 64, 856–864.

D'Esposito, M., Deouell, L.Y., Gazzaley, A., 2003. Alterations in the BOLD fMRI signal withageing and disease: a challenge for neuroimaging. Nat. Rev. Neurosci. 4, 863–872.

Devanandan, M.S., Heath, P.D., 1975. Proceedings: a short latency pathway from fore-arm nerves to area 4 of the baboon's cerebral cortex. J. Physiol. 248, 43P–44P.

Faes, L., Pinna, G.D., Porta, A., Maestri, R., Nollo, G., 2004. Surrogate data analysis for assessingthe significance of the coherence function. IEEE Trans. Biomed. Eng. 51, 1156–1166.

Feige, B., Aertsen, A., Kristeva-Feige, R., 2000. Dynamic synchronization between mul-tiple cortical motor areas and muscle activity in phasic voluntary movements.J. Neurophysiol. 84, 2622–2629.

Halliday, D.M., Rosenberg, J.R., Amjad, A.M., Breeze, P., Conway, B.A., Farmer, S.F., 1995.A framework for the analysis of mixed time series/point process data-theory andapplication to the study of physiological tremor, single motor unit dischargesand electromyograms. Prog. Biophys. Mol. Biol. 64, 237–278.

Hari, R., Salenius, S., 1999. Rhythmical corticomotor communication. NeuroReport 10,R1–R10.

Hodgson, J.A., Finni, T., Lai, A.M., Edgerton, V.R., Sinha, S., 2006. Influence of structureon the tissue dynamics of the human soleus muscle observed in MRI studies duringisometric contractions. J. Morphol. 267, 584–601.

Jerbi, K., Lachaux, J.P., N'Diaye, K., Pantazis, D., Leahy, R.M., Garnero, L., Baillet, S., 2007.Coherent neural representation of hand speed in humans revealed by MEG imag-ing. Proc. Natl. Acad. Sci. U. S. A. 104, 7676–7681.

Kilner, J.M., Baker, S.N., Salenius, S., Jousmäki, V., Hari, R., Lemon, R.N., 1999. Task-dependentmodulation of 15–30 Hz coherence between rectified EMGs from human hand andforearm muscles. J. Physiol. 516, 559–570.

Kim, J.S., Chung, C.K., 2007. Robust source analysis of oscillatory motor cortex activitywith inherently variable phase delay. NeuroImage 37, 518–529.

Korvenoja, A., Kirveskari, E., Aronen, H.J., Avikainen, S., Brander, A., Huttunen, J., Ilmoniemi,R.J., Jääskeläinen, J.E., Kovala, T., Mäkelä, J.P., et al., 2006. Sensorimotor cortex localiza-tion: Comparison of magnetoencephalography, functional MR imaging, andintraoperative cortical mapping. Radiology 241, 213–222.

Krings, T., Reinges, M.H., Erberich, S., Kemeny, S., Rohde, V., Spetzger, U., Korinth, M.,Willmes, K., Gilsbach, J.M., Thron, A., 2001. Functional MRI for presurgical planning:problems, artefacts, and solution strategies. J. Neurol. Neurosurg. Psychiatry 70,749–760.

Lucier, G.E., Ruegg, D.C., Wiesendanger, M., 1975. Responses of neurones in motor cor-tex and in area 3A to controlled stretches of forelimb muscles in cebus monkeys.J. Physiol. 251, 833–853.

Macefield, V.G., Hager-Ross, C., Johansson, R.S., 1996. Control of grip force during restraintof an object held between finger and thumb: responses of cutaneous afferents fromthe digits. Exp. Brain Res. 108, 155–171.

Mäkelä, J.P., Forss, N., Jääskeläinen, J., Kirveskari, E., Korvenoja, A., Paetau, R., 2006.Magnetoencephalography in neurosurgery. Neurosurgery 59, 493–510.

Mima, T., Hallett, M., 1999. Corticomuscular coherence: a review. J. Clin. Neurophysiol.16, 501–511.

Murayama, N., Lin, Y.Y., Salenius, S., Hari, R., 2001. Oscillatory interaction between humanmotor cortex and trunk muscles during isometric contraction. NeuroImage 14,1206–1213.

Nichols, T.E., Holmes, A.P., 2002. Nonparametric permutation tests for functional neu-roimaging: a primer with examples. Hum. Brain Mapp. 15, 1–25.

Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inven-tory. Neuropsychologia 9, 97–113.

Piitulainen, H., Bourguignon,M., De Tiège, X., Hari, R., Jousmäki, V., 2013. Corticokinematiccoherence during active and passive finger movements. Neuroscience, 2013, in press.

Pohja, M., Salenius, S., Hari, R., 2005. Reproducibility of cortex-muscle coherence.NeuroImage 26, 764–770.

Pollok, B., Gross, J., Dirks, M., Timmermann, L., Schnitzler, A., 2004. The cerebral oscilla-tory network of voluntary tremor. J. Physiol. 554, 871–878.

Pollok, B., Sudmeyer, M., Gross, J., Schnitzler, A., 2005a. The oscillatory network of sim-ple repetitive bimanual movements. Brain Res. Cogn. Brain Res. 25, 300–311.

Pollok, B., Gross, J., Müller, K., Aschersleben, G., Schnitzler, A., 2005b. The cerebral oscil-latory network associated with auditorily paced finger movements. NeuroImage24, 646–655.

Salenius, S., Hari, R., 2003. Synchronous cortical oscillatory activity during motor ac-tion. Curr. Opin. Neurobiol. 13, 678–684.

Salenius, S., Salmelin, R., Neuper, C., Pfurtscheller, G., Hari, R., 1996. Human cortical40 Hz rhythm is closely related to EMG rhythmicity. Neurosci. Lett. 213, 75–78.

Salenius, S., Portin, K., Kajola, M., Salmelin, R., Hari, R., 1997. Cortical control ofhuman motoneuron firing during isometric contraction. J. Neurophysiol. 77,3401–3405.

Taulu, S., Kajola, M., Simola, J., 2004. Suppression of interference and artifacts by thesignal space separation method. Brain Topogr. 16, 269–275.

Yousry, T.A., Schmid, U.D., Alkadhi, H., Schmidt, D., Peraud, A., Buettner, A., Winkler, P.,1997. Localization of the motor hand area to a knob on the precentral gyrus. A newlandmark. Brain 120 (Pt 1), 141–157.