Sage Journals: Discover world-class research

Abstract

Background. Mirrored feedback has potential as a therapeutic intervention to restore hand function after stroke. However, the functional (effective) connectivity of neural networks involved in processing mirrored feedback after stroke is not known. Objective. To determine if regions recruited by mirrored feedback topographically overlap with those involved in control of the paretic hand and to identify the effective connectivity of activated nodes within the mirrored feedback network. Methods. Fifteen patients with chronic stroke performed a finger flexion task with their unaffected hand during event-related functional magnetic resonance imaging (fMRI). Real-time hand kinematics was recorded during fMRI and used to actuate hand models presented in virtual reality (VR). Visual feedback of the unaffected hand motion was manipulated pseudorandomly by either actuating the VR hand corresponding to the moving unaffected side (veridical feedback) or the affected side (mirrored feedback). In 2 control conditions, the VR hands were replaced with moving nonanthropomorphic shapes. Results. Mirrored feedback was associated with significant activation of regions within and outside the ipsilesional sensorimotor cortex, overlapping with areas engaged when patients performed the task with their affected hand. Effective connectivity analysis showed a significantly interconnected ipsilesional somatosensory and motor cortex in the mirrored feedback condition. Conclusions. Mirrored feedback recruits ipsilesional brain areas relevant for control of the affected hand. These data provide a neurophysiological basis by which mirrored feedback may be beneficial as a therapy for restoring function after stroke.

Keywords

mirrored feedback motor control fMRI virtual reality visuomotor integration

The excitability of the ipsilesional corticospinal system after stroke is a marker of long-term functional recovery, ¹ which may explain why therapies that foster activation of the ipsilesional motor cortex have therapeutic value.^2,3 We suggest that providing mirrored feedback during movement, by visually mirroring the actual limb movement in real time, can facilitate activity of the ipsilesional motor cortex via rich intrahemispheric^4-6 and interhemispheric^7-10 connections among frontoparietal areas. Mirror therapy was originally introduced by Ramachandran and Rogers-Ramachandran¹¹ to reduce phantom limb pain in amputees and may benefit the recovery of hand function after stroke.^12-14 The core feature of mirror therapy is the incongruence between actual performance and mirrored feedback, which can be delivered by means of an actual mirror or through virtual reality (VR).

To our knowledge, only one group has so far investigated the neural correlates of mirror therapy in populations with chronic stroke. Michielsen and coworkers¹⁵ showed that mirror therapy may activate the precuneus and posterior cingulate cortex and that 6 weeks of bimanual exercises with mirrored feedback leads to significant increases in the blood oxygen level–dependent signal and its relateralization toward the ipsilesional side (compared to similar training without mirrored feedback).¹² A number of investigations have revealed similar effects in healthy individuals,^16,17 along with increases in corticospinal excitability (measured with motor evoked potentials).^18-21 Together, these findings suggest that mirrored feedback can modify the activation of sensorimotor brain areas in healthy populations and those with chronic stroke. However, the effective connectivity for mirrored feedback within the sensorimotor network, that is, how nodes in the sensorimotor network functionally interact in response to mirrored feedback, is still unknown. Studies by our group^22-24 and others^25,26 have shown that feedback of hand motion presented through high-fidelity VR can facilitate classic visuomotor networks in healthy and patient populations. Therefore, the aims of this study in patients with chronic stroke were the following: (1) To identify areas of the visuomotor system where activation is facilitated by mirrored feedback of unaffected hand motion. We hypothesized that mirrored feedback would activate the ipsilesional motor cortex. (2) To identify the functionally interconnected brain networks during mirrored feedback. We hypothesized that regions in the ipsilesional posterior parietal cortex and motor cortex would show increased functional connectivity elicited by mirrored feedback. (3) To test the relationship between changes in mirrored feedback–induced brain activation and the level of hand function. We hypothesized that the degree of activation elicited by mirrored feedback would depend on the level of hand function. (4) To test if regions with activation elicited by mirrored feedback overlap with those that are involved in volitional control of the affected hand. We hypothesized that mirrored feedback would facilitate brain regions that are relevant for control of the affected hand.

For this, patients with chronic stroke underwent event-related functional magnetic resonance imaging (fMRI). Participants performed simple finger movements during scanning, with their finger motion recorded using instrumented data gloves. Kinematic data from the gloves were used to actuate, in real time, the motion of virtual hand models. In any given trial, the VR hand motion corresponded either to the moving unaffected hand (veridical feedback) or to the affected hand (mirrored feedback). The use of a VR mirror setup with simultaneous performance of fMRI and hand kinematics allowed us to (1) ensure that patients focused exclusively on mirrored feedback without being distracted by seeing motion of the other hand, as might occur when using a physical mirror setup; (2) use an event-related design to pseudorandomize the conditions (which would be impossible if using a physical mirror); (3) identify and exclude from analysis any trials confounded by inadvertent motion of the fingers; and (4) deliver a set of control conditions to rule out confounders in activation that may be attributed to motor output, visual field, or arbitrary visual motion effects.

