Brain activation of lower extremity movement in chronically impaired stroke survivors

www.elsevier.com/locate/ynimg

NeuroImage 26 (2005) 184–194

Brain activation of lower extremity movement in chronically impaired

stroke survivors

Andreas R. Luft,a,b,d,T Larry Forrester,c Richard F. Macko,d,e Sandy McCombe-Waller,c

Jill Whitall,c Federico Villagra,c and Daniel F. Hanleya

aDivision of Brain Injury Outcomes, Department of Neurology, Johns Hopkins University, Baltimore, MD 21218, USAbDepartment of General Neurology, Hertie Inst. For Clinical Brain Research, University of Tubingen, Tubingen, GermanycDepartment of Physical Therapy and Rehabilitation Science, University of Maryland, Baltimore, MD 21201, USAdDepartment of Neurology, University of Maryland, Baltimore, MD 21201, USAeDepartment of Veterans Affairs and the Baltimore VA Geriatric Research Education and Clinical Center (GRECC) and Division of Gerontology,

University of Maryland School of Medicine, Baltimore, MD 21201, USA

Received 25 August 2004; revised 9 January 2005; accepted 12 January 2005

Available online 16 March 2005

Lower extremity paresis poses significant disability to chronic stroke

survivors. Unlike for the upper extremity, cortical adaptations in

networks controlling the paretic leg have not been characterized after

stroke. Here, the hypotheses are that brain activation associated with

unilateral knee movement in chronic stroke survivors is abnormal,

depends on lesion location, and is related to walking ability. Functional

magnetic resonance imaging of unilateral knee movement was obtained

in 31 patients 26.9 months (mean, IQ range: 11.3–68.1) after stroke and

in 10 age-matched healthy controls. Strokes were stratified according

to lesion location. Locomotor disability (30 ft walking speed) did not

differ between patient groups (9 cortical, 12 subcortical, 10 brainstem

lesions). Significant differences in brain activation as measured by

voxel counts in 10 regions of interest were found between controls and

patients with brainstem ( P = 0.006) and cortical strokes ( P = 0.002),

and between subcortical and cortical patients ( P = 0.026). Statistical

parametric mapping of data per group revealed similar activation

patterns in subcortical patients and controls with recruitment of

contralateral primary motor cortex (M1), supplementary motor area

(SMA), and bilateral somatosensory area 2 (S2). Cortical recruitment

was reduced in brainstem and cortical stroke. Better walking was

associated with lesser contralateral sensorimotor cortex activation in

brainstem, but stronger recruitment of ipsilateral sensorimotor and

bilateral somatosensory cortices in subcortical and cortical patients,

respectively. A post hoc comparison of brainstem patients with and

without mirror movements (50%) revealed lesser recruitment of

ipsilateral cerebellum in the latter. Subcortical patients with mirror

movements (58%) showed lesser bilateral sensorimotor cortex activa-

tion. No cortical patient had mirror movements. The data reveal

adaptations in networks controlling unilateral paretic knee movement

1053-8119/$ - see front matter D 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.neuroimage.2005.01.027

* Corresponding author. Abteilung Allg. Neurologie Hertie Institute fqrKlinische Hirnforschung Universit7t Tqbingen Hoppe-Seyler-Str. 3 72076

Tqbingen, Germany. Fax: +49 7071 967857.

E-mail address: [email protected] (A.R. Luft).

Available online on ScienceDirect (www.sciencedirect.com).

in chronic stroke survivors. These adaptations depend on lesion

location and seem to have functional relevance for locomotion.

D 2005 Elsevier Inc. All rights reserved.

Keywords: Stroke; Hemiparesis; Functional imaging; Lower extremity;

Locomotion; Mirror movement

Introduction

Locomotor impairment reduces the quality of life of chronic

stroke survivors. More than upper-extremity dysfunction, it

increases the burdens and costs of post-stroke care (Mayo et al.,

1999). Rehabilitation techniques to improve locomotion were

developed (Macko et al., 1997; Smith et al., 1999; review in

Moseley et al., 2003) and are currently tested in randomized clinical

trials. Successful development and optimization of such therapies

requires an understanding of adaptive and compensatory mecha-

nisms which the brain utilizes to improve lower extremity paresis.

Such mechanisms are well known to exist for the upper

extremity (Chollet et al., 1991; Cramer et al., 1997; Luft et al.,

2004; Marshall et al., 2000; Seitz et al., 1998; Weiller et al., 1992,

1993). These studies suggest that intact brain regions in the

lesioned or the non-lesioned hemisphere–in particular premotor

cortex, postcentral gyrus, and motor cortices in the non-lesioned

hemisphere–are recruited during paretic movement. It is assumed

that this activation reflects a–more or less successful–compensation

for the lesion (Calautti and Baron, 2003; Hallett, 2001). The

characteristics of the activation pattern depend on the level of

recovery (Ward et al., 2003), on lesion location (Luft et al., 2004),

and on the duration of recovery after the index stroke (Feydy et al.,

2002). Here, we study chronically disabled patients long after their

stroke to eliminate the effects of duration.

A.R. Luft et al. / NeuroImage 26 (2005) 184–194 185

For the lower extremity, comparable data are not available.

Considering the differences in use patterns of upper (mainly

unilateral, e.g., reaching) and lower extremities (mainly bilateral,

e.g., walking) and the differences in brain activation associated

with healthy arm and leg movement (Luft et al., 2002), a transfer of

findings from arm to leg seems unjustified. Few studies have

investigated brain activation during lower extremity movements.

Miyai et al. (2001, 2002, 2003) used near infrared spectroscopy to

map brain activation during walking in healthy subjects and stroke

patients before and after a rehabilitation program. Functional

magnetic resonance imaging (fMRI) of the lower extremity,

however, poses technical difficulties due to head motion artifacts.

We developed and tested a paradigm that allows for functional

magnetic resonance imaging of unilateral knee movement in

patients with limited range of motion (Luft et al., 2002).

The objective of this study was to identify and characterize

adaptation in neural networks controlling lower extremity move-

ment in chronically impaired stroke survivors. Specifically, brain

activation patterns of paretic and non-paretic knee movement are

compared to healthy controls, the relationship between activation

pattern and lesion location is characterized, and correlations

between activation and walking ability are investigated.

