Effects of motor imagery training after chronic, complete spinal cord injury

Abnormalities in brain motor system function are present following spinal cord injury (SCI) and could reduce effectiveness of restorative interventions. Motor imagery training, which can improve motor behavior and modulate brain function, might address this concern but has not been examined in subjects with SCI. Ten subjects with SCI and complete tetra-/paraplegia plus ten healthy controls underwent assessment before and after 7 days of motor imagery training to tongue and to foot. Motor imagery training significantly improved the behavioral outcome measure, speed of movement, in non-paralyzed muscles. Training was also associated with increased fMRI activation in left putamen, an area associated with motor learning, during attempted right foot movement in both groups, despite foot movements being present in controls and absent in subjects with SCI. This fMRI change was absent in a second healthy control group serially imaged without training. In subjects with SCI, training exaggerated, rather than normalized, baseline derangement of left globus pallidus activation. The current study found that motor imagery training improves motor performance and alters brain function in subjects with complete SCI despite lack of voluntary motor control and peripheral feedback. These effects of motor imagery training on brain function have not been previously described in a neurologically impaired population, and were similar to those found in healthy controls. Motor imagery might be of value as one component of a restorative intervention.

groups, despite foot movements being present in controls and absent in subjects with SCI. This fMRI change was absent in a second healthy control group serially imaged without training. In subjects with SCI, training exaggerated, rather than normalized, baseline derangement of left globus pallidus activation. The current study found that motor imagery training improves motor performance and alters brain function in subjects with complete SCI despite lack of voluntary motor control and peripheral feedback. These eVects of motor imagery training on brain function have not been previously described in a neurologically impaired population, and were similar to those found in healthy controls. Motor imagery might be of value as one component of a restorative intervention.
Keywords Motor system · Imagery · Plasticity · Putamen · Globus pallidus · Spinal cord injury Spinal cord injury (SCI) is a common neurological condition associated with substantial motor disability. Several therapeutic approaches are under investigation towards reducing disability after SCI. Some focus on SCI repair (Raineteau et al. 2002;Dobkin and Havton 2004), while others aim to drive muscles or devices with signals derived from cortical recordings (Carmena et al. 2003;Friehs et al. 2004;Scherberger et al. 2005).
Such approaches generally assume intact brain motor systems for driving limb movements. However, although brain mapping studies have found preservation of many normal features of motor system function after SCI, abnormalities of brain motor system function have also been described. Examples include both sub-and supra-normal activation in motor system foci (Sabbah et al. 2002;Alkadhi et al. 2005;Cramer et al. 123 2005), higher threshold and longer latency for motor evoked potentials to muscles below an incomplete lesion (Davey et al. 1998;Curt and Dietz 1999), and enhanced excitability of motor pathways targeting muscles above the level of SCI (Topka et al. 1991). These abnormalities likely represent eVects of impaired corticospinal tract integrity and peripheral feedback, as well as reactive/adaptive changes in brain function.