Methods

Participants

Fifteen right-handed patients with hemiparesis due to stroke (5 right hemiplegics and 5 females; mean age, 54 ± 12 years) (Table 1) participated after providing institutionally approved informed consent. One patient had excessive head movement during fMRI, and those data were excluded. Patients’ clinical status was assessed with the Chedoke-McMaster Rating Scale²⁷ and the Wolf Motor Function Test,²⁸ including its distal subtest (dWMFT).²⁹ Lesion location and volume (Figure 1A) were characterized from high-resolution T1-weighted MRI scans using MRIcron software .³⁰ Inclusion criteria for participation were the following: 30 to 80 years of age, a diagnosis of a first-time stroke at least 6 months before the study date, and normal or corrected-to-normal vision. Exclusion criteria included a history of any neurological pathological abnormality other than stroke, visuospatial deficits such as neglect or visual agnosia, impaired proprioception or sensation, any contraindications to being able to enter an MRI environment, any orthopedic trauma to the hand (not related to the stroke) that would hinder volitional control of movement, and receipt of Botox (Allergan Inc, Irvine, California) injections within the past 3 months.

Table 1.

Patients’ Clinical Information.

Patient	Age, y	Sex	Months from 1^st stroke	CMA/CMH	dWMFT, s	Lesion
S1	63	Female	53	6/4	175.5	Left frontal, parietal
S2	55	Male	41	5/4	53	Left thalamus
S3^a	49	Male	144	5/4	85	Left basal ganglia
S4	74	Male	9	6/6	21.4	Right frontal
S5	70	Female	96	7/5	40	Right corona radiata
S6	58	Male	132	5/4	96.5	Right frontal, parietal, temporal
S7	37	Male	92	4/3	82.5	Right pons
S8	69	Female	18	7/7	NA	Right pons
S9	68	Male	78	6/6	33.1	Right occipital, parietal
S10^a	48	Female	148	4/3	102.67	Right frontal, parietal
S11	41	Female	70	6/6	28.7	Right frontal, parietal, temporal
S12	43	Male	11	4/4	120.4	Left pons
S13	41	Male	158	6/6	44.7	Left frontal, parietal, temporal
S14	53	Male	156	6/6	44.1	Right pons
S15^a	39	Female	14	4/3	NA	Right parietal, temporal

CMA/CMH, Chedoke-McMaster Rating Scale for the hand and arm; dWMFT, distal subtest of the Wolf Motor Function Test; NA, not available.

Patients who were unable to move the paretic hand and therefore did not perform experiment 2.

Figure 1.

(A) Mapping of each patient’s lesion. Each row shows 5 different patients with lesions filled in with color. Blue horizontal lines in the midsagittal slice mark the levels of the axial slices to the left. (B) The experimental setup and feedback conditions in virtual reality (VR), viewed from the patient’s perspective. Motion of the right unaffected hand (upper left inset) would actuate the right (veridical) or left (mirrored) VR hand or a left/right nonanthropomorphic shape (control). All conditions were randomly interleaved within a functional magnetic resonance imaging run.

Setup

Patients wore left and right MRI-compatible recording gloves (Fifth Dimension Technologies Inc, Pretoria, South Africa).The gloves use fiber-optic sensors to measure metacarpophalangeal and proximal interphalangeal finger joint angles. The gloves were interfaced with a VR environment developed with Virtools software (Dassault Systemes, Vélizy-Villacoublay, France).³¹ Patients viewed the VR environment in the scanner through a back-projected image on a semitransparent screen, which enabled them to see a first-person view of left and right virtual hand models in semipronated positions (thumb toward the viewer) (Figure 1). Joint angle data streaming from the recording gloves actuated (in real time) motion of the virtual hand models. For example, motion of the index finger produced a corresponding motion of the virtual index finger. All joint angle data were also saved for statistical analysis of the movement kinematics.

Task

Patients with stroke were required to perform whole-hand finger flexion-extension movements with their unaffected hand (experiment 1) or with their affected hand (experiment 2). During the task, data from the recording gloves were streamed to actuate the virtual hand models. Visual feedback was manipulated as described below. All patients were familiarized with the task and feedback conditions before imaging.

Visual Feedback Conditions for Experiment 1

Experiment 1 was designed to answer aims 1 to 3 (see introductory section). To identify brain activation elicited by mirrored feedback, patients performed the task with the unaffected hand. The event-related fMRI protocol consisted of 4 feedback conditions that were presented in a pseudorandom order (Figure 1). Each condition was repeated for 8 trials per fMRI run. Four fMRI runs were performed per patient. Trial duration was 5 seconds, separated by an intertrial rest period that randomly varied from 4 to 7 seconds. The visual feedback conditions were the following:

Veridical feedback condition: The actuated virtual hand corresponded to the same side as the moving hand (eg, moving the right hand actuated the right virtual hand).

Mirrored feedback condition: The actuated virtual hand was on the opposite side of the moving hand (eg, moving the right hand actuated the left virtual hand).

Control feedback conditions were added to subtract potentially confounding activation related to the position of virtual objects in the visual field, gaze direction, and observed motion. For this, the virtual hand models were replaced with nonanthropomorphic objects (ellipsoids that were similar in size and color to the virtual hands). One of the 2 objects rotated about an oblique axis at 1 Hz.

Veridical control condition: The moving nonanthropomorphic object was on the same side as the moving hand.

Mirrored control condition: The moving nonanthropomorphic object was on the opposite side to the moving hand.