Fig. 1. For each group of patients, the average lesion distribution is

superimposed onto an averaged anatomical image. Shades of red to yellow

indicate in how many of 9 cortical, 10 brainstem, and 12 subcortical stroke

patients, a particular area was lesioned (red—injury less frequent, yellow—

more frequent).

Materials and methods

Subjects

Thirty-one patients (age: 65.6 F 8.6 years, mean F SD, 13

female, 18 male), who suffered a first-ever stroke 26.9 months

(median, IQ range: 11.3–68.1 months) before inclusion, are

reported here. All patients were hemiparetic with residual leg

movement from a single ischemic stroke. They were selected

based on the presence of a single remote stroke longer than 6

months before study entry, the presence of residual hemiparesis,

completion of all conventional rehabilitation programs, and

adequate language and neurocognitive function to understand

informed consent and study-related instructions. Patients with

more than one symptomatic stroke, other neurological disease,

chronic pain, emotional disorders, and/or metal implants were

excluded. Four additional patients, initially screened and found

eligible, had to be excluded because of obesity (subject did not

fit the restraining device, n = 2), or post hoc because video

footage of the head during scanning revealed visible head

movement (n = 2).

Locomotor impairment was assessed at the time of study

enrollment by measuring the time to walk 30 ft. Six 30-ft timed

walks at bself-selectedQ and bfastest comfortableQ paces were

conducted with patients using the same assistive device and/or

orthoses as when walking at home. Twenty-seven of 31 patients

(87%) made used of assistive devices (single point cane: 18,

quad-point cane: 7, walker: 2). Sixteen patients (52%) had ankle

foot orthoses. Because medical records for patients with remote

strokes were not available, the degree of initial impairment

(acutely after stroke) was assessed by asking the patient to

categorize the initial disability as either bno movementQ, bminimal

movement/not able to lift leg off bedQ, bable to lift leg off bed, but

not able to stand of walkQ, bable to stand/walk with weaknessQ, orbnormalQ (score 0 to 4).

To quantitatively assess the degree of involuntary co-contraction

of the intact limb during voluntary movement of the paretic limb

(mirror movement), bilateral EMG recordings of the vastus

medialis and lateralis muscles obtained during voluntary isometric

flexion of the paretic knee were analyzed. These data were collected

outside the MRI scanner within 2 weeks of fMRI testing as part of

another study. A threshold for co-contraction was defined as a

signal-amplitude of mean + 3 SD above relaxation levels (rectified

EMG). The number of times this threshold was surpassed in the

intact limb during voluntary contraction of the paretic limb in ten

sampling epochs (5 s duration, 2 s interepoch pause) was quantified

as the percentage of the 10. The criterion for claiming mirror limb

co-contraction was relevant activation in both contralateral muscles

in N50% of epochs. These data were compared to video recordings

of each patient during fMRI scanning.

The patient sample was stratified according to whether the

lesion involved the sensorimotor cortex with or without subcortical

regions (further called cortical group), subcortical regions only

(subcortical group, dsubcorticalT regions were all areas medial to

the insular cortex and ventral to the corpus callosum, i.e., basal

ganglia and descending pathways), or the brainstem (medulla, pons

or midbrain, with the red nucleus as cranial limitation). Fig. 1

presents the average stroke locations per group. Lesion volume was

measured by manual delineation of the hypointensity on high-

resolution T1-weighted images that were obtained as an anatomical

reference for functional scans.

Patients were compared to a group of ten healthy volunteers

(6 male, 4 female, age 58 F 7.3 years, mean F SD). A subset of

these data was previously published (Luft et al., 2002). Written

informed consent was obtained from all subjects in accordance

with the Declaration of Helsinki. The study protocol and the

A.R. Luft et al. / NeuroImage 26 (2005) 184–194186

consent forms were approved by the participating institutions

(Johns Hopkins University and University of Maryland).

fMRI technique and paradigms

fMRI technique and paradigms were described before (Luft et

al., 2002). In brief, a custom-made Plexiglas scaffold was used to

ensure constant movement range and plane of motion and to limit

concomitant head motion by fixation of pelvis, thigh, and trunk.

fMRI scanning was performed using a 1.5-T scanner (Gyroscan,

Philips, Eindhoven, The Netherlands) at the F.M. Kirby Center for

Functional Brain Imaging, Kennedy Krieger Institute, Baltimore. A

T1-weighted 3D-MPRAGE sequence (TR 8.2 ms, TE 3.7 ms, flip

angle 88, NEX 1, matrix 256 � 256, voxel-resolution: 1 � 1 � 1

mm3) was obtained as anatomical reference. BOLD-weighted EPI

sequences (TR 3 s, TE 40 ms, matrix: 64 � 64, slice thickness 5

mm, pixel resolution 3.75 � 3.75, 30–36 slices to obtain whole

brain coverage, interleaved slice timing) were scanned in axial

orientation while the subject moved either the paretic or the non-

paretic knee (right vs. left in controls). For each functional

measurement, sixty scans were acquired over a period of 3 min

without interscan delay; during this time, three cycles of rest and

knee movements were completed. Knee movements were acous-

tically cued by computer-generated beeps (one every 3 s, beep

duration 0.1 s) triggering 10 movements per cycle. The low pace

allowed all patients to perform the movement without fatiguing (as

tested before scanning).

The movement consisted of unilateral knee extension starting at

308 of flexion and ending at 10 to 158 of flexion. The non-pareticand the paretic limb movement were scanned in sequence. The

movement range was adjusted before scanning to match the

patient’s ability with the paretic limb (measured at the toe: 18 F3.2 cm mean F SD, range: 11.5–23.5 cm), again, with the goal of

avoiding fatigue, and to homogenize the effort required for

movement among patients. The same range of motion was used

for the non-paretic limb. Knee movements were videotaped to

monitor for mirror movements. A second camera positioned above

the subject’s face (built into the scanner) recorded head motion. If

head motion was visible in these recordings, the subject was

excluded from further analysis (this was the case for 2 who are not

included in the 31 patients reported here). Subjects kept their eyes

closed during scanning.

Statistical analysis of fMRI data

BrainVoyager (Brain Innovation B.V., Maastricht, The Nether-

lands) was used to analyze functional imaging data. Image sets

were first corrected for head motion and slice timing and were

temporally smoothed (high pass filter, linear trend removal). Mean

head motion as determined by the correction algorithm was 2.5 F0.1 mm (mean F SEM). In all subjects, maximum movement was

smaller than 3 mm. The efficacy of motion correction was assessed

by viewing the corrected functional image series of every subject

as a movie; residual motion was not observed. Image series were

warped to Talairach space, first, by applying rigid body trans-

formations (translation, rotation), followed by manual definition of

Talairach landmarks, and third, by scaling of 12 landmark-defined

subvolumes into Talairach space.