Normalization of brain motor system function might be useful as an adjunct to therapies aiming to restore movement after SCI, and motor imagery training might be one means to achieve this. This form of training involves internal reactivation of the representation of a speciWc motor action without any overt motor output, and is governed by the principles of central motor control (Sharma et al. 2006). Motor imagery can modify brain function and improve motor performance in healthy controls; in subjects with neurological impairment, motor imagery can improve motor performance but little is known regarding eVects on brain function (Pascual-Leone et al. 1995;Gerardin et al. 2000;Page et al. 2001;Jackson et al. 2003Jackson et al. , 2004Crosbie et al. 2004;Dijkerman et al. 2004;Johnson-Frey 2004;Lacourse et al. 2004aLacourse et al. , 2005Page et al. 2005). Ability to perform motor imagery is preserved after SCI (Decety and Boisson 1990;Lacourse et al. 1999), though the capability of motor imagery to modify motor behavior or brain motor system function in subjects with SCI has not been studied to date.
The current study examined the eVects of a 7-day course of motor imagery training on a body area that did (tongue) and an area that did not (right foot) retain voluntary motor control after SCI. The content of training was imagined movement of tongue, and separately of right foot, through a 5-part sequence using audiovisual cueing provided by an in-home computer. Assessments were made before and after training. The primary behavioral outcome measure was the rate with which this same 5-part sequence was actually performed. Functional MRI was used to probe brain function. Though the intervention revolved around imagined movement, the fMRI task employed attempted movement given that the long-term goal is to improve brain function related to moving the plegic limb. Neurophysiological assessments were made using transcranial magnetic stimulation (TMS).
We hypothesized that behavioral eVects of motor imagery would be similar among subjects with complete SCI and healthy controls for tongue movements, as imagery was expected to remain intact for supralesional muscles. For foot movements, imagery eVects on behavior were hypothesized to be diVerent among these two subject groups, with gains in controls but no change in subjects with SCI-imagery was not expected to modify chronic complete plegia, but when used as an adjunct, the eVects of imagery on brain function might improve the impact of other restorative interventions.
The main focus of this study revolved around the hypothesis that motor imagery training would induce similar changes in brain motor system activation during attempted movement among controls and subjects with SCI, despite absence of voluntary movement in the latter. This was based on prior studies in healthy subjects showing that motor system activation can be modiWed by a 1-week course of motor imagery practice, i.e., by an intervention that lacked physical movement rehearsal (Pascual-Leone et al. 1995;Jackson et al. 2003;Lacourse et al. 2004a).
As a corollary, we further hypothesized that motor imagery training would reduce abnormalities of brain motor system function seen at baseline in subjects with SCI, a question of particular interest given the paucity of data examining eVects of imagery training on brain function in a neurologically impaired population. This hypothesis arose from the fact that motor imagery is known to improve motor behavior, and that in healthy subjects much of this improvement arises from changes in brain function. Brain motor system function is known to be deWcient after SCI, therefore the ability of motor imagery to induce improvements was examined. The main focus was on contralateral putamen, thalamus, primary sensory, and primary motor cortex, given the key nature of these areas in movement generation and learning (Jueptner and Weiller 1998;Milliken et al. 1999;Maillard et al. 2000;Scholz et al. 2000;Doyon et al. 2003;Cramer et al. 2005). Motor imagery training was also hypothesized to reduce TMS thresholds for foot motor system activation over time based on a previous study of hand motor imagery eVects in healthy subjects (Pascual-Leone et al. 1995).

