This Article Abstract Full Text (PDF) All Versions of this Article: 37/7/1941    most recent 01.STR.0000226902.43357.fcv1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Sharma, N. Articles by Baron, J.-C. Search for Related Content PubMed PubMed Citation Articles by Sharma, N. Articles by Baron, J.-C. Related Collections Computerized tomography and Magnetic Resonance Imaging Rehabilitation, Stroke

(Stroke. 2006;37:1941.)
© 2006 American Heart Association, Inc.

Progress Reviews

Motor Imagery

A Backdoor to the Motor System After Stroke?

From the Department of Clinical Neurosciences (N.S., J.-C.B.), University of Cambridge, England; and the Centre for Rehabilitation and Ageing (V.M.P.), St George’s, University of London, England.

Correspondence to Prof Jean-Claude Baron, University of Cambridge, Department of Clinical Neurosciences, Addenbrooke’s Hospital Box 83, Cambridge CB2 2QQ, UK. E-mail jcb54{at}cam.ac.uk

	Abstract

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
Background and Purpose— Understanding brain plasticity after stroke is important in developing rehabilitation strategies. Active movement therapies show considerable promise but depend on motor performance, excluding many otherwise eligible patients. Motor imagery is widely used in sport to improve performance, which raises the possibility of applying it both as a rehabilitation method and to access the motor network independently of recovery. Specifically, whether the primary motor cortex (M1), considered a prime target of poststroke rehabilitation, is involved in motor imagery is unresolved.

Summary of Review— We review methodological considerations when applying motor imagery to healthy subjects and in patients with stroke, which may disrupt the motor imagery network. We then review firstly the motor imagery training literature focusing on upper-limb recovery, and secondly the functional imaging literature in healthy subjects and in patients with stroke.

Conclusions— The review highlights the difficulty in addressing cognitive screening and compliance in motor imagery studies, particularly with regards to patients with stroke. Despite this, the literature suggests the encouraging effect of motor imagery training on motor recovery after stroke. Based on the available literature in healthy volunteers, robust activation of the nonprimary motor structures, but only weak and inconsistent activation of M1, occurs during motor imagery. In patients with stroke, the cortical activation patterns are essentially unexplored as is the underlying mechanism of motor imagery training. Provided appropriate methodology is implemented, motor imagery may provide a valuable tool to access the motor network and improve outcome after stroke.

Key Words: magnetic resonance imaging, functional • recovery of function • rehabilitation • stroke • tomography, emission computed

	Introduction

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
Motor impairment after stroke is a major cause of permanent disability. Recovery of the hand is crucial in order to perform activities of daily living but is often variable and incomplete.¹ Neuroscience-based rehabilitation is gaining strength as a way to improve outcome, even in situations where the deficit appears to be permanent.² Understanding the effect of rehabilitative techniques on brain plasticity is potentially important in providing a neural substrate to underpin rehabilitation and hence in developing novel rehabilitation strategies.

Active movement training (AMT) such as constraint-induced therapy (CIT) benefits from translational research; following a primary motor cortex (M1) infarct in primates, AMT can prevent loss of the peri-infarct hand territory and can even force expansion into adjacent areas.³ In chronic stroke patients CIT may be able to overcome "learned nonuse", resulting in cortical hand expansion and functional gains at 6 months.² When applied directly to humans, consistent relationships have emerged between motor recovery and cortical activation patterns after stroke, such that predominant ipsilesional activation, ie, positive Laterality Index, is associated with better recovery (see reviews by Ward and Cohen,⁴ Calautti and Baron⁵). Moreover AMT can result in both cortical reorganization and improved motor performance. Specifically in the chronic stage of stroke, CIT-like procedures can increase ipsilesional premotor and somatosensory cortex activation,⁶ whereas AMT can result in increases in ipsilesional M1 activation with a positive Laterality index shift.⁷ The idea is therefore emerging that "forcing" the use of the idle motor network, particularly M1, promotes long-term recovery, provided M1 is present and not isolated from the anterior horn cell.

Nevertheless, AMT has limitations; rodent models suggest CIT should be restricted during the early phase of stroke because it may result in infarct expansion,⁸ although this has not been encountered clinically.⁹ Difficulties implementing CIT in the clinical setting have resulted in a modified CIT protocol.¹⁰ However, poor motor performance still excludes almost 4 of 5 otherwise eligible subacute stroke patients from CIT.¹¹ Attempts to bridge the period of poor motor performance by combining different AMTs, such as electromyography-triggered neuromuscular stimulation and modified CIT,^12,13 have produced unconvincing results with some severely affected subjects verbalizing frustration.

Alternative strategies such as passive mobilization can be used in persons with hemiplegia to access the motor system although it produces mainly proprioceptive input to motor pathways.¹⁴ Yet, in healthy subjects passive mobilization neither improves performance nor induces the cortical plasticity shown with active movement, supporting a key role for voluntary drive in motor learning and neuro-rehabilitation.¹⁵

In principle a therapy based on attempted movement is conceivable because it combines both M1 activation¹⁶ and voluntary drive. The authors are unaware of any published trials that use attempted movement as a rehabilitation invention; this perhaps reflects the practical difficulties in avoiding patient frustration and fatigue.¹² Moreover, in subjects with hemiplegia, attempted movement is likely to be complicated by proximal or undesirable movement, and furthermore there would be no way to assess compliance (see below). Overall, the clinical uses of AMT in stroke are limited by their dependence on a degree of motor performance.

Motor imagery represents an intriguing new "backdoor" approach to accessing the motor system and rehabilitation at all stages of stroke recovery.¹⁷ Unlike active and passive motor therapies, motor imagery, in principle, is not dependent on residual function but still incorporates voluntary drive. Importantly, in the primate, M1 is directly involved in motor imagery as suggested by direct cellular recordings.¹⁸ In patients with stroke, motor imagery may therefore provide a substitute for executed movement as a means to activate the motor network.