Experiment 2 was used to answer aim 4 (see introductory section). To identify brain activation associated with volitional control of the affected hand, patients moved only the affected hand. Only veridical feedback was presented. Movement and rest periods were presented in alternating order (16 trials per functional run). Two functional imaging runs were conducted. Three patients could not move the fingers of the affected hand consistently and therefore did not perform experiment 2 (Table 1).

Movement Kinematics

Kinematic analysis was used to verify that movements were consistent across the feedback conditions. Trials with inadvertent motion of the stationary hand were not included in the analysis. For each trial, movement onset and offset were defined as the time at which the mean angular velocity of the 4 metacarpophalangeal joints exceeded and then fell below 5% of the mean peak angular velocity. Movement time was the interval between onset and offset. Movement onset and movement time were included in the fMRI general linear model on a trial-by-trial basis for convolution of the blood oxygen level–dependent signal of each event with the hemodynamic response function. To verify that movements remained consistent, the peak angle (angular excursion) and movement time were averaged across trials for each functional imaging run and each feedback condition and were submitted to a 3-way repeated-measures analysis of variance (rmANOVA). Factors (levels) were object viewed (virtual hands, control), feedback type (veridical, mirrored), and fMRI run (run 1, run 2, run 3, run 4). Statistical significance was set at .05 (Bonferroni corrected).

fMRI Acquisition and Analysis

The fMRI data were acquired using a 3-T Siemens Allegra (Munich, Germany) head-only scanner with a standard head coil. Parameters for the high-resolution structural images were the following: repetition time (TR), 2 seconds; echo time (TE), 4.38 milliseconds; voxel size, 0.938 × 0.938 × 1 mm; number of slices, 176; and slice thickness, 1 mm. Parameters for the functional images were the following: T2*-weighted echo planar imaging sequence; TR, 2 seconds; TE, 30 milliseconds; field of view, 100 mm; voxel size, 3 mm³; number of slices, 32; interslice time, 62 milliseconds; and number of volumes, 175. All fMRI data were preprocessed and analyzed in SPM8, (fil.spm@ucl.ac.uk, London) . The first 2 dummy scans were acquired to account for field inhomogeneities but not included in the analysis.

Functional volumes were realigned to the first volume, slice time corrected, coregistered to the structural image, and normalized to a standardized brain in the Montreal Neurological Institute (MNI) space. A binary mask of each patient’s lesion was used to optimize brain normalization with the DARTEL³² toolbox in SPM8, consistent with established approaches.^32-36 Functional scans were smoothed using an 8-mm full width at half maximum Gaussian kernel. Separate general linear models were created in SPM8 for each experiment and patient.

Analysis was conducted at the single-patient (fixed-effects) and group level (random-effects). For group-level analysis, fMRI scans of patients with left hemisphere lesions were flipped to the right. This is consistent with previous approaches for creating group-level means in patient populations who have lesions in different hemispheres.³⁷ Statistical significance for all analyses was set at P < .05 (cluster-level family-wise error corrected [voxel extent k > 10]).

The following contrasts were created for experiment 1:

Contrast 1: Main effect of VR hand feedback. This contrast identified regions recruited by observing virtual hand motion versus nonanthropomorphic control (CTRL) objects.

(HAND[mirror] + HAND[veridical]) > (CTRL[mirror] + CTRL[veridical])

Contrast 2: Effect of mirrored feedback. This contrast identified regions recruited by mirrored feedback.

HAND[mirror] > (HAND[veridical] + CTRL[mirror] + CTRL[veridical])

The following contrasts were created for experiment 2:

Contrast 3: Effect of paretic hand motion. This contrast identified regions recruited for controlling movement of the affected hand.

HAND[veridical] > rest

Contrast 4: Conjunction analysis. This contrast identified regions that were recruited both by mirrored feedback (contrast 2) and by volitional control of the paretic hand (contrast 3). For the conjunction analysis, a binary mask of contrast 3 images was created (threshold of t > 1.7) and multiplied by contrast 2 images at a patient-by-patient level. To reduce the possibility of type II error in this sample size, a nonparametric pseudo–t test was computed for the overlap images using SnPM.³⁸ The SPM8 parameters were set to zero variance smoothing and the maximum number of iterations. Results were evaluated at a cluster-wise threshold of P < .05 (family-wise error corrected).

Effective Connectivity

Effective connectivity explains the task-specific changes in functional connectivity between a region of interest and the rest of the brain. Effective connectivity was quantified using the generalized psychophysiological interaction (PPI) toolbox for SPM8.^39-41 Three a priori–defined regions of interest within the ipsilesional sensorimotor cortex^42,43 were specified, based on their importance in sensorimotor control, for their predictive validity of stroke recovery⁴⁴ and their involvement in mirrored feedback paradigms in healthy populations (see introductory section)⁴⁵ and in contrast 2 of the current study. A fourth adjacent seed (supplementary motor area [SMA]) was added based on published reports that it has significantly altered connectivity with the motor cortex in populations with stroke.⁴⁶ The mean ± standard deviation of each seed location in the MNI space was the following:

Precentral gyrus corresponding to the motor cortex (Brodmann area 4 [BA4]): 40 ± 4.2, −13 ± 4.5, 47 ± 2.1