Subjects
Subjects with SCI had injury between C5 and T10, SCI >1 year prior to study enrollment, complete infralesional motor loss i.e., American Spinal Injury Association (ASIA) Impairment Scale Grade A or B, age <80 years, and no additional neurological diagnoses. Control subjects were age-matched to subjects with SCI and neurologically healthy. Subjects could have no major psychiatric diagnoses, nor contraindications to 123 MRI or TMS; the latter include certain metals in the body, pregnancy, and medications that lower seizure threshold such as tricyclic antidepressants or neuroleptics. The study was approved by the Institutional Review Board at the University of California, Irvine.
Study participation was 9 days long. On day 1, subjects underwent behavioral, fMRI, and TMS evaluations; days 2-8, 7 days of home motor imagery training for right foot and for tongue; and day 9, repeat of all baseline evaluations. Subjects with SCI and one set of healthy controls underwent complete testing and training. A second control group underwent serial fMRI scans only.

Behavioral assessments
A medical history was obtained from each subject, as well as determination of handedness (OldWeld 1971) and footedness (Coren 1993). ASIA scores and Impairment Grade were measured. Maximum right foot (a) dorsiXexion force and (b) tapping rate across 10° were measured.
The primary behavioral outcome measure to assess eVects of motor imagery training was the maximum rate of physical tapping of a 5-part sequence, determined separately for tongue and for right foot. Note that the 5part sequence used to test behavior was also the one that was mentally rehearsed as content of imagery training. Tapping rate was measured using separate foot and tongue sensor plates. Each plate had four numbered, spatially separate contact stations that together subtended 45°. For the tongue, the head was restrained and the plate was suspended in front of the mouth so that the tongue could reach each of the four stations without head or jaw movement. For the right foot, a sock was donned that limited sensor plate contact to the plantar hallux. A sensor plate on the Xoor allowed the toe to reach each station by rotating the foot in the plane of the Xoor with the heel held stationary. Each sensor plate station detected tongue, or right hallux, contact, which was ampliWed (2000£, Grass-Telefactor), digitally converted (Maclab, ADInstruments), and recorded. Subjects repeated 5-part sequence tapping onto the sensor plate at maximum rate over 30 s. Two 5part sequences were evaluated, in random order, on tongue then on right foot: "14213", the sequence also used as motor imagery training content; and "41342", a control sequence that was evaluated but not practiced.
Functional MRI scanning Subjects next prepared for fMRI by watching two videos, outside the scanner. Video 1 helped subjects understand the task to be performed. A right foot was shown hovering above an object, then moving in plan-tarXexion to crush the object.
Video 2 alternated 30-s blocks of rest and active state. For rest, subjects viewed images of a foot at rest, (new image every 3 s) and were instructed to remain at rest. For active state, subjects viewed images of a foot hovering above an object (new object every 3 s) and were instructed to attempt plantarXexion to crush each new object, regardless of whether their right foot was able to move or not. Bilateral ankle splints ) that went from lower tibia to toes were also placed, restricting movement to 10° of ankle dorsiXexion/plantarXexion and preventing lateral leg rotation. During this rehearsal, surface EMG leads were attached to bilateral tibialis anterior muscles, with signal Wltered (band pass 30/1,000 Hz), ampliWed, digitalized, and recorded.
During MRI scanning, subjects wore protective headphones, video goggles, the ankle splints, and had arms at side plus head stabilized. Scanning (1.5 T, Philips) began with a whole brain high-resolution volumetric anatomical scan (in-plane resolution 0.94 mm 2 , slice thickness 1 mm). Functional scans were acquired next, with a gradient-echo echoplanar imaging sequence, axial slices = 4 mm thick with 1 mm gap, TE = 40 ms, TR = 3 s, Xip angle = 80°, FOV = 24 cm, and acquisition matrix 128 £ 128. A total of 170 volumes were acquired while video 2 guided subjects through alternating rest with attempted right foot movement. During fMRI, a member of the study team observed the subject, noting movement in any of the four extremities.

Transcranial magnetic stimulation
Each subject's volumetric MRI scan was registered to the subject's head using Brainsight (Rogue Research, Inc.). Surface EMG leads were placed over the right tibialis anterior. A 70-mm Wgure of eight coil was placed over left medial central sulcus, with coil handle parallel to the sagittal sulcus. Using frameless stereotaxic TMS, the coil was systematically moved across a 5 £ 8 cm 2 region that included mesial aspects of both hemispheres to identify the site with the lowest threshold for motor evoked response (>50 V in at least 4 of 7 trials). Stimulation was adjusted in 1% TMS machine output increments.

Motor imagery training
The day after baseline assessments, a member of the research team set up a computer in each subject's 123 home. Instructions for the motor imagery were read aloud, then the subject demonstrated correct participation of training, i.e., imagery without actual movement. The subject then practiced motor imagery training for two 60-min sessions per day (30 min for right foot and 30 min for tongue in the a.m.; same in p.m.) for 7 days, recorded by the computer's camera. During training, audiovisual computer output guided the subject to imagine movement through the "14213" sequence, with a replica of the sensor plates having been placed adjacent to the body part undergoing motor imagery. A research team member phoned the subject every 3 days to oVer assistance. Compliance was judged to be full or not for each session based on three factors: (1) subject interviews during phone calls, (2) length of each practice session, with only sessions of full duration considered compliant; and (3) subject attention, assessed by whether subjects responded correctly to computer-generated questions regarding the number of sequence repetitions performed during a pre-announced test period each day.
Exam, MRI, and TMS were repeated the day after completion of motor imagery training.