Additional parallels between motor imagery and executed movement are worth mentioning; there is close temporal coupling between motor imagery and executed movement, ie, the time taken to mentally perform an action closely mirrors the executed movement.¹⁹ During imagined movement, the reduction in accuracy with increasing speed (ie, Fitt Law) is maintained²⁰ and the asymmetry between dominant and nondominant hand is also preserved.²¹ Motor imagery produces similar autonomic changes as executed movement, with significant increase in heart and respiratory rates.^22,23,24 Nevertheless, these principles can be disrupted by structural lesions in specific brain areas,²⁵ the implications of which will be discussed later.

In athletes, a structured program of motor imagery movement, motor imagery training, can lead to an improvement in performance. The effects are independent of subclinical muscle activation.²⁶ Perhaps unsurprisingly, motor imagery training produces less practical improvement compared with physical training.²⁷ Although widely used in conjunction with physical training, motor imagery training can independently improve motor performance and produce similar cortical plastic changes,²⁸ providing a useful alternative when physical training is not possible.

As will be discussed, a few early studies suggest that motor imagery training may improve functional recovery after stroke.^29–33 However, how effective it is and in what setting it is best used remains unclear and will form the second section of this review. But in order to effectively target the correct patient population it is imperative to understand how motor imagery training may work, and hence how motor imagery allows access to the motor network. It is widely assumed that motor imagery activates similar pathways as executed movement,³⁴ but as will be seen there are some important differences. Particular attention will be given to whether motor imagery activates M1 because this would imply greatest efficiency, although as will be discussed effects on nonprimary motor areas may also provide gains.

This systematic review of the functional imaging and motor imagery training literature will focus on hand function, because of the lateralized and large cortical representation as well as the obvious key importance of the hand in activities of daily living, and will consist of 4 sections: the first will briefly discuss general cognitive considerations, particularly when investigating motor imagery in the context of structural lesions and relevant methods of assessment; the second will review studies investigating the effectiveness of motor imagery training at restoring upper-limb motor function after stroke; the third will address the cortical network activated during motor imagery in healthy subjects and the potential implications for patient selection. In the fourth section the motor imagery functional imaging studies in patients with stroke will be reviewed. Finally, the results will be discussed and summarized, concluding with suggestions for future research.

	Cognitive Considerations

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
Motor imagery can be defined as a dynamic state during which the representation of a specific motor action is internally reactivated within working memory without any overt motor output, and that is governed by the principles of central motor control.^19,20,35 Operationally, this can be considered as occurring from the 1st person perspective.³⁶ The interested reader is referred to dedicated reviews on the cognitive neuroscience theories of motor imagery.^36,37

Motor Imagery in Stroke
Motor imagery is an integral part of the wider motor system that can be represented by internal models or programs, which develop over time and are consistently changing.³⁸ In chronic disease states temporal coupling is often preserved; in Parkinson disease the asymmetrical bradykinesia is mirrored during motor imagery,³⁹ as is the performance in the chronic fatigue syndrome.⁴⁰ However, in patients with stroke it is important to consider both the ability to perform motor imagery accurately and temporal coupling because, depending on the site and extent of the stroke, either or both may be affected.

Because the parietal lobe is involved in the preservation and generation of a kinaesthetic model,⁴¹ it is perhaps not surprising that parietal damage can reduce motor imagery accuracy.^42,43 By excluding subjects with parietal or premotor lesions, Johnson⁴⁴ used a prehensile task to suggest that both subacute and chronic hemiplegic patients⁴⁵ may still perform motor imagery. Moreover, the authors even suggest that subjects with chronic stroke may be more accurate during motor imagery with the affected limb, a "hemiplegic advantage".

Late disruption of M1 (hand area) by transcranial magnetic stimulation (TMS) can also reduce motor imagery accuracy and increases response time during the mental rotation of hands but not feet; the authors suggest that M1 is not only involved in motor imagery but is required.⁴⁶ Conversely, others have reported that M1 is not necessary to perform motor imagery,²⁵ whereas direct stimulation of M1 via an implanted electrode array may not disrupt motor imagery accuracy but increase response time.⁴⁷

Temporal uncoupling may occur after either parietal²⁵ or frontal lobe damage⁴⁴ but can be preserved after cerebellar stroke.⁴⁸ In partially recovered patients, Malouin et al⁴⁹ speculated that stroke could result in temporal uncoupling of the nonaffected limb while being preserved in the affected limb. However, the authors acknowledge that this unexpected finding could be explained by the extensive strokes included in the study causing either cortical-cortical disruption of the motor imagery network or a deficit in working memory, which in itself may reduce the effectiveness of motor imagery training.⁵⁰ However, Sabate et al⁵¹ reported that temporal coupling can be preserved after stroke.

Despite the limited samples and the wide variety of lesion location in these studies, overall, the ability to perform motor imagery may be preserved after stroke, but it appears that both accuracy and temporal coupling can be disrupted. We have referred to this as chaotic motor imagery.

Chaotic motor imagery can be defined as an inability to perform motor imagery accurately or if having preserved accuracy, demonstration of temporal uncoupling. Further characterization of motor imagery performance in subjects unable to perform accurately or with temporal uncoupling is likely to produce incongruent results. Chaotic motor imagery may be limb-specific, affecting distal but not proximal movement²⁵ probably attributable to differences in cortical organization. Although we primarily use the term "chaotic motor imagery" here to designate inability to perform motor imagery in people with central nervous system damage, it is possible that it also applies to a small percentage of the normal population.

Motor Imagery: Assessment of Ability, Performance and Compliance
It follows from the definition of motor imagery that because of its concealed nature, a subject may surreptitiously use alternative cognitive strategies that, if not screened for, could confound investigations and produce conflicting results. Applied to the normal population these alternative strategies can be distilled into 4 components: the inability to perform motor imagery or inaccuracy; a simple failure to comply; the concealed use of alternative strategies such as counting or visual imagery; and a failure to suppress movement. In addition, when applied to the study of patients with stroke, chaotic motor imagery should be considered.