Primary somatosensory cortex (BA1): 62 ± 1.7, −18 ± 1.4, 40 ± 5.3

Secondary somatosensory cortex (BA2): 41 ± 7, −30 ± 4, 50 ± 3

SMA: 0 ± 8, −4 ± 16, 67 ± 9

The procedure for PPI analysis is well published.^40,41 Briefly, the time series of each seed is extracted, deconvolved with the boxcar function of the task events for each condition, and entered as a regressor in the general linear model. The strength of the regression between the seed and the remaining voxels in the brain represents the effective connectivity map for that seed (referred to as a contribution map). Contribution maps were generated for each condition and contrasted at the single-patient level using the following contrast:

Contrast 5

HAND[mirror] > (HAND[veridical] + CTRL[mirror] + CTRL[veridical])

These contrast images were statistically tested at the group level using a 1-sample t test.

Relationship Between Brain Activation and Hand Function

To ascertain if the magnitude of the mirrored feedback–based activation was related to the degree of hand function, a regression analysis was performed on a patient-by-patient basis between the t values of the significantly activated clusters of contrast 2 and contrast 5 and the dWMFT clinical score.

Results

Consistency of Hand Kinematics Across Conditions

The mirror experiment was performed with the unaffected hand. Hand kinematics was inspected to identify missed trials and inadvertent movements. Such trials comprised only 2% of all trials and were excluded from analyses. A χ² test revealed no significant difference in the number of trials removed from each condition (P > .05).

Hand kinematics was also analyzed to verify if movements were consistent across conditions. A 3-way rmANOVA for mean angular velocity and movement duration did not show significant effects of object viewed (virtual hands, CTRL), feedback type (veridical, mirrored), or fMRI run (run 1, run 2, run 3, run 4) (P > .05). Finger excursion (movement amplitude) tended to increase from 0.7 ± 0.39 rad in the first run to 0.89 ± 0.41 rad in the last run of the CTRL condition (fMRI run × object viewed interaction: F_3,39 = 4.67; P = .01), although this did not occur when patients performed the task with real-time feedback of the VR hands. These analyses suggested that patients performed slightly larger amplitude movements in the control condition but that, otherwise, movements were generally comparable across veridical and mirrored feedback conditions, making any differences in subsequent fMRI results unlikely to be confounded by movement kinematics.

Main Effect of Movement on Brain Activation

Motion of the unaffected hand (across all feedback types) was associated with activation in a distributed network predominantly involving the contralateral frontoparietal network that is traditionally recruited during visually guided movement. Motion of the affected hand was associated with significant activation of the bilateral sensorimotor cortex, parietal lobe, ipsilateral occipitotemporal cortex, and contralateral insula (Table 2).

Table 2.

fMRI Results for All Contrasts.

Region	Side	t Statistic	P Value (FWE Corrected)	x, y, z	k
Main effect of hand feedback versus control feedback
Inferior parietal	Left	6.29	.0000	−27, −54, 51	293
Midoccipital	Left	6.08	.0250	−30, −84, 36	112
Main effect of mirror feedback
Superior parietal	Left	5.89	.0000	−30, −54, 51	899
Inferior parietal	Left	5.26		−39, −39, 45
Supramarginal	Left	4.86		−60, −39, 33
Postcentral gyrus	Right	5.52	.0160	36, −33, 51	426
Effective connectivity (seed: iA5)
Precentral gyrus	Right	4.62	.035	39, −12, 48	148
Postcentral gyrus	Right	4.26		51, −15, 36
Postcentral gyrus	Right	4.01		21, −30, 63
Main effect of mirror feedback overlap with experiment 2
Postcentral gyrus	Right	4.61	.0049	36, −30, 45	131
Inferior parietal	Left	3.93	.0005	−36, −42, 48	211
Midoccipital	Left	3.49	.0227	−30, −66, 36	64

Significant clusters that are family-wise error (FWE) corrected (P < .05; extent > 10) at the voxel level. Note that “right” corresponds to the ipsilesional side after flipping all patients with left-sided lesions to the right side (to allow for group-level comparisons). fMRI, functional magnetic resonance imaging; k, cluster size in voxels; x, y, z, coordinates (mm) of the peak voxel in the Montreal Neurological Institute space.

Main Effect of Viewed Object Type on Brain Activation

The type of object viewed (virtual hands > CTRL) was associated with significant activation in the contralesional posterior parietal cortex, extending to the occipitoparietal area, middle occipital cortex, and ipsilesional somatosensory cortex (Table 2).

Brain Activation Related to Mirrored Feedback

Mirrored feedback was associated with significant activation of a large cluster in the ipsilesional postcentral gyrus, corresponding to BA1 (Figure 2 and Table 2). The cluster extended rostrally to the anterior bank of the central sulcus (BA4) (eg, primary motor cortex) and caudally to the precuneus, adjacent to a region identified by Michielsen et al¹⁵ in their mirror paradigm. Figure 2 (also Table 2) shows that mirrored feedback also elicited significant activation in the contralesional cortex, including the postcentral gyrus (BA1-BA3), superior bank of the intraparietal sulcus and precuneus (BA7), and supramarginal gyrus (BA40). A conjunction analysis (Figure 3) revealed that activation for mirrored feedback (during motion of the unaffected hand) was topographically similar to regions activated by volitional motion of the affected hand under veridical feedback conditions. The overlapping regions included the ipsilesional postcentral gyrus (corresponding to the junction of BA3 and BA5) and contralesional superior parietal lobule (BA7) (Table 2).

