2.1.

Subjects

Twenty right-handed healthy subjects (five females, 21 to 40 years old; mean age 27.7) without any history of neurological or psychiatric disorders were investigated. Handedness was assessed by the German version of the Edinburgh Handedness Inventory.²⁵ The participants were employees or students of the Charité and participated voluntarily in the experiment with financial compensation. The study was approved by the Ethics Committee of the Charité—Universitätsmedizin Berlin, and all subjects gave informed consent prior to the experiment.

2.2.

Experimental Design

Mirror therapy has been applied using different setups. Originally a “mirror box” was used, which provides the visual image of two hands moving simultaneously.³ Subsequent studies mainly employed a real mirror located at the patients’ midsagittal plane.¹⁵ In order to allow a systematic and randomized variation of the seen and moved hand without changing other cofactors, we implemented a mirror-like paradigm in accordance with the previous study of Dohle et al.¹⁰ by means of a video chain (Fig. 1). The lateralized cerebral activation elicited by this setup was not shown to be confounded by hemifield stimulation.¹⁹ Subjects had to either hold their hand static or perform a finger-thumb opposition movement sequence under visual control. Subjects could not see their hand directly, as it was covered by a black paperboard in front of them, but viewed it via a video chain. The subject’s hand was filmed online with a video webcam (Logitech Webcam Pro 9000). By means of a software package (Logitech Webcam Software v1.1, frame rate: 50 Hz), the image was displayed in real time on a screen (Acer, $1280\times 1024$ pixel, frame rate: 60 Hz) in front of the subjects with 0-deg eccentricity. Thus, the delay was maximal 20 ms which is well below the threshold of subjective awareness.²⁶

Fig. 1

Photography of the experimental setup: (1) Subject wearing an EEG cap with functional near-infrared spectroscopy (fNIRS) detectors and sources; (2) subject’s hand inside the black paperboard covered by a black drapery being filmed by (3) a web-cam inside the black paperboard; (4) visual feedback the subject gets of his hand, in this case mirror-like illusion (MIR) on (5) a precuneus (PC) monitor. Not in the picture: control PCs.

The distance between subject and screen was approximately 55 cm. The size of the outer frame was $15\times 20\mathrm{cm}$ ($640\times 480$ pixel). The size of the online film of the hand was individually different, because we adopted this to the real size of the subject’s hand, but for all subjects it fitted into the outer frame, which means the stimulus was not bigger than $15\times 20\mathrm{cm}$, i.e., not bigger than 21-deg horizontal and 16-deg vertical visual angle.

There were two movement tasks: The first one (“movement”) consisted of moving the thumb and the index finger sequentially at three different opposition distances. The opposition distances reached from just not touching both fingers to maximal comfortable distance and movement sequences were performed at an individually chosen rate of approximately 1 Hz. The second task (“static”) consisted of holding the hand in a static posture with the thumb and index finger opposing each other at a distance of a few millimeters. Both finger movement tasks were demonstrated and explained verbally by the investigator. During these tasks, the filmed image of the hand could be inverted horizontally in such a way that the subjects’ right hand appeared as if it was their left hand and vice versa (“mirror”). With the horizontal inversion of the video image, either normal visual feedback (NOR) or a MIR could be induced. This results in the following four randomized presented conditions:

As previous studies showed no effect of the MIR neither on eye movements¹⁰ nor on muscle activity, as assessed with electromyography,^18,27 these parameters were not recorded in this experiment.

For each hand, a total number of 80 trials under these four conditions were arranged in randomized sequences. Each trial was initiated by an acoustic cue (“start” or “stop”) and lasted 10 s followed by a rest period with an average duration of 15 s (jittered from 10 to 20 s), where only a black screen was shown. This rest period served as a baseline for the correction in the analysis. Each condition was presented 20 times, yielding in a total experimental time of 38 min per hand. Each hand was examined separately, and the whole experiment lasted approximately an hour and a half. During the whole experiment, the subjects were instructed to watch the screen, even during the rest period and to focus the hand during the active tasks (“static” and “movement”). The subjects’ adherence to the task was ensured by the investigator, who was present during the entire experiment.

By comparing corresponding channels in both hemispheres, a $2\times 2\times 2\times 2$ study design with the four factors movement (movement/static), mirror (MIR/NOR), hand (left/right), and hemisphere (ipsilateral/contralateral to the moving hand) was applied, resulting in 16 different conditions.