In order to select appropriate subjects for clinical studies, particularly functional imaging and motor imagery training, a dichotomic approach to motor imagery, able versus unable, is imperative; however, it should be noted that motor imagery is a complex cognitive process.^37,52 As such, important information which further characterizes the quality and vividness of the motor imagery effected by the enrolled subjects performed should be obtained.^22,53

Inability to Perform Motor Imagery
When cognitive screening for the ability to perform motor imagery accurately is applied to a normal population, a small fraction of subjects will fail the assessment. Historically, questionnaires (eg, Mental Imagery Questionnaire (MIQ)⁵⁴) were used which were aimed primarily at athletes, subjectively selecting those who were skilled at motor imagery. The questionnaire requires subjects to perform complex tasks physically and then using motor imagery, after which subjects rate their performance on both a kinaesthetic and visual scale. Because of the difficulties in performing the tasks, a modified MIQ has been produced which comprises of simpler gestures,⁵⁵ though it still does not provide objective evidence of motor imagery ability.

An interesting alternative is the Controllability of Motor Imagery Scale⁵⁶: subjects follow a series of instructions each specifying a single mental movement of a limb; at the end the subject must assume the position produced by the combination of movement. Those unable to do so are excluded, providing objective evidence of motor imagery ability. Although the use of alternative strategies such as visual imagery is not addressed, it remains a more useful assessment than the MIQ.

Perhaps the most appropriate method to assess motor imagery accuracy is to use a prehensile task, similar to Johnson^44,45 because this provides objective evidence and intrinsically activates the motor imagery network.⁵⁷

Simple Failure to Comply
Unlike executed movement paradigms during which compliance can be checked by observation of the task, motor imagery is concealed and requires more complex monitoring. Hanakawa⁵⁸ used visually presented numbers to "guide" subjects through a prelearned finger-tapping sequence. In this procedure, subjects are instructed which finger to begin and then presented a number; the subject then moves the corresponding steps through the sequence, and at end of block is asked to confirm the position. Although this method does provide evidence of compliance, it encourages the use of alternative techniques, particularly counting.

We have used a variant of this method; the motor imagery task is paced and unexpectantly stops and the subject has to confirm their position.⁵⁹ Although this technique would be pragmatically difficult to include into a motor imagery training program, it is of particular interest in functional imaging studies when it can be incorporated into the paradigm, with those who fail to correctly confirm being excluded. Regardless of these details, the task must be of sufficient complexity to contain a number of possible positions so a correct response is unlikely to be attributable to chance.

A similar method is the application of mental chronometry, ie, (re)action times for instance; the number of foot taps in various time blocks can be compared with the number of imagined taps.^50,60,61 This method is particularly useful during motor imagery training,²⁸ providing an objective assessment of compliance; however, in the way it has been used so far it is unable to exclude a noncompliant individual.

Nonetheless, it remains that all of these tasks can easily be solved by the use of alternative strategies such as counting, visual imagery or 3rd person imagery.

Concealed Use of Alternative Strategies
Because the aim of motor imagery is to activate the motor networks, it is crucial that subjects perform the mental task from the 1st person perspective, in contrast to 3rd person perspective or visual imagery.³⁶ Likewise, they should also be instructed not to count, as also by doing so a subject could fool an exclusion criteria based on confirmation of position. This has not been addressed in the literature but a form of objective assessment is clearly needed.

Failure to Suppress Movement
Several methods can be used to monitor overt or covert movement during motor imagery. The electromyogram (EMG) has been used during training before the functional imaging sessions^62,63 to ensure subjects do not move. This presents practical challenges during motor imagery training and functional MRI (fMRI), although it is possible in the latter⁵⁸ and during positron-emission tomography (PET) studies. Close inspection of the subject is the simplest way to highlight overt movement, though once again this is difficult in the MRI bore and small amplitude movements may be missed. Alternatively, a dynamometer,⁶⁴ an accelerometer⁶⁵ or goniometer⁶⁶ could be used in the fMRI session but not easily during motor imagery training. Specifically for the fMRI sessions, video observation may present an alternative method. We have successfully used fiber-optic MR-friendly gloves.⁵⁹

Chaotic Motor Imagery
The identification of subjects performing chaotic motor imagery is not addressed by any published assessments, yet this is possibly a very important confound in both the interpretation of previous studies^50,51 and the application of motor imagery training and motor imagery to patients with stroke. Crucially, the potential screening tools discussed above would not be suitable for subjects performing chaotic motor imagery. Moreover, including subjects who perform chaotic motor imagery into a motor imagery training program may dilute any positive affects.

Furthermore, a lack of more general cognitive ability (dysphasia, dementia, neglect or inattention) would make an assessment of motor imagery ability very difficult or impossible.

For these reasons at this stage, theoretically a subject whose stroke is subcortical and does not involve M1 or the parietal lobe are likely to benefit the most from early motor imagery training. However, it is possible that many other factors may influence motor imagery training, such as gender, handedness, affected hemisphere, dysphasia, precise lesion location and time elapsed since stroke.

	Review of the Literature

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
Motor Imagery Training Effectiveness
We will review studies investigating the effectiveness of motor imagery training at restoring motor function after stroke.

Inclusion Criteria
A systematic literature search using the electronic database on PubMed was performed. Only studies that focus exclusively on motor imagery of the upper-limb function and rehabilitation after stroke will be considered. Although other articles have been published^67–75 only 5 fulfilled the criteria.^29,30–33

Clinical Data
Details of the number and characteristics of subjects included in the 5 primary studies are given in Table 1. Age and gender ratio are similar across studies but there are large differences in the time since stroke.

Lesion location and affected hemisphere is poorly reported across all the studies, as is any clinical classification or description of the stroke. Motor function is not always described in sufficient detail; of note is that all the studies excluded subjects with hemiplegia. General cognitive assessment was performed in 4 studies, 3^29,30,33 using the Mini Mental State Examination. It is likely that with such a wide range of cognitive ability combined with a lack of topographic data, a portion of subjects will have been performing chaotic motor imagery.