Figure 2.

Group mean of the main effect of mirrored feedback (contrast 2) (family-wise error corrected P < .05; minimal cluster size k = 10). For averaging purposes, left-sided lesions were flipped to the right (see Methods) such that the right hemisphere (R) represents the ipsilesional hemisphere. Surface plots were generated using render and canonical templates in SPM8. Note that mirrored feedback activation was observed in the ipsilesional precentral gyrus, corresponding to the hand area of the motor cortex, along with other distributed frontal and parietal areas.

Figure 3.

Group mean of the conjunction analysis between contrast 2 (experiment 1) and contrast 3 (experiment 2) to identify topographic overlap between brain regions recruited during motion of the unaffected hand with mirrored feedback and those controlling motion of the affected hand.

Effective Connectivity During Mirrored Feedback

Effective connectivity was used to quantify task-based network interactions between 4 seed regions (BA4, BA1, BA2, SMA) and the rest of the brain (see Methods: Effective Connectivity for details). Figure 4A and 4B show that the connectivity between the BA1 seed and the primary motor cortex (BA4) and between the BA1 seed and the primary sensory cortex (BA3) were significantly stronger for mirrored feedback, relative to the other conditions. Effective connectivity between the other seed regions and the whole brain did not differ between conditions.

Figure 4.

(A) Effective connectivity of the psychophysiological interaction between the Brodmann area (BA1) seed region (white circle) (Montreal Neurological Institute space: 63, −18, 39) and the rest of the brain (family-wise error corrected P < .05; minimal cluster size k = 10) overlaid on a partially inflated cortical surface (Caret software, http://www.nitrc.org/projects/caret/, St. Louis, MO) from the right lateral and top view. (B) Bar plots showing the strength of effective connectivity between the BA1 seed and BA4 (39, −12, 48) and between the BA1 seed and BA3 (51, −15, 36) in each of the 4 conditions. (C) Linear regression between the t values corresponding to the strength of effective connectivity (y-axis) and level of hand function (x-axis). Hand function is plotted as the log-normed distal subtest of the Wolf Motor Function Test (dWMFT) score (1-log₁₀ dWMFT), with larger values indicating better hand function.

Relationship Between Brain Activity and Affected Hand Function

No significant correlation was noted between the level of hand function (log-normed dWMFT) and the degree of activation elicited by mirrored feedback, relative to the other conditions. However, Figure 4C (bottom) demonstrates that a significant correlation between BA1-BA3 effective connectivity and the log-normed dWMFT score (F_1,11 = 14.14; P = .0037, Bonferroni corrected) was present, suggesting that stronger connectivity was elicited in persons with greater impairment. The correlation between BA1-BA4 effective connectivity and the dWMFT score (Figure 4C, top) showed a similar trend but did not reach statistical significance after correction (F_1,10 = 5.94; P = .0375).

Discussion

Our data show that mirrored feedback during movement of the unaffected hand is associated with significant activation of the ipsilesional sensorimotor cortex. The observation of virtual hands versus nonanthropomorphic shapes recruited higher order visual networks in the lateral occipital cortex similar to other fMRI studies.^22,26 The fMRI contrast, with subtraction of the control conditions, as well as the analysis of finger kinematics acquired concomitantly with the fMRI data, suggests that this activation is unlikely to be confounded by visual motion, gaze effects,⁴⁷ location of objects in a particular hemifield, any potential differences in movement kinematics between conditions, or inadvertent mirror movements.⁴⁸ Finally, the presence of significant activation even after subtraction of activation in the control conditions (in which patients performed the same movements) suggests that the cortical response was most likely elicited by feedback processes rather efferent signals from the contralesional motor cortex.

Previous work from independent laboratories suggests that the sensorimotor system of the hemisphere contralateral to the mirrored feedback hand can be facilitated in healthy individuals. While this facilitation seems to include the corticospinal system itself,^18,20,49,50 work on fMRI, electroencephalography, and magnetoencephalography also suggests that it may extend to higher order visuomotor areas in the parietal, posterior temporal, and occipital areas as well.^15-17,45,51 Further substantiating this view is a recent elegant study in which healthy individuals trained on fine motor tasks over 4 days using their right hand with either veridical or mirrored feedback.⁵² The mirrored feedback group demonstrated superior performance on the task when tested with the untrained left hand and an increase in activity in the right dorsal premotor cortex, left ventral premotor cortex, and left SMA. Together, these data in healthy individuals lend support to the idea that the mirrored feedback responses observed in our study may be similar, although not identical, to those noted in the above-mentioned studies.