Data analysis
Parametric statistics were used. For EMG data, the root mean square values were determined for the Wrst 20 s of the Wrst two rest and active blocks. Values for rest cycles were averaged, as were values for active cycles. For each muscle during each task, the ratio of active/rest EMG signal was then determined. For behavioral data, eVects of time were examined using repeated measures ANOVA. Groups were directly compared using the Student t-test.
From each fMRI scan, several measures of left hemisphere motor system activation were extracted, including volume, location, and magnitude of activation. To derive these three measures, an activation map for individual subjects was created, for each of the two visits, using SPM2. The Wrst two functional volumes were removed because of tissue non-saturation. Next, remaining images were realigned, coregistered to the volumetric scan, spatially normalized, transformed into MNI stereotaxic space, and spatially smoothed (4 mm FWHM). Images during attempted right foot movement were contrasted with rest for the activation map of each individual subject. Images showing excess head motion, deWned as zero areas of brain activation, activation along more than 90° at the edge of the brain, or massive activation across all brain areas, were excluded.
Activation volume was measured (P < 0.001, using small volume correction) within left posterior putamen, thalamus, globus pallidus, and foot primary sensorimotor cortex (Fig. 1). The posterior aspect of putamen was examined because repetitive movements preferentially activate this portion (Jueptner and Weiller 1998). Posterior putamen, thalamus, and globus pallidus were drawn by hand on the SPM single-subject anatomical template, on which deep gray matter is well contrasted. Primary sensorimotor cortex for foot extended from precentral to postcentral sulcus, vertex to cingulate sulcus. Activation location was measured within each of these four areas as MNI x, y, and z coordinates for center of activation, i.e., the center of mass. Activation magnitude was measured using MarsBaR (Brett et al. 2002) to extract task-related fMRI signal change in each subject's map within left posterior putamen, foot primary motor cortex, and foot primary sensory cortex. For activation magnitude, foot primary motor and sensory cortex were separated by the central sulcus, and included only the subset of voxels known to activate during foot movement in a prior control group ; note that none of the control subjects or subjects with SCI in the current study were part of that prior study. For each measure, a within group comparison evaluated the eVect of time, for each of the three subject groups individually, as well as for the combined group representing all subjects treated with motor imagery training. Next, the two treated subject groups, and the two healthy control groups, were directly compared. No correction was made for multiple comparisons.

Subjects
Of the ten subjects with SCI enrolled, eight were ASIA A (no infralesional sensorimotor function) and two were ASIA B (infralesional sensory but not motor function; Table 1). SCI level was C5 in Wve, C6 in one, T4 in one, T6 in two, and T10 in one. Subjects were studied 66 § 18 months after SCI (mean § SEM). There were no signiWcant diVerences between the two treated groups in age or footedness (Table 1).
Three treated control subjects declined to complete the protocol resulting in recruitment of three new enrollees, and Wve fMRI studies were contaminated by excess head motion. This left 12 baseline and 9 post-training scans for treated healthy controls; 9 and 8 scans, respectively, for subjects with SCI; and 10 123 and 10 scans, respectively, for the untreated healthy controls.