Study Design, Interventions and Outcome Measures
The details of design of the 5 studies are outlined in Table 2. Three studies^29,32,33 were randomized controlled trials although sample size was not based on a power calculation. Where applicable, groups were reasonably balanced, with no significant differences in motor performance. Three studies^29,32,33 used a blinded assessor to undertake the outcome measures.

All studies provided a control intervention except 1³⁰ whose replicated case design removed the need. Both the control intervention and motor imagery training were provided in addition to a background rehabilitation therapy that in 2 studies^31,33 was related to the motor imagery training. Only 2 studies^31,33 specifically instructed subjects to perform motor imagery; the remaining were either vague or encouraged visual imagery.

Both the control intervention and motor imagery training were balanced for frequency and duration. Across the studies the control intervention varied markedly, as did the motor imagery training; of note 1 study³² used a motor imagery training program based on the functional retraining program used as a control intervention, thus limiting inclusion to stroke patients with a degree of movement.

Outcome measures varied between studies limiting direct comparison, although all measured motor function of the paretic upper limb. In addition 2 studies^31,32 used the trained task (task used in motor imagery) and an untrained task to assess outcome.

Subject Performance
Only 1 study²⁹ assessed motor imagery ability using the MIQ; however, this was not used in their subsequent study.³³ None of the studies assessed the use of alternative strategies, subtle movement or monitored compliance, which may be particularly relevant for instructions delivered via prerecorded tape.

Findings
The heterogeneity in subject characteristics, interventions and outcome measures across the 5 studies precludes any general conclusion, but overall all studies found that motor imagery training had encouraging effects on motor function compared with the control condition. One study³² found a significant advantage for the motor imagery training group for both untrained and trained tasks, maintained at 1 month, whereas another study³¹ found a significant advantage only for the trained task. In addition, 1 study²⁹ demonstrated an improved outcome in the motor imagery training group (no probability value given) on the Fugl-Meyer and Action Research Arm Test scores 1 week after end of training, and another³⁰ showed a significant change (no probability value given) in Motricity Index in 8 of 9 subjects during the intervention. Further, still another study³³ confirmed a significant increase in the Action Research Arm Test immediately after training in the motor imagery training group. In addition to showing that motor imagery had positive effects on motor function, these studies also suggest that the effects might generalize beyond the tasks being trained. With only 1 study³² following-up subjects after 1 month after the training period, it is unclear, however, whether motor imagery training is able to produce long-term gains.

Functional Imaging Studies
Healthy Subjects
A systematic search on PubMed was performed. For the purpose of this review, to explore the neural mechanism of motor imagery and allow an appropriate frame of reference, only studies that contrasted motor imagery of a hand action with rest, as opposed to another more or less complex cognitive task,^76–84 will be included in this review (also see Discussion section). Furthermore, to allow meaningful comparison only articles with published stereotactic coordinates based on whole-brain voxel-based analysis of motor imagery contrasted with rest will be analyzed in detail. Although numerous studies have been published^{28,34,52,60,61,85–91} only 5^{92,62,63,93,94} fulfilled these criteria. However, 6 otherwise relevant Region of Interest (ROI)–based studies^{58,66,95–97} will be also commented on.

Main Studies
   General Features.
Demographics are given in Table 3. The objective of the studies varied; in only one study⁶² was it the primary aim to investigate the motor imagery network. Of note 1 study⁹⁴ included subjects who were not completely right-handed; as will be seen, this is of particular relevance when considering cortical activation patterns.

   Functional Imaging Paradigm.
The activation task and pacing varied markedly across the studies; of note 1 study⁹² used a motor imagery task that required the subject to imagine their arm outstretch while performing the finger task. Only 1 study⁶² explicitly states that the auditory stimuli were continued during rest.

   Subject Performance.
Table 4 documents the methods each study used to tackle the difficult cognitive challenges. Overall, the methods used were sparse. Instructions to subjects were limited; only 1 study⁹² was specific and only 1⁹³ asked subjects not to perform visual imagery or count. All the studies except 1⁹³ attempted to assess motor imagery ability; however, this particular study was unique in that they monitored compliance during the imaging paradigm. This was achieved by asking subjects to confirm their position after the finger-tapping sequence during both the training and the positron-emission tomography session. Interestingly, the block duration was fixed (50 seconds) but the tapping sequence varied in length ensuring a different end position, yet whether the subject was informed of this is unclear. The remaining studies had no confirmation of performance. No study assessed alternative strategies. All studies, bar 1,⁹⁴ monitored for subtle movement in either the task training, imaging paradigm or both.

   Data Analysis.
All the studies used versions of the Statistical Parametric Mapping (SPM) software, either SPM 96^62,92 or SPM 99.^63,93,94 Two studies^62,94 used random effect analysis, with the remaining using fixed effects analysis. Statistical thresholds varied; 2 studies^63,92 used P<0.01 (corrected), 2^62,94 P<0.001 (corrected) and 1⁹³ P<0.001(uncorrected).

   Activation Patterns.
When needed, the published activation coordinates were transformed into Talairach space using Mni2tal⁹⁸ and then checked manually against the Talairach atlas. Table 5 summarizes cortical activation patterns in M1, the secondary motor structures and motor-related areas.

Only 1 study,⁹⁴ documents M1 activation during motor imagery, which was bilateral and localized to the hand area; however, the activation was less than during executed movement. The use of random effect analysis on only 8 subjects, as performed in 1 study,⁶² is conservative and may have contributed to the failure to demonstrate M1 activation, particularly as they report that 50% of the subjects showed M1 activation on individual analysis (precise location not reported). Dorsal premotor was activated in all studies bar 1,⁹⁴ whereas 1 study⁹³ reported ventral premotor activation. Two studies^62,93 reported S1 activation and 3 (pre) supplementary motor cortex activation.^63,92,94 Three studies^62,93,94 demonstrated dorsolateral prefrontal cortex activation. The contralateral inferior parietal lobule was consistently activated as was the ipsilateral cerebellum (with 2 exceptions^62,92).