Given the therapeutic potential of mirror therapy, the question therefore arises whether the brains of patients with stroke have the capacity to respond to mirrored feedback similar to what has been noted in healthy individuals. Interestingly, studies investigating the effectiveness of mirror therapy have yielded inconsistent results^12,13,53-57 (for review, see Thieme et al¹⁴ and Ezendam et al⁵⁸), suggesting that the responsiveness of the lesioned nervous system, particularly the sensorimotor system, may be questionable. Indeed, the only study, to date, that we are aware of testing for activation associated with mirrored feedback in patients with stroke found significant activation induced by mirrored feedback in the posterior cingulate cortex and precuneus but not the sensorimotor cortex.¹⁵ Differences between our design and those of Michielsen et al¹⁵ (such as using a blocked vs event-related design and a real vs virtual mirror setup) may have led to the distinctions in the observed results. Indeed, some have shown that mirrored feedback presented through VR may potentiate corticospinal excitability more strongly than when presented through a real mirror,²⁵ although the reason for this still remains unclear. In this regard, a limitation of our design was the absence of a comparison group with a real mirror, which was logistically impossible in an event-related fMRI design, precluding us from drawing such comparisons.

Effective Connectivity of Networks Processing Mirrored Feedback

A central aim of this study was to characterize the effective connectivity of the network interactions invoked by mirrored feedback. Our data show significant effective connectivity interactions between the ipsilesional somatosensory cortex (BA1) and the sensorimotor cortex in the caudal and rostral banks of the central sulcus (BA3 and BA4). Notably, we found that BA1-BA3 effective connectivity was correlated with the level of hand function, suggesting a testable prediction that mirror therapy may have the greatest yield for more impaired individuals, at least for individuals in the chronic stage. If true, then this may provide a basis for testing the potential of mirror therapy in patients with more acute stroke who are generally more impaired and have the greatest potential for recovery. It is interesting to note that no change in effective connectivity was observed between an adjacent region, the SMA, and the sensorimotor cortex, highlighting the specificity of mirrored feedback–induced interactions. However, these speculations need to be directly tested in empirical studies. Moreover, it remains unclear at this point how lesions to nodes in this network may affect the mirrored feedback–induced changes in activation.

Mirrored Feedback Recruits Regions Involved in Volitional Control of the Affected Hand

A notable finding in our study was that regions activated by mirrored feedback share a degree of topographical overlap with regions involved in volitional control of the affected hand. These data suggest that mirrored feedback may facilitate motorically relevant regions of the sensorimotor system in patients with stroke with moderate hemiparesis. Because movement-based interventions for lower functioning individuals are limited by the amount of remaining volitional motion after stroke, activation of the sensorimotor system through mirrored feedback may be particularly suitable for severely hemiplegic patients or for those in the early stages after stroke. This finding is consistent with recent fMRI work in amputee patients, showing that mirrored feedback paired with mental imagery of moving the amputated limb can recruit sensorimotor territories that would represent the amputated extremity.⁵⁹

Conclusion

Our study demonstrates that mirrored feedback can invoke activity and strengthen functional interactions within topographically relevant sensorimotor areas of the lesioned hemisphere. Further study of the neural mechanisms and causal interactions that underlie mirrored feedback is necessary to develop therapeutic interventions that boost activity in select brain networks to potentially enhance brain reorganization.

Footnotes

Acknowledgements

The authors thank Michal Kafri and Thushini Manuweera for review and feedback on this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by National Institutes of Health grants K01-HD059983 (E.T.) and R01-HD58301 (S.V.A.) and by National Institute on Disability and Rehabilitation (Research Rehabilitation Engineering Research Center) grant H133E050011 (S.V.A.).

References

Trompetto

Assini

Buccolieri

Marchese

Abbruzzese

. Motor recovery following stroke: a transcranial magnetic stimulation study. Clin Neurophysiol. 2000;111:1860-1867.

Hummel

Celnik

Giraux

. Effects of non-invasive cortical stimulation on skilled motor function in chronic stroke. Brain. 2005;128:490-499.

Calautti

Naccarato

Jones

. The relationship between motor deficit and hemisphere activation balance after stroke: a 3T fMRI study. Neuroimage. 2007;34:322-331.

Rushworth

Behrens

Johansen-Berg

. Connection patterns distinguish 3 regions of human parietal cortex. Cereb Cortex. 2006;16:1418-1430.

Lewis

Van Essen

. Mapping of architectonic subdivisions in the macaque monkey, with emphasis on parieto-occipital cortex. J Comp Neurol. 2000;428:79-111.

Camarda

Matelli

Glickstein

Rizzolatti

. [Intrinsic and extrinsic connections of the postarcuate premotor area of the monkey]. Boll Soc Ital Biol Sper. 1984;60:823-829.

Dancause

Barbay

Frost

Mahnken

Nudo

. Interhemispheric connections of the ventral premotor cortex in a new world primate. J Comp Neurol. 2007;505:701-715.

Baumer

Bock

Koch

. Magnetic stimulation of human premotor or motor cortex produces interhemispheric facilitation through distinct pathways. J Physiol. 2006;572:857-868.

Killackey

Gould

3rd Cusick

Pons

Kaas

. The relation of corpus callosum connections to architectonic fields and body surface maps in sensorimotor cortex of new and old world monkeys. J Comp Neurol. 1983;219:384-419.

10.

Rouiller

Babalian

Kazennikov

Moret

Wiesendanger

. Transcallosal connections of the distal forelimb representations of the primary and supplementary motor cortical areas in macaque monkeys. Exp Brain Res. 1994;102:227-243.