Behavioral assessments
The two groups showed no signiWcant diVerence in compliance with training. Full compliance during foot imagery training was present in 91 § 4% of sessions by subjects with SCI versus 97 § 2% of sessions by treated controls (mean § SEM, P > 0.2). Full compliance was present for tongue imagery training in 90 § 4% of sessions by subjects with SCI versus 98 § 1% of treated controls (P > 0.1). Though not formally measured, a number of subjects with SCI reported fatigue in the plegic right foot immediately after training.
Over time, both treated groups showed improvement on the two tongue motor tasks (for patients, P < 0.0005 for the practiced task and P < 0.05 for the unpracticed task; for controls, P < 0.0001 and P < 0.001, respectively; Table 2). For the practiced tongue task, controls showed signiWcantly larger gains over time than subjects with SCI (P < 0.05 for time £ group interaction). For the unpracticed tongue task, there was no diVerence between groups in gains over time. Directly comparing the two treated groups found no signiWcant diVerences on either task at either time point.
Study of the foot motor tasks in controls found a sig-niWcant gain over time in the practiced (P < 0.01), but not the unpracticed, task. Subjects with SCI had no foot movement at either time point and thus no gains.
Training produced greater gains for the practiced task than for the unpracticed task. This was true both for the foot task (P < 0.009, controls only) and for the tongue task (examining both groups, P < 0.05, 2-tailed t-tests, with repeated measures time £ group interaction not signiWcant).
Treatment-related gains on the foot motor task in controls did not generalize to other foot motor measures. Neither maximum force (154 § 22 N at baseline) nor tapping rate (3.7 § 0.3 taps/s) changed over time in controls, with nil values at both time points in subjects with SCI.

EMG
For both healthy control groups, at both time points, right tibialis anterior EMG during fMRI task rehearsal was signiWcantly (P < 0.05) increased during movement as compared to rest, but left tibialis anterior EMG did not increase with movement. In subjects with SCI, EMG did not increase with attempted movement in either muscle at either time point. Fig. 1 Representative slices from the regions of interest used in data analysis, all on left side of brain: a thalamus, b globus pallidus, c posterior putamen, d foot primary sensorimotor cortex, e the subset of foot primary motor cortex known to activate (at threshold Z > 3) during right foot movement from a prior study , f the subset of foot primary sensory cortex known to activate (at threshold Z > 3) during right foot movement from a prior study   During fMRI, all controls performed isolated foot movements as instructed. In subjects with SCI, neither the right foot nor other extremities showed visible movement during fMRI with the exception of occasional minor right hand twitches in one subject with SCI pre-training and in two subjects post-training.

Activation site
Over time, neither the MNI x, y, nor z location coordinate changed within left primary sensorimotor cortex, globus pallidus, thalamus, or posterior putamen. This was true in all three subject groups individually, as well as for the combined group representing all subjects treated with motor imagery training. Directly comparing the two treated groups at each time point disclosed one Wnding, that the x-coordinate for primary sensorimotor cortex activation at the second fMRI scan was more medial in subjects with SCI as versus treated controls (¡1 vs. ¡5, P < 0.03); note that prior to therapy, the same coordinates were present but these did not reach signiWcance. Directly comparing the two healthy control groups showed no signiWcant diVerences.

Task-related signal change
Over time, in the left posterior putamen, among all subjects treated with motor imagery training, taskrelated signal increased an average of 138 § 67% (P < 0.056), without a signiWcant time £ group interaction. The non-treated control group did not show this signiWcant change in left posterior putamen signal over time. Also, training was associated with no signiWcant eVect on task-related signal change within left M1 or S1 (Table 3). Over time, there were no other changes in any group or brain region. Directly comparing the two treated subject groups, as well as the two healthy control groups, at each time point disclosed no signiWcant diVerences in any brain region.

Activation volume
Over time, regional activation volume did not change signiWcantly within left primary sensorimotor cortex, globus pallidus, thalamus, or posterior putamen (Table 3). This was true in all three subject groups individually, as well as for the combined group representing all subjects treated with motor imagery training. Subjects with SCI showed larger activation than the treated controls after training in left globus pallidus (P < 0.054) and posterior putamen (P < 0.05, see Fig. 2); note that in both regions, a trend in the same direction was present before training. There were no other signiWcant diVerences in any brain region at either time point when the two treated subject groups, and when the two healthy control groups, were directly compared.