The predicted time course of voxel intensities was that of a

box-car function with hemodynamic response modification. To

identify activated voxels, linear regression models including data

of single or of all subjects were computed. For single-subject

region of interest (ROI) analysis, the data were spatially

smoothed (4 mm FWHM) and probability thresholds of 0.0001

(uncorrected for multiple comparisons) were employed. Activated

voxels were counted in ten ROIs using software we developed

based on Matlab (Mathworks Inc, Natick, MA, USA). The ten

ROIs were defined in each patient by manually identifying the

following anatomical landmarks in T1 images: The precentral

gyrus was delineated medially from the fundus of the central

sulcus (in the axial plane) to the lateral surface. To include the leg

area, it extended into the interhemispheric fissure. While the

posterior border of the leg area was usually identified by the

medial extension of the central sulcus, the anterior border

separating it from SMA proper, was defined arbitrarily according

to Brodmann’s map. The inferior border of the leg area was the

cingulate sulcus; the inferior boundary of the precentral ROI

laterally was defined as the first axial slice (browsing in dorso-

ventral direction) not showing the hand notch configuration

(inverted omega). SMA proper was anterior to the leg area,

extending to a vertical plane through the anterior commissure

(Zilles et al., 1995). The postcentral ROI was defined as the

postcentral gyrus excluding the area corresponding to the

posterior parietal cortex (Grefkes et al., 2001). The cerebellum

was divided into three ROIs, vermis (medial to the cerebellar

tonsils), left, and right cerebellar hemisphere. The supramarginal

ROI was defined according to Brodmann’s area 40 posterior to

the postcentral gyrus, superior to the Sylvian fissure, anterior of

the angular gyrus, and inferior of the intraparietal fissure.

Superior, middle, and inferior frontal gyri ROIs included the

respective gyri with the following boundaries: medial—medial

hemispheric margin, posterior—precentral ROI, inferior—Sylvian

fissure with horizontal (axial) anterior extension, anterior—

coronal plane through the anterior margin of the genu of the corpus

callosum. To avoid ROI-overlap, previously defined ROIs were

subtracted from original image sets before tracing subsequent ROIs.

Based on the number of activated voxels per ROI, a lateralization

index (LI) was computed as (countcontralateral � countipsilateral)/

(countcontralateral + countipsilateral).

For multi-subject regression models, the data were smoothed

using a kernel of 8 mm FWHM. Data of patients with left

hemisphere stroke were flipped about the mid-sagittal plane.

Hence, the paretic leg was always considered to be the left and

vice versa for the non-paretic. In controls, right- and left-sided

movement data were averaged after appropriate flipping so that the

orientation corresponded with that of patients. Multi-subject

regression models were computed for paretic or non-paretic limbs

with and without inclusion of the mean of self-selected and

maximum walking speeds as covariate. The inclusion of the

covariate allowed for an assessment of correlations between

functional measures of gait and brain activation. Fixed effects

analysis was used with a uniform probability threshold of 0.05

(corrected for multiple comparisons across the entire brain). To

compare activation maps of patients with mirror movements to

those without, difference maps (orthogonal contrasts) were

computed; significance of the difference was then tested by

computing t tests for a set of voxels that were manually selected

in the difference statistical maps. All functional maps were

superimposed onto averaged T1-weighted image sets derived from

all subjects per group.

To localize foci of activation in multi-subject maps, the

Talairach Daemon (Lancaster et al., 1997) with a query range of


5 mm was used. In the following, the term bprecentral gyrusQ refersto the ROI, whereas bM1Q indicates that activation was mapped to

the posterior rim of the precentral gyrus, i.e., to Brodmann’s area 4

according the Talairach Daemon.

Statistical analysis of numerical data

Baseline characteristics of the subject sample were compared

between groups of patients using ANOVA. Normal distribution and

homogeneity of variances was tested using Schapiro–Wilk and

Levene’s tests. Regional voxel counts were analyzed using a full

factorial ANOVA model including group (three patient groups and

healthy controls), side (ipsilesional versus contralesional), ROI (ten

ROIs), leg (paretic vs. non-paretic), and mirror (presence/absence

of mirror movements) as independent variables. Tukey’s honestly

significant difference (HSD) tests were applied in post hoc analysis

to test for differences between groups. Hemispheric dominance

was assessed by testing whether the LI was different from zero

(one sample t test). Significance was assumed on the 5% level.

Data are reported as mean F SEM unless noted otherwise.

Fig. 2. Lateralization of activation is compared between groups for selected

regions of interest. Lateralization in the medial precentral gyrus ROI (leg

representation) is observed in subcortical and brainstem stroke and to a

lesser degree in healthy controls (statistical trend, P = 0.059). Cortical

stroke patients demonstrate no lateralization. The opposite pattern is seen

for the supramarginal gyrus. For non-paretic limb movement, precentral

gyrus activation is lateralized only in cortical stroke patients. (*P b 0.05).

Results

Patient sample characteristics

Ten patients were classified as brainstem strokes, nine as

cortical, and twelve as subcortical (Table 1). Age, gender,

retrospective ratings of the initial motor deficit, and walking

velocities were similar between groups. Cortical stroke patients had

longer time intervals between stroke and study entry than

subcortical or brainstem patients. Right-sided strokes were more

frequent in the cortical stroke group, likely reflecting a selection

bias due to the entry criterion of adequate language and neuro-

cognitive function (aphasia is common in cortical stroke patients

with left hemisphere lesions). Lesion size was expectedly different

between groups. High correlations (r N 0.95, P b 0.05) were found

between self-selected and maximum walking speeds. No correla-

tions were observed between lesion volume and walking speed

between and within groups.

Region of interest analysis (voxel counts)

Comparison patients—controls (ANOVA)

Voxel counts derived from region of interest (ROI) analysis of

single-subject activation maps were not normally distributed and

Table 1

Patient characteristics

Brainstem

Number 10

Age 69.8 (2.1)

Gender (female) 3

Side of stroke (right) 5

Time since stroke (months) 19.0 (10.3–46.1)

Lesion size (ml) 0.37 (0.07)

Initial deficita 1.37 (0.38)

Walking speed, self-selected (m/s) 0.48 (0.07)

Walking speed, maximum (m/s) 0.59 (0.08)

Values represent mean F SEM, except for btime since strokeQ: median (IQ rangea Rated retrospectively by the patient on a scale of 0–4 (0: no movement, 4: full

had unequal variances; therefore, the data were square root-

transformed for further analysis.