ROI Studies
   General Features.
The 6 relevant ROI-based studies are detailed in Table 6. The objective of the studies varied; 3 studies^58,96,99 focused primarily on aspects of the motor imagery network. The studies are similar with regards to subject demographics, except the inclusion of left-handed subjects by 2 studies.^66,96 Across the studies, motor imagery ability was not assessed and 2 studies^96,99 did not explicitly instruct subjects to perform motor imagery. During the training phase, different methods were used to detect overt movement: observation,^95,97 goniometer,⁶⁶ and EMG.⁹⁹ In this particular study⁹⁹ the subjects continued training until they subjectively scored >4 on a kinaesthetic scale with a silent EMG.

   Design.
All studies used fMRI and either auditory or visual pacing. Of note, 1 study⁵⁸ used MRI-compatible EMG to detect subtle movement, and moreover objectively assessed compliance. Subjects learned the simple finger tapping sequence and at the beginning of each block were presented with a visual number which corresponded to the start position, after which they confirmed (using a button box) their position.

   ROIs.
All studies report prior hypotheses to justify the ROI analysis. There is a wide variation among studies in the ROIs used. One study⁵⁸ used functional ROIs, 2 studies^96,99 created ROIs from anatomical landmarks; as a result the anterior border of M1 was defined as the precentral sulcus. This will have included premotor areas,¹⁰⁰ which may be of relevance. The statistical thresholds and methods varied across studies and were generally liberal with 3 studies using signal based analysis.^58,66,96

   Results.
The results are shown in Table 6. Two studies^66,96 reported contralateral M1 activation and 1 suggested bilateral activation,⁹⁹ Although 1 study⁹⁷ found activation of the contralateral precentral gyrus, it was anterior to the M1 hand area. Likewise, 1 study⁹⁵ suggests that motor imagery failed to activate contralateral M1, activating ipsilateral precentral gyrus (dorsal premotor) instead. This is consistent with their earlier study⁵⁸ that failed to report activation in M1, although they reported activation in 1 of the 10 subjects studied.

Overall, the findings are in keeping with the previously presented literature suggesting robust premotor activation but weak or fleeting activation of M1 during motor imagery. Consistent with this, in the 3 studies that reported M1 activation during motor imagery, when the activation was compared with executed movement it was found to be much reduced^96,99 and in 1 instance, similar,⁶⁶ but in this study subjects we not monitored adequately for subtle movement during the imaging session.

Stroke Patients
A detailed systematic search on PubMed was performed. The search revealed only 1 study in stroke.¹⁰¹ Lehéricy et al¹⁰¹ used fMRI and motor imagery to examine the role of prefrontal, premotor and motor cortex in secondary dystonia. Although the study does not assess the relationship between motor imagery and hand motor recovery, it does raise some methodological considerations relevant to the review. The cortical activation patterns will not be described in detail.

The study compared 6 patients with adult-onset subcortical stroke and unilateral dystonia to 7 control subjects. Both controls and patients were screened using the MIQ. No further details of current functional level were given; parietal lobe dysfunction was suggested but not described in detail. The same paradigm as Gerardin et al⁶² was used (see above). Subjects were observed not to move during training and there was no monitoring within the paradigm. Overall, the authors suggest an over-activation of the premotor and parietal areas during motor imagery of the dystonic hand compared with controls, in whom they report less sensorimotor cortex and more prefrontal activation during motor imagery than executed movement, but no further details are reported.

	Discussion

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
Motor Imagery Training
The findings of the 5 studies reviewed suggest that motor imagery training might have an encouraging effect on motor function after stroke. However, interpretation of these results is limited by small sample sizes and heterogeneity in subject characteristics, motor imagery interventions and outcome measures used. In addition it is likely that some subjects may have been performing chaotic motor imagery as a result of the lesions involving the motor imagery network, yet despite this they still managed to achieve a positive effect on functional performance. This does raise an interesting paradox: motor imagery training, which is less effective than physical practice in controls,²⁷ may produce greater functional gains than executed movement in patients with stroke who can move, as claimed by Liu et al.³² This observation being isolated, however, it should be considered with caution. Furthermore, a possible explanation was alluded to by Crosbie et al³⁰: subjects often continue motor imagery training unsupervised, increasing the frequency of intervention. Alternatively, it is conceivable that patients performing chaotic motor imagery are more sensitive to and have the most to gain from training. Clearly more research is needed in this area.

Although motor imagery training is theoretically independent of motor recovery, in the studies reviewed the control intervention required motor performance. In doing so, they excluded subjects with severe hemiparesis, who may have a different response to motor imagery training. For instance, 1 intriguing paradigm could be to use motor imagery at the stage of complete hand paresis so as to activate the motor network as a substitute to CIT, and to switch to modified CIT as soon as sufficient hand movement has returned.

Furthermore, the optimal period to start motor imagery training is unclear. Unlike AMT, motor imagery could potentially be started early after stroke when the brain may be most "receptive" to intervention, although it is becoming increasingly clear that active motor training may be effective even in the chronic stage of stroke. Nonetheless, whether motor imagery training is capable of producing long-term gains is unknown.

In summary, motor imagery training is a promising intervention to improve motor function after stroke, but adequately powered clinical trials are needed in groups of well-characterized, stroke subjects. Furthermore, adequate screening and monitoring tools should be used to ensure that subjects included in these trials are able to and do perform motor imagery. Finally, the most appropriate target population, ie, those with severe hemiparesis, have not been studied as yet.

Functional Imaging
The findings from the 5 main studies detailed here highlighted robust and consistent activation of the secondary motor network and ipsilateral cerebellum during motor imagery, but only weak or inconsistent M1 activation.