Transcranial magnetic stimulation
In patients, no TMS response in right tibialis anterior could be elicited, either before or after motor imagery training. In controls, TMS response could be elicited from seven of the ten subjects, while in three subjects no response could be elicited at either TMS session, even at 100% of device output, at any scalp position. Among these seven, at baseline, mean TMS threshold was 78 § 6% of device output, latency was 30.8 § 1.6 ms, and motor evoked potentials averaged 97 § 31 Vs. None of these TMS measures changed signiWcantly over time.

Discussion
Potential therapeutic interventions to reduce disability after SCI often assume intact brain motor system function 123 (Raineteau et al. 2002;Carmena et al. 2003;Dobkin and Havton 2004;Friehs et al. 2004;Scherberger et al. 2005). However, abnormalities of brain motor system function have been described after chronic SCI (Topka et al. 1991;Sabbah et al. 2002;Alkadhi et al. 2005;Cramer et al. 2005), and these might require attention to achieve maximum therapeutic eYcacy from new interventions. The current study hypothesized that motor imagery training would have similar eVects in controls and subjects with plegic SCI on activation of the brain motor system related to attempted movement. In support of this, controls, who could move the right foot during fMRI, and subjects with SCI, who could not, showed increased left putamen activation after a week of foot motor imagery training. These eVects of imagery training upon brain function have not been previously described in a neurologically impaired population. Together, these results suggest that brain motor system function can be modulated independently of voluntary motor control and peripheral feedback, and that motor imagery training might have value as an adjunct to restorative interventions targeting SCI. Motor imagery training was used in the current study rather than physical practice. As compared to attempted physical movement, motor imagery is associated with reduced magnitude of brain activation (Porro et al. 1996), and as compared to physical exercise, motor imagery training produces smaller behavioral gains (Lacourse et al. 2004b). However, imagery oVers certain advantages. From a study design point of view, this permitted closer matching of the intervention between subject groups. Also, motor imagery training might have clinical value in subjects with a weak limb, complementing the gains oVered by physical exercise (Jeannerod and Frak 1999;Pomeroy et al. 2005;Sharma et al. 2006). For example, motor imagery allows a subject to mentally experience completion of a task in a paretic/plegic limb, subjects can sculpt a single movement step in a limb impaired by synkinetic movements, and imagery-based training can be administered independent of motor impairment level. Reports of fatigue after training by some subjects with SCI suggest that improvement in mental Wtness is also a possibility. The data support the hypotheses related to training eVects on behavior, showing similar gains in tongue movement in both treated groups (Table 2), and showing gains in foot movement only in the healthy controls. Comparison of therapeutic behavioral gains in the current control group with published reports is limited by several factors. In treated controls of the current study, movement rate of the practiced task increased 36-75%. Three prior studies also employed the behavioral outcome measure as the content of 1week motor imagery training in healthy controls. Lacourse et al (2004a) found an 86% increase in movement rate after training, but these authors' higher value might in part be due to study of a diVerent body region, the right hand. Jackson et al (2003) studied the foot, but the endpoint was reaction time, which improved (decreased) by 10% after training. Pascual-Leone et al. (1995) found a 43% improvement (decrease) in right hand errors during performance of a Wne motor skill.
The data also supported the hypothesis that training would be associated with similar changes in brain activation across controls and subjects with SCI. The training-related increase in left putamen activation was similar across the two treated groups, with no signiWcant diVerence between the these two groups (Table 3), despite the fact that controls but not subjects with SCI could actually move the foot during scanning as instructed. The increase in putamen activation might reXect motor learning (Jueptner and Weiller 1998;Doyon et al. 2003), an interpretation supported by prior reports that motor imagery in healthy subjects improves motor performances (Pascual-Leone et al. 1995;Jackson et al. 2003;Lacourse et al. 2004a). The current data do not distinguish whether the motor system changes are related to planning, processing, or execution. However, the Wndings do suggest that at least some components of motor learning can take place in the complete absence of voluntary motor control.