ANOVA revealed significant effects for group [F(3,1273) =

4.88, P = 0.002], ROI [F(9,1273) = 6.85, P b 0.0001], leg

[F(1,1273) = 7.14, P = 0.008], and mirror [F(1,1273) = 19.74, P b

0.0001], but not for side of ROI [F(1,1273) = 0.045, P = 0.83].

Significant interactions were group � leg [F(1,1273) = 2.78, P =

0.04] and group � mirror [F(1,1273) = 13.61, P b 0.0001]. Post

hoc Tukey’s HSD tests demonstrated differences between healthy

controls and patients with cortical (P = 0.002) and brainstem

strokes (P = 0.006). Also, activation patterns in subcortical and

cortical patients were significantly different (P = 0.026). A trend

was found for subcortical versus brainstem patients (P = 0.062).

Lateralization indices

For paretic movement, ROI lateralization indices demonstrated

significant contralateral dominance in the precentral ROI for

patients with subcortical [t(11) = 2.51; P = 0.031, Fig. 2] and

Cortical Subcortical P

9 12

64.3 (3.0) 63.2 (2.6) 0.17

4 6 0.65

9 7 0.04

78.7 (16.9–117.9) 24.3 (11.2–34.6) 0.02

55.9 (20.9) 2.15 (0.62) 0.001

1.42 (0.48) 0.17 (0.17) 0.07

0.54 (0.09) 0.47 (0.06) 0.81

0.71 (0.12) 0.67 (0.11) 0.77

).

movement).


brainstem strokes (t(9) = 2.43; P = 0.041). A trend for

contralateral dominance was found in healthy controls [t(9) =

2.16, P = 0.059]. No dominance was seen in cortical stroke

patients—neither in the entire group nor in a subgroup of four

patients who showed bilateral precentral gyrus activation. Instead,

significant lateralization was found for the supramarginal ROI

(BA 40, t(8) = 3.91, P = 0.004, Fig. 2). For non-paretic

movement, contralateral dominance in the precentral gyri was

observed in the cortical stroke group (t(8) = 3.55, P = 0.0094,

Fig. 2).

Correlation between voxel counts and lesion volume

Correlations between voxel counts and lesion volume were

found for the paretic limb in brainstem stroke: Ipsilateral

(contralesional) superior frontal gyrus activation (premotor

cortex) was positively related to lesion volume (r = 0.80,

P = 0.01); the lateralization index showed a negative

correlation (r = �0.76, P = 0.018). In subcortical stroke

patients, activation in the inferior frontal gyrus associated with

non-paretic movement was positively correlated with lesion

volume (r = 0.86, P = 0.001).

Fig. 3. Activation patterns of paretic (red–yellow), non-paretic (blue), and healthy

templates. Image data of patients with left-sided stroke are flipped about the mid-s

control subjects, activation patterns of left- and right-sided knee movement were

Whereas during paretic limb movement, subcortical stroke patients and, to a lesser

almost no cortical activation is observed in patients with cortical stroke. For n

activation is seen in all groups, but is also markedly different from control.

Activation maps associated with paretic movement

Multi-subject statistical mapping revealed more widespread

activation in subcortical than in cortical or brainstem patients

(Fig. 3, see Table 2 for Talairach coordinates). In subcortical

stroke, strong recruitment of the leg representation of the primary

motor cortex (M1) contralateral to the paretic limb and–to a

lesser extent–the corresponding region in the ipsilateral hemi-

sphere was observed. Recruitment was also found in bilateral

SMA, in the supramarginal gyrus corresponding to Brodmann’s

area (BA) 40 (secondary somatosensory area, S2), and in the

anterior cingulate gyrus (BA 24). Further activation was observed

in contralateral thalamus and in ipsilateral cerebellum (vermis and

anterior lobe).

Patients with brainstem stroke recruited the M1 leg representa-

tion (Fig. 3). Activation was also seen in contralateral SMA and

premotor cortex (BA 6), ipsilateral S2, and contralateral thalamus.

Cerebellar activation was found in the ipsilateral hemisphere

(declive).

Cortical stroke patients as a group revealed little overall brain

activation and no recruitment of M1. In 4 patients, bilateral motor

control knee movement (green) are superimposed onto averaged anatomical

agittal plane so that the lesioned hemisphere is always the right. For healthy

averaged (after appropriate flipping so that the moving limb is the left).