11.

Ramachandran

Rogers-Ramachandran

. Synaesthesia in phantom limbs induced with mirrors. Proc Biol Sci. 1996;263:377-386.

12.

Michielsen

Selles

van der Geest

. Motor recovery and cortical reorganization after mirror therapy in chronic stroke patients: a phase II randomized controlled trial. Neurorehabil Neural Repair. 2011;25:223-233.

13.

Dohle

Pullen

Nakaten

Kust

Rietz

Karbe

. Mirror therapy promotes recovery from severe hemiparesis: a randomized controlled trial. Neurorehabil Neural Repair. 2009;23:209-217.

14.

Thieme

Mehrholz

Pohl

Behrens

Dohle

. Mirror therapy for improving motor function after stroke. Cochrane Database Syst Rev. 2012;3:CD008449.

15.

Michielsen

Smits

Ribbers

. The neuronal correlates of mirror therapy: an fMRI study on mirror induced visual illusions in patients with stroke. J Neurol Neurosurg Psychiatry. 2011;82:393-398.

16.

Diers

Christmann

Koeppe

Ruf

Flor

. Mirrored, imagined and executed movements differentially activate sensorimotor cortex in amputees with and without phantom limb pain. Pain. 2010;149:296-304.

17.

Matthys

Smits

Van der Geest

. Mirror-induced visual illusion of hand movements: a functional magnetic resonance imaging study. Arch Phys Med Rehabil. 2009;90:675-681.

18.

Fukumura

Sugawara

Tanabe

Ushiba

Tomita

. Influence of mirror therapy on human motor cortex. Int J Neurosci. 2007;117:1039-1048.

19.

Nojima

Mima

Koganemaru

Thabit

Fukuyama

Kawamata

. Human motor plasticity induced by mirror visual feedback. J Neurosci. 2012;32:1293-1300.

20.

Garry

Loftus

Summers

. Mirror, mirror on the wall: viewing a mirror reflection of unilateral hand movements facilitates ipsilateral M1 excitability. Exp Brain Res. 2005;163:118-122.

21.

Funase

Tabira

Higashi

Liang

Kasai

. Increased corticospinal excitability during direct observation of self-movement and indirect observation with a mirror box. Neurosci Lett. 2007;419:108-112.

22.

Adamovich

August

Merians

Tunik

. A virtual reality-based system integrated with fMRI to study neural mechanisms of action observation-execution: a proof of concept study. Restor Neurol Neurosci. 2009;27:209-223.

23.

Tunik

Adamovich

. Remapping in the ipsilesional motor cortex after VR-based training: a pilot fMRI study. Conf Proc IEEE Eng Med Biol Soc. 2009;2009:1139-1142.

24.

Tunik

Saleh

Adamovich

. Visuomotor discordance during visually-guided hand movement in virtual reality modulates sensorimotor cortical activity in healthy and hemiparetic subjects. IEEE Trans Neural Syst Rehabil Eng. 2013;21:198-207.

25.

Kang

Park

Kim

. Upper extremity rehabilitation of stroke: facilitation of corticospinal excitability using virtual mirror paradigm. J Neuroeng Rehabil. 2012;9:71.

26.

Perani

Fazio

Borghese

. Different brain correlates for watching real and virtual hand actions. Neuroimage. 2001;14:749-758.

27.

Gowland

Stratford

Ward

. Measuring physical impairment and disability with the Chedoke-McMaster Stroke Assessment. Stroke. 1993;24:58-63.

28.

Wolf

Thompson

Morris

. The EXCITE trial: attributes of the Wolf Motor Function Test in patients with subacute stroke. Neurorehabil Neural Repair. 2005;19:194-205.

29.

Fluet

Merians

Qiu

. Robots integrated with virtual reality simulations for customized motor training in a person with upper extremity hemiparesis: a case study. J Neurol Phys Ther. 2012;36:79-86.

30.

Rorden

Brett

. (2000). Stereotaxic display of brain lesions. Behavioural Neurology, 12, 191-200.

31.

Rorden

Karnath

Bonilha

. Improving lesion-symptom mapping. J Cogn Neurosci. 2007;19:1081-1088.

32.

August

Lewis

Chandar

Merians

Biswal

Adamovich

. fMRI analysis of neural mechanisms underlying rehabilitation in virtual reality: activating secondary motor areas. Conf Proc IEEE Eng Med Biol Soc. 2006;1:3692-3695.

33.

Ashburner

. A fast diffeomorphic image registration algorithm. Neuroimage. 2007;38:95-113.

34.

Ashburner

Friston

. Unified segmentation. Neuroimage. 2005;26:839-851.

35.

Brett

Leff

Rorden

Ashburner

. Spatial normalization of brain images with focal lesions using cost function masking. Neuroimage. 2001;14:486-500.

36.

Andersen

Rapcsak

Beeson

. Cost function masking during normalization of brains with focal lesions: still a necessity? Neuroimage. 2010;53:78-84.

37.

Crinion

Ashburner

Leff

Brett

Price

Friston

. Spatial normalization of lesioned brains: performance evaluation and impact on fMRI analyses. Neuroimage. 2007;37:866-875.

38.