2.3.

fNIRS Data Acquisition

During the experiment, the blood oxygenation at the surface of the subjects’ brain was measured with a fNIRS system which offers up to 16 detectors and 16 emitters (NIRScout 16-16, NIRx Medizintechnik GmbH, Berlin, Germany) at two wavelengths (850 and 760 nm). Based on previous findings on the role of the PC during movement mirroring,^10,11 we chose fiber optode positions to cover the bilateral occipito-parietal and precentral areas of the subject’s head, providing a total of 38 useful channels where source and detector were placed at a distance of between 2.5 and 3 cm from each other. This arrangement is shown in Fig. 2. To guarantee optimal safety in optode localization and convenience for the subjects, the emitters and detectors were integrated into a commercially available EEG cap ( www.easycap.de) with 128 possible positions. fNIRS data were continuously sampled at 3.13 Hz.

Fig. 2

Display of optode positions, measurement channels (numbered in green), including main EEG positions from the standard 10–20 system (white labels). A number of fNIRS sources are at the same location as main standard EEG position.

2.4.

Preprocessing of fNIRS Data

Data analysis was performed using software routines employing the software packet Matlab (Mathworks Inc., Natick, Massachusetts, version 7.5.0.342 R2007b). fNIRS data were corrected for movement artifacts by a semiautomated approach, which replaces contaminated data segments by linear interpolation.²⁸ Subsequently, attenuation changes of both wavelengths were transformed to concentration changes of oxy- and deoxygenated hemoglobin (HbO and HbR) using a modified Beer–Lambert law [differential pathlength factors: 5.98 (higher wavelength: 850 nm), 7.15 (lower wavelength: 760 nm), extinction coefficients for HbO 2.53/1.49 (higher/lower wavelength) and HbR 1.80/3.84 (higher/lower wavelength), and an interoptode-distance of 3 cm].²⁹ Data were then band-pass filtered between 0.2 and 0.016 Hz (using a 3rd order Butterworth filter) to attenuate for heartbeat, breathing-related changes, and drifts.

For effect estimation, a general linear model (GLM) was run in all subjects. Regressors were composed by convolution of stimulus onset and length with a hemodynamic response function.³⁰ For each subject, condition and channel, the GLM provided two beta values (for HbO and HbR, respectively), indicating the strength of the modulation of the hemodynamic response. It should be noted that the stronger activity is indicated by positive beta values for HbO and by negative beta values for HbR.^31,32

2.5.

Definition of the Regions of Interest

For visual inspection and demonstration, we employed the freeware MATLAB toolbox NFRI ( http://brain.job.affrc.go.jp/tools/) described by Singh et al.,³³ which takes EEG 10 to 20 positions as references to estimate the brain regions underlying the channel locations. This toolbox also enables statistical results for each channel to be plotted on the surface of a schematic brain (already used for Fig. 2).

Regions of interest (ROI) were defined corresponding to the activation foci of the above-mentioned studies: in the occipito-parietal cortex (PC), the three channels displayed in Fig. 3(a), right sides were chosen (further referred to as “PC-ROI”) for both hemispheres [i.e., for the left hemisphere (LH): channel numbers 23, 24, and 31; for the right hemisphere (RH): channel numbers 25, 26, and 32; see also Fig. 2]. The M1-ROI was defined as channels around EEG positions C3 and C4, known to cover the precentral regions of the brain.³⁴ Therefore, the four midsagittal channels on either hemisphere were selected (i.e., for the LH: channel numbers 1, 2, 7, and 8; for the RH: channel numbers 3, 4, 9, and 10; see also Fig. 2). This selection was in agreement with the analysis of statistical effect sizes (see Appendix).

Fig. 3

Left: Activation foci from previous imaging studies projected onto the surface of a standardized brain for (a) mirror effect in occipito-parietal area, and (b) movement effect in precentral area. Right: Channels chosen as regions of interest (ROI) for further statistical analysis in the area of the PC (a) and precentral areas (b).

2.6.

Time Courses

For both ROIs, baseline-corrected time courses were calculated averaged across subjects and smoothed (moving window of 10 s) for the MIR and NOR conditions separately.

2.7.

Statistical Analysis

Statistical analysis was performed with the software packet PASW Statistics (SPSS Inc., Chicago, Illinois, version 18.0.0). For statistical group analysis, a repeated measure four-way analysis of variance (ANOVA) with the factors movement (movement/static), mirror (MIR/NOR), hand (left/right), and hemisphere (ipsilateral/contralateral) was conducted for the mean beta values of the ROIs. As M1 activation is much greater in the movement condition compared with the static condition,¹⁰ a further three-way repeated measures ANOVA with the factors mirror (MIR/NOR), hand (left/right), and hemisphere (ipsilateral/contralateral) was applied on the mean beta values of the M1-ROI for the trials with movement only. Level of significance was always set at $p<0.05$. The ANOVA analysis was not corrected for multiple-comparison. Post-hoc analysis was performed using paired $t$-tests for all significant interactions of the ANOVAs. As the ANOVA is classified as a so called Omnibus test, the post-hoc tests were not Bonferroni corrected.³⁵

3.1.