degree, brainstem patients recruit sensorimotor cortex and SMA bilaterally,

on-paretic limb movement, consistent contralateral primary motor cortex

Table 2

Talairach coordinates of activation foci per group and limb

Group Arm Region BA Side x y z

Brainstem Paretic Medial M1 4 Contra 4 �29 59

SMA 6 Contra 4 �18 59

Anterior cingulate 24 Contra 4 �14 59

Posterior cingulate 30 Contra 4 �47 15

Precuneus 7 Contra 13 �47 50

Precentral gyrus 6 Ipsi �48 �2 40

Parietal lobule 40 Ipsi �49 �46 24

Posterior cingulate 30 Ipsi �4 �47 15

Thalamus Contra 20 �20 12

Cerebellum (declive) Ipsi �30 �60 �19

Non-paretic Medial M1 4 Contra �3 �34 60

SMA 6 Contra �4 �24 50

Anterior cingulate 24 Ipsi 5 �13 39

Postcentral 3 Contra �31 �20 45

Parietal lobule 40 Ipsi 56 �29 27

Parietal lobule 40 Contra �44 �39 28

Posterior cingulate 30 Ipsi 4 �49 13

Thalamus Ipsi 14 �16 12

Cerebellum (declive) Contra �25 �75 �28

Subcortical Paretic Medial M1 4 Contra 9 �35 58

Medial M1 4 Ipsi �8 �37 57


SMA 6 Ipsi �10 �13 53

Anterior cingulate 24 Bilateral F10 �11 42

Lateral M1 4 Contra 33 �18 42

Precentral gyrus 6 Contra 43 �5 24


Parietal lobule 40 Contra 48 �34 27


Superior temporal gyrus 41 Contra 52 �23 8

Superior temporal gyrus 41 Bilateral F57 0 1

Thalamus Contra F13 �16 1

Caudatum Ipsi �27 �10 6

Anterior vermis 0 �45 �15

Cerebellum (ant. lobe) Ipsi �22 �45 �24



Anterior cingulate 24 Contra �8 �12 37


Precentral gyrus 6 Ipsi 44 �13 33

Parietal lobule 40 Contra �53 �35 34

Parietal lobule 40 Ipsi 50 �32 34

Superior temporal gyrus 22 Contra �51 �23 5

Superior temporal gyrus 22 Ipsi 50 �18 5

Thalamus Contra �16 �18 6

Anterior vermis 4 �45 �17

Cerebellum (ant. lobe) Contra �27 �53 �20

Cerebellum (ant. lobe) Ipsi 31 �56 �18

Dentate nucleus Contra �22 �58 �30

Dentate nucleus Ipsi 12 �56 �30

Cortical Paretic Anterior cingulate 31 Ipsi �2 �34 32


Posterior cingulate 30 Bilateral F4 �47 16

Thalamus Bilateral F17 �13 2

Putamen Ipsi �28 �15 2

Vermis 0 �79 �20

Cerebellum (ant. lobe) Contra 23 �52 �28






Posterior cingulate 30 Bilateral F4 �42 12

(continued on next page)


Fig. 4. Multi-subject activation maps of subcortical stroke patients stratified

according to the presence of mirror movements. Larger ipsi- and

contralateral motor cortex recruitment is found for patients without mirror

movements (orange). Patients with such movements (green) have stronger

recruitment of contralateral SMA.

Table 2 (continued)

Group Arm Region BA Side x y z

Volunteer Right/Left Medial M1 4 Contra 8 �39 59


Anterior cingulate 24 Contra 12 �10 40

Anterior cingulate 24 Ipsi �9 �9 41

Precentral gyrus 6 Contra 46 �3 33




Thalamus Contra 21 �19 1

Thalamus Ipsi �10 �17 �3

Cerebellum (ant. lobe) Ipsi �21 �29 �21

Abbreviations: Contra—contralateral to darmT, Ipsi—ipsilateral to darmT, BA—Brodmann’s area, x/y/z—Talairach coordinates.


cortex activation was seen in single-subject analysis (data not

shown). Apart from M1, activation was found in ipsilateral

thalamus, BA 40, and cingulate gyrus (Fig. 3).

Activation maps associated with non-paretic movement

Recruitment of M1 contralateral to the non-paretic limb was

observed in all groups of patients (Fig. 3). Further activation was

identified in SMA, cerebellum, cingulum, and in postcentral,

supramarginal, and angular gyri (Table 2).

Influence of mirror movements on the activation pattern

Presence or absence of mirror limb co-contraction in EMG

recordings outside the scanner correlated well with video footage

during fMRI scanning (Spearman correlation coefficient: 0.85, P b

0.0001). One patient with subcortical stroke had no visible

movement on video despite relevant co-contraction of the

unimpaired vastus medialis (100% of epochs) and vastus lateralis

muscles (60% of epochs); another patient with subcortical stroke

had some mirror movement in the foot, but EMG analysis did not

reveal co-contraction.

Among ten patients with brainstem stroke, five (50%; one

patient had visible movement during fMRI, but EMG was

unavailable) were classified as having mirror movements. These

were compared to the five patients without such movements: The

difference GLM revealed significantly stronger activation in the

ipsilateral anterior lobe of the cerebellum in patients with mirror

movements (Talairach coordinates:�26/�41/�28; t = 8.37, P b

0.0001). No differences were found in the cerebral hemispheres.

Walking speed was not different in brainstem patients with mirror

movements versus those without.

In twelve subcortical patients, mirror movements were identi-

fied in seven (58%). These were compared to five patients without

such movements (Fig. 4). Patients with mirror movements showed

significantly higher activation in contralateral SMA (Talairach

coordinates: 3/�10/68; t = 2.97, P b 0.003) and in the ipsilateral

posterior lobe of the cerebellum (�18/�81/�24; t = 4.75; P b

0.0001). Lower activation was found in the contralateral post-

central gyrus (45/�27/39; t = 3.78; P = 0.0002), and in bilateral

primary motor cortex (contralateral: 26/�25/56; t = 3.02; P =

0.003; ipsilateral:�16/�24/52; t = 4.36; P b 0.0001) and anterior

cerebellar lobes (contralateral: 40/�44/�25; t = 2.66; P = 0.008;

ipsilateral:�36/�39/�26; t = 4.340; P b 0.0001). Walking speed

was not different in subcortical patients with mirror movements

versus those without.

None of the patients with cortical stroke met the criteria for co-

contraction in EMG or showed mirror movements on video

footage. The frequency of mirror movement differed significantly

between patient groups as revealed by logistic regression analysis

[stroke location as a predictor for the presence of mirror

movements: v2(2) = 9.03, P = 0.011]. The power of this test was

0.89.

Brain activation related to walking ability

Activation foci determined by a linear model with mean

walking speed as covariate are listed for each group of patients

in Table 3. Interestingly, walking ability was not only related to

brain activation observed during paretic but also to activation

associated with non-paretic knee movement. For paretic move-

ment, less contralateral M1 activation was associated with better

walking in patients with brainstem strokes. In contrast, stronger

ipsilateral M1 and S1 activation was related to better walking in

subcortical stroke. In cortical stroke, the same was found for

bilateral S1. Covariate analysis after patient stratification according

to the presence of mirror movements (subcortical and brainstem

patients) did not reveal significant results, likely because sample

sizes were too small.