To what extent these results are affected by methodological pitfalls is unclear. None of the studies reviewed complied with all the quality controls discussed above (see also implications for future studies below). Across the studies, subject screening and monitoring during both the training and imaging paradigm varied markedly in its purpose, with no study adequately assessing all facets of noncompliance. Hanakawa et al⁵⁸ was the closest to optimal by including monitoring of performance during scanning, yet even with the additional statistical powering of ROI analysis found no M1 activation whatsoever in 9 of their 10 subjects. In contrast, Lacourse et al⁹⁴ used whole-brain analysis and documented strong bilateral M1 activation; however, some of their subjects were left-handed. Moreover, attributable to the lack of appropriate screening/monitoring, the validity of the results is uncertain.

Although the use of rest (with cues on) as a baseline condition provides an appropriate contrast and eases comparison between paradigms, it can be difficult to control.¹⁰² Nonetheless, motor imagery studies that use cognitive baseline conditions rather than rest, which are difficult to place in context, are also undecided about the role of M1. Using visual imagery as a baseline, Porro et al^81,82 and Kuhtz-Buschbeck et al⁷⁶ report M1 activation, although significantly less than during executed movement; conversely, Dechent et al⁷⁸ used the same paradigm but failed to show M1 activation, although suggested fleeting M1 activation on the initiation of motor imagery. Nonetheless, by contrasting body parts specific activation, Ehresson et al⁸⁸ have suggested that motor imagery engages somatotopically organized areas of the M1, as did also Stippich et al.⁸⁷ Alternative methods of accessing the motor imagery network such as Luria hand rotation task⁵⁷ activate secondary motor structures but not M1.

The repeated failure to consistently activate M1 in functional imaging studies when other investigative methods, such as magnetoencephalography (MEG),^103,104 electroencephalogram (EEG),^105,106 and TMS,^107–110 are supportive of its involvement is puzzling. The higher spatial resolution of fMRI may be a contributing factor. However, it is plausible that during motor imagery, M1 has a different role beside executive, and so its activation is less sustained than during actual movement, producing a much-reduced signal. Importantly, in the primate, M1 is involved in spatial processing during motor imagery.¹⁸ In addition, M1 can participate in the processing of both serial order information^111,112 and task-related spatial information,¹¹³ whereas in humans it can be modulated by attention¹¹⁴ and has the capacity to store short-term procedural information,¹¹⁵ implying motor imagery may involve generation of motor representations without movement.⁹²

Yet, despite differences in both subject selection and baseline tasks, the secondary motor structures, normally involved in motor learning, preparation, programming and memorizing,³⁶ have consistently shown activation. Therefore, an alternative explanation for the weak and inconsistent M1 activation during motor imagery may be that the premotor areas inhibit an otherwise executive M1. In primates, premotor areas can modulate M1 in an inhibitory manner as well as excitatory.¹¹⁶ TMS studies in humans suggest that premotor-M1 interactions can also be inhibitory.^117,118 Furthermore, connectivity analysis of fMRI data sets supports an inhibitory effect of supplementary motor cortex and superior parietal lobe on M1 during motor imagery.⁹⁶ Although inhibition by secondary motor areas at the level of the anterior horn cell has been considered,³⁶ this would be expected to be associated with high or even excessive M1 activation.

Because it is widely believed that effective training and rehabilitation methods after stroke should involve M1, does the apparently weak and inconsistent activation of M1 during motor imagery make it a doomed rehabilitative approach? Not necessarily so, for 3 distinct reasons. Firstly, this finding needs to be confirmed in methodologically flawless fMRI studies. Secondly, even a weak activation might be sufficient to prevent learned nonuse and "prime" the motor representations enough to result in minimal active movement, bridging the gap across to modified CIT.¹¹⁹ Thirdly, it remains possible that any effects of motor imagery are mediated by the premotor areas bypassing M1 via direct corticospinal projections, or improving preparation and programming rather than execution.²⁸ Further speculation about the role of M1 and the secondary motor areas in healthy volunteers is difficult until the issues of subject selection and noncompliance have been dealt with.

The study of secondary dystonia does not clarify the involvement of M1; nonetheless, it highlights some of the difficulties in applying motor imagery to patients with stroke: (1) patients had parietal dysfunction and may have been performing chaotic motor imagery even though they still scored well on the MIQ; (2) the potential difficulties in interpreting cortical activation patterns when subject compliance cannot be objectively confirmed; and (3) dystonia after adult onset stroke is a very unusual condition whose relevance to stroke in general is uncertain.

An interesting parallel are motor imagery studies in patients with complete spinal cord injury (SCI), where the lesion resembles subcortical stroke to a greater extent. Functional imaging^120,121 studies in SCI patients have reported marked M1 activation during motor imagery of the feet. Alkadhi et al¹²⁰ surprisingly reported a similar M1 cluster size in SCI patients as during executed movement in controls, who, notably, failed to activate M1 during motor imagery. This provides further support that the premotor areas may be inhibiting an otherwise executive M1 during motor imagery. Interestingly, healthy volunteers were screened with the MIQ, but SCI subjects were unable to perform the movements required by the MIQ, highlighting the need for alternative methods of assessment. It is possible, however, that motor imagery performed by SCI subjects differs in some fundamental aspects from that performed by healthy subjects.

	Implications for Study Design

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
Subject Selection
Based on the above review, subjects, particularly those with stroke, should have an objective assessment of their capacity to perform motor imagery and exclude chaotic motor imagery before inclusion into either motor imagery training or functional imaging studies. They should also be given clear instructions; these should include specifying 1st person motor imagery and advice against the use of alternative strategies such as counting or visual imagery. Attempts should be made to identify noncompliant subjects.

Motor Imagery Training
Because motor imagery is, in principle, independent of motor performance, it would seem prudent that the control intervention for future motor imagery training studies does not require motor performance. This will further characterize the effects of motor imagery training, as will monitoring for uncontrolled motor imagery training use. Pragmatically, monitoring for noncompliance during motor imagery training is challenging; however, the use of mental chronometry²⁸ is likely to be an effective means. Finally, further research should compare the effects of intervention in groups of low and high performers, to test the "bridging" hypothesis.

Functional Imaging Paradigm Design
Motor imagery may prove to be a consistent activation task that can be used across the full range of recovery and is largely independent of motor performance. The activation task should be simple and quick enough to be completed by patients with stroke while allowing objective evidence of compliance. To allow meaningful interpretation, particularly in subjects with stroke, the baseline condition should be rest (with cues).