Effects at the PC-ROI

Baseline-corrected time courses for the PC-ROI are shown in Fig. 4, upper panel. In the hemisphere ipsilateral to the moving hand, we observed a stronger change in concentration of both HbO and HbR (Fig. 4, left side) in the MIR condition compared with the NOR condition. This difference peaks at about 10 s after stimulus onset. In contrast, in the hemisphere contralateral to the moving hand (Fig. 4, right side), there is a less concentration change in the MIR compared with the NOR condition for both wavelengths.

Fig. 4

Time courses averaged over all subjects ($n=20$) of the PC- (a, b) and M1-ROIs (c, d), contra- (a, c) and ipsilateral (b, d) to the moving hand. The duration of stimulus presentation is shaded in gray. HbO is presented in red, while HbR is coded in blue. Time courses have error bars with standard error of the mean.

Mean beta values for the PC-ROI are shown in Fig. 5(a)–5(c), upper panel and in Table 1. Mean beta values indicated a stronger activation in the MIR than in the NOR condition ipsilateral. In contrast, activation was reduced in the MIR as compared with the NOR condition contralateral. Mean beta values for HbR indicate the same as for HbO, respectively. Figure 5(c) also displays the mean interhemispheric difference between the ipsi- and contralateral hemisphere for MIR and NOR conditions.

Fig. 5

Mean beta values and differences, and their corresponding standard errors at the PC-ROI (top row, a–c) and M1-ROI (bottom row, d–f). Left column ($a+d$): mean beta values for HbO of the ipsi- and contralateral hemisphere for the MIR and NOR conditions; middle column ($b+e$): mean beta values for HbR of the ipsi- and contralateral hemisphere for the MIR and NOR condition; right column ($c+f$): mean interhemispheric differences in the MIR and NOR conditions for HbO and HbR (multiplied by $-1$). Note that, in the figures of the right column, values $>0$ indicate activation in the contralateral $>$ ipsilateral hemisphere, and values $<0$ indicate activation ipsilateral $>$ contralateral.

Table 1

Mean beta values and standard errors (in brackets) for PC-ROI (top rows) and M1-ROI (bottom rows). Left column: mean beta values for HbO. Right column: mean beta values for HbR.

The repeated measures four-way ANOVA over the mean beta values of the PC-ROI showed a significant main effect of the factor movement [$F(1,19)=4.92$; $p<0.05$, ${\eta }^2=0.21$] for HbO. The mean beta values indicate more activation in movement conditions as compared with static conditions (Table 1).

Furthermore, the ANOVA showed a significant two-way interaction of the factors $\text{mirror}\times \text{hemisphere}$ [$F(1,19)=4.59$; $p<0.05$, ${\eta }^2=0.20$] for HbO. Post-hoc analysis of this two-way interaction revealed no significant effect in the paired $t$-tests. This two-way interaction of the factors $\text{mirror}\times \text{hemisphere}$ was similar for HbR, but failed to reach significance [$F(1,19)=1.71$; $p=0.21$, ${\eta }^2=0.08$].

No other effects reached significance in the PC-ROI.

3.2.

Effects at the M1-ROI

Baseline-corrected time courses for the M1-ROI are shown in Fig. 4, lower panel. There is only a slight difference in the change of concentration in both hemispheres for the MIR compared with the NOR condition for HbO and almost no difference between both conditions for HbR (Fig. 4, lower panel, left and right).

Mean beta values for the M1-ROI are shown in Fig. 5(d)–5(f), lower panel and in Table 1. As expected, the mean beta values for static trials were not significantly different from zero, whereas mean beta values for movement trials were clearly increased. Correspondingly, the four-way ANOVA over the mean beta values showed a significant main effect of the factor movement both for HbO [$F(1,19)=5.06$; $p<0.05$, ${\eta }^2=0.21$] and for HbR [$F(1,19)=5.51$; $p<0.05$, ${\eta }^2=0.23$].

The three-way ANOVA of the movement trials only (discarding the static trials) showed a main effect of hemisphere [$F(1,19)=4.65$, $p<0.05$, ${\eta }^2=0.20$] for HbR. The mean beta values indicate more activation contralateral compared with ipsilateral for the movement trials. The same effect was observed for HbO, but failed to reach significance [$F(1,19)=3.71$, $p=0.069$, ${\eta }^2=0.16$]. No other effects reached significance for the three-way ANOVA.

The three-way ANOVA of the static trials (discarding the movement trials) showed no significant effects.