Discussion

These findings demonstrate that brain activation associated with

paretic as well as with non-paretic lower extremity movement in

Table 3

Talairach coordinates of activation foci correlated with walking ability per group and limb

Group Arm r Region BA Side x y z

Brainstem Paretic � M1 4 Contra 46 �8 47

� Precuneus 7 Bilateral F3 �49 63

� Insula 13 Ipsi �32 �21 6

� Middle temporal gyrus 21 Ipsi �63 �36 �7

� Cingulum 24 Contra 17 �12 43

� Posterior cingulum 30 Ipsi �4 �44 21

� S2 40 Contra 47 �35 43

� Cerebellum (post. lobe) Bilateral F11 �69 �24

Non-paretic � Paracentral lobule 5 Ipsi 25 �39 45

� Precentral gyrus 6 Ipsi 49 �13 35a

� Cingulum 31 Ipsi 9 �39 36

� Middle temporal gyrus 39 Ipsi 48 �59 8

� Cerebellum (post. lobe) Contra �33 �68 �26

� Cerebellum (nodulus) Ipsi 10 �53 �26

+ Middle frontal gyrus 8 Ipsi 33 21 47

+ Cingulum 32 Contra �13 12 38

Subcortical Paretic � S1 3 Contra 26 �28 67

� M1 4 Contra 21 �23 67

� Precentral gyrus 6 Contra 28 �13 67

� Superior fronal gyrus 6 Contra 30 39 33

� Superior parietal lobule 7 Bilateral F26 �58 57

� Middle frontal gyrus 8 Contra 41 17 43

+ S1 3 Ipsi �20 �30 62

+ M1 4 Ipsi �20 �25 62

+ Pallidum Bilateral F16 �9 �5

Non-paretic � M1 4 Ipsi 38 �19 62

� Middle frontal gyrus 6 Ipsi 30 1 51

� Precentral gyrus 6 Ipsi 39 �3 28

� Precuneus 7 Ipsi 4 �67 49

� Cuneus 18 Bilateral F15 �68 16

� S2 40 Ipsi 47 �45 51

� Cerebellum (post. lobe) Contra �30 �78 �30

+ Precuneus 7 Contra �20 �70 43

+ Medial frontal gyrus 8 Ipsi 15 24 46

+ Middle frontal gyrus 9 Contra �34 21 33

+ Cuenus 18 Ipsi 22 �85 23

+ Temporal lobe 21 Contra �46 �30 �3

+ Anterior cingulate 24 Contra �7 5 39

+ S2 40 Ipsi 55 �28 26

+ Inferior frontal gyrus 45 Ipsi 55 18 14

+ Cerebellum (ant. lobe) Ipsi 19 �34 �16

Cortical Paretic + S1 3 Contra 13 �39 72b

+ S1 3 Ipsi �21 �37 72

+ Superior parietal lobule 7 Ipsi �32 �47 62

+ Precuneus 7 Bilateral F5 �71 39

+ Putamen Contra 23 �1 11

+ Cerebellum (post. lobe) Ipsi �28 �80 �30

Non-paretic � Precentral gyrus 6 Ipsi 41 �7 53c

� S2 40 Ipsi 44 �32 52

+ Precentral gyrus 6 Contra �35 �3 41c

+ Precuneus 7 Contra �4 �57 39

+ Middle frontal gyrus 9 Contra �27 18 32

+ Insula 13 Contra �36 �37 17

+ Superior temporal gyrus 22 Contra �41 �57 17

+ S2 40 Contra �38 �34 41

+ Superior temporal gyrus 41 Ipsi 48 �23 8

+ Putamen Ipsi 21 �3 9

+ Cerebellum (post. lobe) Bilateral F16 �73 �26

Results of region of interest analysis (voxel counts) support ANCOVA map findings: anegative correlation between precentral ROI and mean walking speed

(r = �0.92, P = 0.001); bpositive correlation between lateralization index of postcentral ROI and mean walking speed (r = 0.75, P = 0.032); cpositive

correlation between lateralization index of precentral ROI and mean walking speed (r = 0.70, P = 0.035).

Abbreviations: r—correlation (positive/negative), Contra—contralateral to darmT, Ipsi—ipsilateral to darmT, BA—Brodmann’s area, x/y/z—Talairach

coordinates.



chronically hemiparetic stroke patients differs substantially from

activation in healthy controls. The differences depend on stroke

location: in patients with cortical lesions, recruitment of sensor-

imotor cortices occurs variably and, if present, is bilateral; stronger

S1 activation is associated with better walking ability. In brainstem

stroke, activation is reduced as compared to controls and lesser

activation (of contralateral M1) predicts better walking. Patients

with subcortical lesions have recruitment of sensorimotor cortices

that is very similar to healthy control. Higher ipsilateral sensor-

imotor cortex activation in these patients relates to better walking.

The presence of mirror movements in this group has an influence

on the cortical activation pattern. Other areas differentially

activated by unilateral knee movement in all patients groups and

normals were SMA, cingulum, cerebellar hemispheres, and S2

(Brodmann’s area 40).

That paretic knee movement is associated with brain activation

patterns that differ from healthy control seems expected given the

number of studies reporting similar findings for hand and elbow

movement (Calautti and Baron, 2003). We have shown previously

that the activation pattern of paretic elbow movement depends on

lesion location (Luft et al., 2004). This is confirmed here for knee

movement. In our previous study, patients with brainstem strokes

were included in the subcortical group due to limited sample size.

The current sample was large enough to separate these groups: a

marginal difference in the activation pattern is found between them

(P = 0.062). The activation pattern in brainstem patients is closer

to that of cortical lesions (group difference: P = 0.84); the latter

was significantly different from subcortical stroke. Both brainstem

and cortical lesions affect descending motor pathways. Brainstem

strokes may destroy a larger percentage of descending fibers as

they occupy smaller space when traversing the brainstem. Possibly,

differences exist between bsubcorticalQ lesions affecting fiber

systems and basal ganglia as opposed to fiber systems alone, but

no study has so far managed to collect a sufficient number of

patients with well-circumscribed lesions to demonstrate such

differences. The observation that in cortical stroke sensorimotor

cortex recruitment is low seems an expected consequence of

cortical damage. Compensatory activation of areas adjacent to the

lesion is often observed (Calautti and Baron, 2003; Cramer et al.,

1997; Luft et al., 2004). The heterogeneity of cortical lesions and

hence, the variability of the geometry of intact cortex surrounding

these lesions, has likely prevented us from detecting the activation

adjacent to the stroke cavity in multi-subject analysis here. Single-

subject analysis, however, demonstrates cortical recruitment in a

subset of patients.

In healthy subjects, brain activation patterns observed during

unilateral knee movement are similar to the patterns associated

with walking as measured by near infrared spectrocscopy (NIRS,

Miyai et al., 2001). Therefore, we assume that knee movement is

an acceptable paradigm to study lower extremity function in

fMRI where walking cannot be performed (Luft et al., 2002). In a

group of six stroke patients (4 with cortical, 2 with subcortical

lesions), Miyai et al. (2002) observed walking-associated recruit-

ment of bilateral premotor cortex and SMA with lateralization

towards the affected hemisphere. Recruitment of primary motor

cortex showed opposite lateralization. Activation in these areas is

found here for subcortical stroke. Cortical stroke, however, shows

significantly less activation in these regions and no evidence for

lateralization. This may explain why Miyai et al. found intact

hemisphere dominance when a group, consisting mainly of

cortical strokes, moved both legs at the same time and unaffected

leg movement was responsible for most of cortical recruitment.