	Conclusions

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 
The literature reviewed suggests that in healthy subjects motor imagery activates the motor network, providing a plausible mechanism for motor imagery training in athletes. No useful information is available as yet in patients with stroke, and the underlying mechanism of motor imagery training-induced improved performance suggested by the literature remains unexplained. However, if the same relationship as SCI patients is maintained in stroke, this may suggest that patients with chronic hemiplegia would involve M1 to a greater degree during motor imagery, providing a potential target group for motor imagery training.

	Acknowledgments

 
Sources of Funding

N.S. was supported by a Brain Entry Scholarship, The Stroke Association (TSA 2003/10) & Sackler Fellowship. V.P. was supported by the Wellcome Trust. J.C.B. was supported by the Medical Research Council (MRC G0001219).

Disclosure

None.

Received February 20, 2006; revision received March 24, 2006; accepted April 10, 2006.

	References

Top Abstract Introduction Cognitive Considerations Review of the Literature Discussion Implications for Study Design Conclusions References

 

1. Duncan PW GL, Matchar D, Divine GW, Feussner J. Measurement of motor recovery after stroke. Outcome assessment and sample size requirements. Stroke. 1992; 23: 1084–1089.[Abstract/Free Full Text]
2. Liepert J, Bauder H, Miltner WHR, Taub E, Weiller C. Treatment-induced cortical reorganization after stroke in humans. Stroke. 2000; 31: 1210–1216.[Abstract/Free Full Text]
3. Nudo RJ WB, SiFuentes F, Milliken GW. Neural substrates for the effects of rehabilitative training on motor recovery after ischemic infarct. Science. 1996; 272: 1791–1794.[Abstract]
4. Ward NS CL. Mechanisms underlying recovery of motor function after stroke. Arch Neurol. 2004; 61: 1844–1848.[Abstract/Free Full Text]
5. Calautti C, Baron J-C. Functional neuroimaging studies of motor recovery after stroke in adults: A review. Stroke. 2003; 34: 1553–1566.[Abstract/Free Full Text]
6. Johansen-Berg H, Dawes H, Guy C, Smith SM, Wade DT, Matthews PM. Correlation between motor improvements and altered fMRI activity after rehabilitative therapy. Brain. 2002; 125: 2731–2742.[Abstract/Free Full Text]
7. Carey JR, Kimberley TJ, Lewis SM, Auerbach EJ, Dorsey L, Rundquist P, Ugurbil K. Analysis of fMRI and finger tracking training in subjects with chronic stroke. Brain. 2002; 125: 773–788.[Abstract/Free Full Text]
8. DeBow SB MJ, Kolb B, Colbourne F. Immediate constraint-induced movement therapy causes local hyperthermia that exacerbates cerebral cortical injury in rats. Can J Physiol Pharmacol. 2004; 82: 231–237.[CrossRef][Medline] [Order article via Infotrieve]
9. Dromerick AW, Edwards DF, Hahn M. Does the application of constraint-induced movement therapy during acute rehabilitation reduce arm impairment after ischemic stroke? Stroke. 2000; 31: 2984–2988.[Abstract/Free Full Text]
10. Page SJ, Sisto S, Johnston MV, Levine P, Hughes M Modified constraint-induced therapy in subacute stroke: A case report. Arch Phys Med Rehabil. 2002; 83: 286–290.[CrossRef][Medline] [Order article via Infotrieve]
11. Grotta JC, Noser EA, Ro T, Boake C, Levin H, Aronowski J, Schallert T. Constraint-induced movement therapy. Stroke. 2004; 35: 2699–2701.[CrossRef][Medline] [Order article via Infotrieve]
12. Fritz SL CY, Malcolm MP, Patterson TS, Light KE. Feasibility of electromyography-triggered neuromuscular stimulation as an adjunct to constraint-induced movement therapy. Phys Ther. 2005; 85: 428–442.[Abstract/Free Full Text]
13. Page SJ, Levine P. Back from the brink: Electromyography-triggered stimulation combined with modified constraint-induced movement therapy in chronic stroke. Arch Phys Med Rehabil. 2006; 87: 27–31.[CrossRef][Medline] [Order article via Infotrieve]
14. Nelles G, Spiekermann G, Jueptner M, Leonhardt G, Muller S, Gerhard H, Diener HC. Reorganization of sensory and motor systems in hemiplegic stroke patients: A positron-emission tomography study. Stroke. 1999; 30: 1510–1516.[Abstract/Free Full Text]
15. Lotze M, Braun C, Birbaumer N, Anders S, Cohen LG. Motor learning elicited by voluntary drive. Brain. 2003; 126: 866–872.[Abstract/Free Full Text]
16. Cramer SC, Lastra L, Lacourse MG, Cohen MJ. Brain motor system function after chronic, complete spinal cord injury. Brain. 2005; 128: 2941–2950.[Abstract/Free Full Text]
17. Jackson PL, Lafleur MF, Malouin F, Richards C, Doyon J. Potential role of mental practice using motor imagery in neurologic rehabilitation. Arch Phys Med Rehabil. 2001; 82: 1133–1141.[CrossRef][Medline] [Order article via Infotrieve]
18. Georgopoulos AP LJ, Petrides M, Schwartz AB, Massey JT. Mental rotation of the neuronal population vector. Science. 1989; 243: 234–236.[Abstract/Free Full Text]
19. Decety J, Jeannerod M, Prablanc C. The timing of mentally represented actions. Behav Brain Res. 1989; 34: 35–42.[Medline] [Order article via Infotrieve]
20. Decety J, Jeannerod M. Mentally simulated movements in virtual reality: Does Fitts’s law hold in motor imagery? Behav Brain Res. 1995; 72: 127–134.[CrossRef][Medline] [Order article via Infotrieve]