Walther

Juenger

Kuhnke

. Motor cortex plasticity in ischemic perinatal stroke: a transcranial magnetic stimulation and functional MRI study. Pediatr Neurol. 2009;41:171-178.

39.

Nichols

Holmes

. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum Brain Mapp. 2002;15:1-25.

40.

Friston

Buechel

Fink

Morris

Rolls

Dolan

. Psychophysiological and modulatory interactions in neuroimaging. Neuroimage. 1997;6:218-229.

41.

O’Reilly

Woolrich

Behrens

Smith

Johansen-Berg

. Tools of the trade: psychophysiological interactions and functional connectivity. Soc Cogn Affect Neurosci. 2012;7:604-609.

42.

McLaren

Ries

Johnson

. A generalized form of context-dependent psychophysiological interactions (gPPI): a comparison to standard approaches. Neuroimage. 2012;61:1277-1286.

43.

Grefkes

Geyer

Schormann

Roland

Zilles

. Human somatosensory area 2: observer-independent cytoarchitectonic mapping, interindividual variability, and population map. Neuroimage. 2001;14:617-631.

44.

Geyer

Schormann

Mohlberg

Zilles

. Areas 3a, 3b, and 1 of human primary somatosensory cortex, part 2: spatial normalization to standard anatomical space. Neuroimage. 2000;11:684-696.

45.

Marshall

Zarahn

Alon

Minzer

Lazar

Krakauer

. Early imaging correlates of subsequent motor recovery after stroke. Ann Neurol. 2009;65:596-602.

46.

Wasaka

Kakigi

. The effect of unpredicted visual feedback on activation in the secondary somatosensory cortex during movement execution. BMC Neurosci. 2012;13:138.

47.

Grefkes

Eickhoff

Nowak

Dafotakis

Fink

. Dynamic intra- and interhemispheric interactions during unilateral and bilateral hand movements assessed with fMRI and DCM. Neuroimage. 2008;41:1382-1394.

48.

Baker

Donoghue

Sanes

. Gaze direction modulates finger movement activation patterns in human cerebral cortex. J Neurosci. 1999;19:10044-10052.

49.

Kim

Jang

Chang

Byun

Son

Ahn

. Bilateral primary sensori-motor cortex activation of post-stroke mirror movements: an fMRI study. Neuroreport. 2003;14:1329-1332.

50.

Touzalin-Chretien

Dufour

. Motor cortex activation induced by a mirror: evidence from lateralized readiness potentials. J Neurophysiol. 2008;100:19-23.

51.

Tominaga

Matsubayashi

Deguchi

. A mirror reflection of a hand modulates stimulus-induced 20-Hz activity. Neuroimage. 2009;46:500-504.

52.

Voisin

Rodrigues

Hetu

. Modulation of the response to a somatosensory stimulation of the hand during the observation of manual actions. Exp Brain Res. 2011;208:11-19.

53.

Hamzei

Lappchen

Glauche

Mader

Rijntjes

Weiller

. Functional plasticity induced by mirror training: the mirror as the element connecting both hands to one hemisphere. Neurorehabil Neural Repair. 2012;26:484-496.

54.

Altschuler

Wisdom

Stone

. Rehabilitation of hemiparesis after stroke with a mirror. Lancet. 1999;353:2035-2036.

55.

Sathian

Greenspan

Wolf

. Doing it with mirrors: a case study of a novel approach to neurorehabilitation. Neurorehabil Neural Repair. 2000;14:73-76.

56.

Sutbeyaz

Yavuzer

Sezer

Koseoglu

. Mirror therapy enhances lower-extremity motor recovery and motor functioning after stroke: a randomized controlled trial. Arch Phys Med Rehabil. 2007;88:555-559.

57.

Moseley

Gallace

Spence

. Is mirror therapy all it is cracked up to be? Current evidence and future directions. Pain. 2008;138:7-10.

58.

Yavuzer

Selles

Sezer

. Mirror therapy improves hand function in subacute stroke: a randomized controlled trial. Arch Phys Med Rehabil. 2008;89:393-398.

59.

Ezendam

Bongers

Jannink

. Systematic review of the effectiveness of mirror therapy in upper extremity function. Disabil Rehabil. 2009;31:2135-2149.

60.

Bogdanov

Smith

Frey

. Former hand territory activity increases after amputation during intact hand movements, but is unaffected by illusory visual feedback. Neurorehabil Neural Repair. 2012;26:604-615.

Mirrored Feedback in Chronic Stroke

Abstract

Keywords

Methods

Participants

Setup

Task

Visual Feedback Conditions for Experiment 1

Movement Kinematics

fMRI Acquisition and Analysis

Effective Connectivity

Relationship Between Brain Activation and Hand Function

Results

Consistency of Hand Kinematics Across Conditions

Main Effect of Movement on Brain Activation

Main Effect of Viewed Object Type on Brain Activation

Brain Activation Related to Mirrored Feedback

Effective Connectivity During Mirrored Feedback

Relationship Between Brain Activity and Affected Hand Function

Discussion

Effective Connectivity of Networks Processing Mirrored Feedback

Mirrored Feedback Recruits Regions Involved in Volitional Control of the Affected Hand

Conclusion

Footnotes

Acknowledgements

Declaration of Conflicting Interests

Funding

References