The contribution of the sensorimotor cortex in the intact

hemisphere to paretic movement in patients with cortical stroke

is disputed (Johansen-Berg et al., 2002; Werhahn et al., 2003).

Our data speak for functional relevance of bilateral S1 in these

patients. Interestingly, higher activation of intact hemisphere S1

was also related to better arm function in cortical stroke patients

in a prior study (Luft et al., 2004). Sensorimotor cortex

recruitment in the intact hemisphere may also be a consequence

of mirror movement. Here, this is unlikely, as such movements

were not observed in cortical stroke patients.

The absence of mirror movements in cortical stroke patients

contrasts a study reporting such movements in the upper

extremities in 2 of 4 cortical stroke patients (Kim et al., 2003).

Although our observation has to be considered with caution given

low sample sizes, statistical significance with adequate power

suggests that mirror movements in the legs are at least infrequent

in patients with cortical stroke. Also unexpected is reduced

contralesional activation in subcortical patients with mirror

movements as compared to those without. This contrasts several

studies documenting activation in contralesional (ipsilateral)

motor areas in conjunction with upper extremity mirror move-

ments (Cohen et al., 1991; Kim et al., 2003; Weiller et al., 1993).

Together, our observations may suggest that the mechanisms for

mirror movements in lower and upper extremities are different.

However, this study was not designed to investigate brain

activation associated with mirror movements. Thus, the post

hoc nature of the assessment, the influence of a number of

uncontrolled confounders such as severity of impairment and

exact lesion site, and the lack of EMG monitoring during fMRI

render the results difficult to interpret and further study is

certainly needed. Given the present analysis, we do, however,

think that the presence of mirror movements does not influence

the main conclusions of the study.

Abnormal activation patterns were found not only for paretic

but also for non-paretic limb movement. Several functional

imaging studies have described alterations in non-paretic move-

ment-associated activation patterns for the upper extremity

(Cramer et al., 1997; Luft et al., 2004). Post-stroke modifications

in neuronal networks controlling the paretic limb, especially

compensatory recruitment of the non-lesioned hemisphere, may

affect cortical areas in control of the non-paretic limb. Also, non-

use of both lower extremities due to impaired walking or altered

limb kinematics and body posture due to hemiparesis may induce

neural adaptations in networks controlling the intact limb.

Higher walking speed as a measure of locomotor ability was

associated with higher or lower activation of several brain regions.

In brainstem stroke, lesser recruitment of contralateral motor

cortices in the ipsilesional hemisphere was associated with better

walking. For subcortical stroke, we did find evidence that higher

activation in contralesional and lower activation in ipsilesional

sensorimotor cortex was linked to faster walking. At first, these

findings seem to challenge the hypothesis that the recovery of

perilesional activation is an indicator of functional improvement

(Calautti and Baron, 2003). However, studies leading to this

hypothesis did not investigate lower extremity movement, the

recovery of which may be substantially different as compared with

the arm. In cortical stroke patients, who presented little activation

overall, stronger recruitment of several areas in both hemispheres

was associated with faster walking. This may be consistent with the

idea that compensatory recruitment of bilateral cortical regions is


required when cortex itself is damaged; the better this compensa-

tion evolves, the better motor abilities can be attained. These

conclusions remain speculative and await further testing in larger

patient samples with levels of impairment more widely distributed

than the sample investigated here.

A limitation of the study is that contractions of the resting limb

resulting from the attempt to stabilize the trunk during knee

movements were not controlled. We designed the Plexiglas

scaffold and the movement task to provide a high degree of

stabilization. Stabilization by contraction of contralateral proximal

muscles is more likely in patients with higher impairment. Because

the degree of impairment was similar between patient groups,

stabilization is not expected to affect the comparison between

patient groups. The comparison between patients and controls may

be affected. But, we think that this error is minimal. If it were not,

patients should show higher contralesional motor cortex recruit-

ment reflecting intact limb activation as compared with controls—

this was not observed.

Conclusion

This study demonstrates that central adaptations occur in

networks controlling the paretic as well as the non-paretic lower

limb after stroke. The nature of these changes depends on lesion

location. Such adaptations may indicate cortical reorganization that

may be specifically accessed by neurorehabilitative interventions

to improve locomotor function in chronically impaired stroke

survivors.

Acknowledgments

We thank Andrew Goldberg, Jim Boyd, Christina Stephenson,

Jill England, and Justin Rajendra for their support. We thank the

F.M. Kirby Center for Functional Brain Imaging, Kennedy Krieger

Institute (Baltimore, MD, USA), and its staff, especially Terry

Brawner, Dr. James Pekar, and Dr. Peter van Zijl. Dr. Luft is

supported by grants from Deutsche Forschungsgemeinschaft (Lu

748/3). Funding for this study was obtained from the National

Institutes of Health (P60AG 12583 NIA University of Maryland

Claude D. Pepper Older Americans Independence Center) the

Baltimore Department of Veterans Affairs Medical Center Geriatrics

Research, Education and Clinical Center (GRECC), Dr. Hanley is

supported by 1RO1 NS 24282-08 and the France-Merrick Founda-

tion, the Johns Hopkins GCRC (grant NCRR #MO1-00052), the

National Center for Research Resources, MO1 RR-00052, and the

Eleanor Naylor Dana Charitable Trust.

References

Calautti, C., Baron, J.C., 2003. Functional neuroimaging studies of motor

recovery after stroke in adults: a review. Stroke 34, 1553–1566.

Chollet, F., DiPiero, V., Wise, R.J., Brooks, D.J., Dolan, R.J., Frackowiak,

R.S., 1991. The functional anatomy of motor recovery after stroke in

humans: a study with positron emission tomography. Ann 29, 63–71.

Cohen, L.G., Meer, J., Tarkka, I., Bierner, S., Leiderman, D.B., Dubinsky,

R.M., Sanes, J.N., Jabbari, B., Branscum, B., Hallett, M., 1991.