21. Maruff P, Wilson PH, De Fazio J, Cerritelli B, Hedt A, Currie J. Asymmetries between dominant and non-dominant hands in real and imagined motor task performance. Neuropsychologia. 1999; 37: 379–384.[CrossRef][Medline] [Order article via Infotrieve]
22. Roure R, Collet C, Deschaumes-Molinaro C, Delhomme G, Dittmar A, Vernet-Maury E. Imagery quality estimated by autonomic response is correlated to sporting performance enhancement. Physiol Behav. 1999; 66: 63–72.[CrossRef][Medline] [Order article via Infotrieve]
23. Jeannerod M, Frak V. Mental imaging of motor activity in humans. Curr Opin Neurobiol. 1999; 9: 735–739.[CrossRef][Medline] [Order article via Infotrieve]
24. Kazuo Oishi TKaTM. Autonomic response specificity during motor imagery. Journal of Physiological Anthropology and Applied Human Science. 2000; 19: 255–261.[CrossRef][Medline] [Order article via Infotrieve]
25. Sirigu A, Duhamel JR, Cohen L, Pillon B, Dubois B, Agid Y. The mental representation of hand movements after parietal cortex damage. Science. 1996; 273: 1564–1568.[Abstract]
26. Zijdewind I, Toering ST, Bessem B, Van Der Laan O, Diercks RL. Effects of imagery motor training on torque production of ankle plantar flexor muscles. Muscle Nerve. 2003; 28: 168–173.[CrossRef][Medline] [Order article via Infotrieve]
27. Pascual-Leone A, Nguyet D, Cohen LG, Brasil-Neto JP, Cammarota A, Hallett M. Modulation of muscle responses evoked by transcranial magnetic stimulation during the acquisition of new fine motor skills. J Neurophysiol. 1995; 74: 1037–1045.[Abstract/Free Full Text]
28. Jackson PL, Lafleur MF, Malouin F, Richards CL, Doyon J. Functional cerebral reorganization following motor sequence learning through mental practice with motor imagery. Neuroimage. 2003; 20: 1171–1180.[CrossRef][Medline] [Order article via Infotrieve]
29. Page SJ, Levine P, Sisto S, Johnston MV. A randomized efficacy and feasibility study of imagery in acute stroke. Clin Rehabil. 2001; 15: 233–240.[Abstract/Free Full Text]
30. Crosbie JH MS, Gilmore DH, Wiggam MI. The adjunctive role of mental practice in the rehabilitation of the upper limb after hemiplegic stroke: A pilot study. Clin Rehabil. 2004; 18: 60–68.[Abstract/Free Full Text]
31. Dijkerman HC LM, Johnston M, MacWalter RS. Does motor imagery training improve hand function in chronic stroke patients? A pilot study. Clin Rehabil. 2004; 18: 538–549.[Abstract/Free Full Text]
32. Liu KP, Chan CC, Lee TM, Hui-Chan CW. Mental imagery for promoting relearning for people after stroke: A randomized controlled trial. Arch Phys Med Rehabil. 2004; 85: 1403–1408.[CrossRef][Medline] [Order article via Infotrieve]
33. Page SJ, Levine P, Leonard AC. Effects of mental practice on affected limb use and function in chronic stroke. Arch Phys Med Rehabil. 2005; 86: 399–402.[CrossRef][Medline] [Order article via Infotrieve]
34. Grezes JDJ. Functional anatomy of execution, mental simulation, observation, and verb generation of actions: A meta-analysis. Hum Brain Mapp. 2001; 12: 1–19.[CrossRef][Medline] [Order article via Infotrieve]
35. Decety J, Grezes J. Neural mechanisms subserving the perception of human actions. Trends in Cognitive Sciences. 1999; 3: 172–178.[CrossRef][Medline] [Order article via Infotrieve]
36. Jeannerod M. Mental imagery in the motor context. Neuropsychologia. 1995; 33: 1419–1432.[CrossRef][Medline] [Order article via Infotrieve]
37. Berthoz A. The role of inhibition in the hierarchical gating of executed and imagined movements. Brain Res Cogn Brain Res. 1996; 3: 101–113.[CrossRef][Medline] [Order article via Infotrieve]
38. Wolpert DM GZ. Computational principles of movement neuroscience. Nat Neurosci. 2000; 3: Suppl: 1212–1217.[CrossRef][Medline] [Order article via Infotrieve]
39. Dominey P, Decety J, Broussolle E, Chazot G, Jeannerod M. Motor imagery of a lateralized sequential task is asymmetrically slowed in hemi-Parkinson’s patients. Neuropsychologia. 1995; 33: 727–741.[CrossRef][Medline] [Order article via Infotrieve]
40. de Lange FP, Kalkman JS, Bleijenberg G, Hagoort P, vd Werf SP, van der Meer JWM, Toni I. Neural correlates of the chronic fatigue syndrome-an fMRI study. Brain. 2004; 127: 1948–1957.[Abstract/Free Full Text]
41. Sirigu A, Daprati E, Pradat-Diehl P, Franck N, Jeannerod M. Perception of self-generated movement following left parietal lesion. Brain. 1999; 122 (Pt 10): 1867–1874.[Abstract/Free Full Text]
42. Tomasino B, Toraldo A, Rumiati RI. Dissociation between the mental rotation of visual images and motor images in unilateral brain-damaged patients. Brain and Cognition. 2003; 51: 368–371.[CrossRef][Medline] [Order article via Infotrieve]
43. Tomasino B, Rumiati RI, Umilta CA. Selective deficit of motor imagery as tapped by a left-right decision of visually presented hands. Brain Cogn. 2003; 53: 376–380.[CrossRef][Medline] [Order article via Infotrieve]
44. Johnson SH. Imagining the impossible: Intact motor representations in hemiplegics. Neuroreport. 2000; 11: 729–732.[Medline] [Order article via Infotrieve]
45. Johnson SH, Sprehn G, Saykin AJ. Intact motor imagery in chronic upper limb hemiplegics: Evidence for activity-independent action representations. J Cogn Neurosci. 2002; 14: 841–852.[Abstract/