Comparison of current results with prior serial brain mapping studies during motor imagery training in healthy subjects is complicated by diVerences in the brain mapping probe, treatment content, and study power. Increased orbitofrontal cortex (Jackson et al. 2003), as well as cerebellar, premotor, and striatal (Lacourse et al. 2004a), activation have been reported. Hypothesized changes in neurophysiology were not observed in the current study but were found in a prior serial brain mapping study by Pascual-Leone et al (1995). These authors gave healthy subjects similar imagery training as in the current study, but to the hand, and found reduced TMS thresholds over time.
The basis for this diVerence is unclear but might relate to the higher baseline thresholds required to stimulate the foot area, which resides within the interhemispheric Wssure, as compared to hand.
The study did not support the hypothesis that motor imagery training would reduce SCI-related baseline abnormalities in globus pallidus activation. A prior study using the same fMRI probe found that subjects with chronic, complete SCI had larger than normal activation volume in left globus pallidus ). In the current, smaller cohort, a baseline trend towards the same Wnding (Table 3) increased in signiWcance after training. The meaning of this Wnding requires further study as well as replication using a diVerent motor task, but globus pallidus activation has been associated with force regulation (Vaillancourt et al. 2004), speed of cognitive processing (Dirnberger et al. 2005), and with sensory gating of motor control (Kaji 2001). The current observations could represent derangement in one of these functions, or a dysfunction related to other movement disorders (Toda et al. 2004). Globus pallidus dysfunction after SCI requires further study and might reXect a process that is a barrier to restoring motor function.
Prior studies have also described abnormalities in leg primary sensorimotor cortex function using group map analyses, some Wnding subnormal (Humphrey et al. 2000;Sabbah et al. 2002;Cramer et al. 2005) and others Wnding supranormal (Alkadhi et al. 2005) activation. The current study was not powered to use these analyses, using a more sensitive region of interest approach instead, and did not identify diVerences between groups at baseline in primary sensorimotor cortex activation (Table 3). The current study did con-Wrm the results of a prior study  showing that left primary leg sensorimotor cortex activation during attempted right foot movement is more medial after chronic SCI as compared to activation during the same tasks in healthy subjects, a Wnding that might reXect a number of processes such as reduced sensory feedback or eVects of motor disuse.
Whether future therapies to improve motor status after SCI focus on spinal cord repair or brain signal recordings, attending to abnormalities of brain function will likely be an important component to maximize gains. The current study found that the training-related increase in putamen activation present in controls was also present in subjects with SCI, but that training did not reduce SCI-associated baseline abnormalities of globus pallidus function. Strengths of the study include homogeneity of subject groups, high compliance with training, and inclusion of a serially scanned non-treated control group. Weaknesses include small 123 sample size, and lack of correction for multiple comparisons. The study would have been improved by addition of a fourth subject group, subjects with SCI who did not undergo motor imagery training and by addition of a Wfth subject group, subjects with SCI who received sham training. The fMRI activation task was selected because of its demonstrated value  for probing the brain motor system of interest, that related to attempted movement. However, this probe might have had reduced sensitivity to treatment-related changes in brain function because treatment and brain activation employed diVerent tasks. Also, use of attempted movement introduced imbalances between subjects with SCI and controls related to lack of proprioceptive feedback in subjects with SCI, and the presence of a diVerent cognitive eVort underlying attempting movement when such movement does or does not occur. Movement in the four extremities was monitored visually during fMRI scanning, but subvisible muscle contractions might have occurred that could inXuence interpretation of motor system activation. In addition, given that numerous variables inXuence motor outcome and brain function after SCI (Humphrey et al. 2000;Cramer et al. 2005), results could vary in other cohorts, e.g., in subjects with more recent SCI, those with incomplete motor deWcits, or those with older age.
However, the results suggest that because motor imagery training is able to modify some brain functions related to attempted movement, despite the absence of overt movements, motor imagery training might have value as a potential adjunct to restorative therapies in subjects with SCI.