Congenital mirror movements. Abnormal organization of motor path-

ways in two patients. Brain 114 (Pt. 1B), 381–403.

Cramer, S.C., Nelles, G., Benson, R.R., Kaplan, J.D., Parker, R.A., Kwong,

K.K., Kennedy, D.N., Finklestein, S.P., Rosen, B.R., 1997. A functional

MRI study of subjects recovered from hemiparetic stroke. Stroke 28,

2518–2527.

Feydy, A., Carlier, R., Roby-Brami, A., Bussel, B., Cazalis, F., Pierot, L.,

Burnod, Y., Maier, M.A., 2002. Longitudinal study of motor recovery

after stroke: recruitment and focusing of brain activation. Stroke 33,

1610–1617.

Grefkes, C., Geyer, S., Schormann, T., Roland, P., Zilles, K., 2001.

Human somatosensory area 2: observer-independent cytoarchitectonic

mapping, interindividual variability, and population map. NeuroImage

14, 617–631.

Hallett, M., 2001. Plasticity of the human motor cortex and recovery from

stroke. Brain Res., Brain Res. Rev. 36, 169–174.

Johansen-Berg, H., Rushworth, M.F., Bogdanovic, M.D., Kischka, U.,

Wimalaratna, S., Matthews, P.M., 2002. The role of ipsilateral premotor

cortex in hand movement after stroke. Proc. Natl. Acad. Sci. U. S. A.

99, 14518–14523.

Kim, Y.H., Jang, S.H., Chang, Y., Byun, W.M., Son, S., Ahn, S.H., 2003.

Bilateral primary sensori-motor cortex activation of post-stroke mirror

movements: an fMRI study. NeuroReport 14, 1329–1332.

Lancaster, J.L., Summerln, J.L., Rainey, L., Freitas, C.S., Fox, P.T., Evans,

A.E., Toga, A.W., Mazziotta, J.C., 1997. The Talairach Daemon, a

database server for Talairach atlas labels. Neuroimage. NeuroImage 5,

S633.

Luft, A.R., Smith, G.V., Forrester, L., Whitall, J., Macko, R.F., Hauser,

T.K., Goldberg, A.P., Hanley, D.F., 2002. Comparing brain activation

associated with isolated upper and lower limb movement across

corresponding joints. Hum. Brain Mapp. 17, 131–140.

Luft, A.R., Waller, S., Forrester, L., Smith, G.V., Whitall, J., Macko, R.F.,

Schulz, J.B., Hanley, D.F., 2004. Lesion location alters brain

activation in chronically impaired stroke survivors. NeuroImage 21,

924–935.

Macko, R.F., DeSouza, C.A., Tretter, L.D., Silver, K.H., Smith, G.V.,

Anderson, P.A., Tomoyasu, N., Gorman, P., Dengel, D.R., 1997.

Treadmill aerobic exercise training reduces the energy expenditure and

cardiovascular demands of hemiparetic gait in chronic stroke patients. A

preliminary report. Stroke 28, 326–330.

Marshall, R.S., Perera, G.M., Lazar, R.M., Krakauer, J.W., Constantine,

R.C., DeLaPaz, R.L., 2000. Evolution of cortical activation during

recovery from corticospinal tract infarction. Stroke 31, 656–661.

Mayo, N.E., Dauphinee, S., Ahmed, S., Gordon, C., Higgins, J., McEwen,

S., Salbach, N., 1999. Disablement following stroke. Disabil. Rehabil.

21, 258–268.

Miyai, I., Tanabe, H.C., Sase, I., Eda, H., Oda, I., Konishi, I., Tsunazawa,

Y., Suzuki, T., Yanagida, T., Kubota, K., 2001. Cortical mapping of gait

in humans: a near-infrared spectroscopic topography study. NeuroImage

14, 1186–1192.

Miyai, I., Yagura, H., Oda, I., Konishi, I., Eda, H., Suzuki, T., Kubota, K.,

2002. Premotor cortex is involved in restoration of gait in stroke. Ann.

Neurol. 52, 188–194.

Miyai, I., Yagura, H., Hatakenaka, M., Oda, I., Konishi, I., Kubota, K.,

2003. Longitudinal optical imaging study for locomotor recovery after

stroke. Stroke 34, 2866–2870.

Moseley, A.M., Stark, A., Cameron, I.D., Pollock, A., 2003. Treadmill

training and body weight support for walking after stroke. Cochrane

Database Syst. Rev. (CD002840). DOI:10.1002/14651858.CD002840.

Seitz, R.J., Hfflich, P., Binkofski, F., Tellmann, L., Herzog, H., Freund,

H.J., 1998. Role of the premotor cortex in recovery from middle

cerebral artery infarction. Arch 55, 1081–1088.

Smith, G.V., Silver, K.H., Goldberg, A.P., Macko, R.F., 1999. bTask-orientedQ exercise improves hamstring strength and spastic reflexes in

chronic stroke patients. Stroke 30, 2112–2118.

Ward, N.S., Brown, M.M., Thompson, A.J., Frackowiak, R.S., 2003.

Neural correlates of outcome after stroke: a cross-sectional fMRI study.

Brain 126, 1430–1448.

Weiller, C., Chollet, F., Friston, K.J., Wise, R.J., Frackowiak, R.S., 1992.

http:\\dx.doi.org\doi:10.1002\14651858.CD002840


Functional reorganization of the brain in recovery from striatocapsular

infarction in man. Ann. Neurol. 31, 463–472.

Weiller, C., Ramsay, S.C., Wise, R.J., Friston, K.J., Frackowiak, R.S., 1993.

Individual patterns of functional reorganization in the human cerebral

cortex after capsular infarction. Ann. Neurol. 33, 181–189.

Werhahn, K.J., Conforto, A.B., Kadom, N., Hallett, M., Cohen, L.G., 2003.

Contribution of the ipsilateral motor cortex to recovery after chronic

stroke. Ann. Neurol. 54, 464–472.

Zilles, K., Schlaug, G., Matelli, M., Luppino, G., Schleicher, A., Qu, M.,

Dabringhaus, A., Seitz, R., Roland, P.E., 1995. Mapping of human and

macaque sensorimotor areas by integrating architectonic, transmitter

receptor, MRI and PET data. J Anat 187 (Pt. 3), 515–537.

Brain activation of lower extremity movement in chronically impaired stroke survivors

Documents

Transcript of Brain activation of lower extremity movement in chronically impaired stroke survivors