neurorehabilitation and neural repairwkdurfee/publications/nnr_2007.pdfneurorehabilitation and...

http://nnr.sagepub.comRepair

Neurorehabilitation and Neural

DOI: 10.1177/1545968306292381 2007; 21; 216 originally published online Mar 9, 2007; Neurorehabil Neural Repair

Scott M. Lewis James R. Carey, William K. Durfee, Ela Bhatt, Ashima Nagpal, Samantha A. Weinstein, Kathleen M. Anderson and

Function and Cortical Reorganization After StrokeComparison of Finger Tracking Versus Simple Movement Training via Telerehabilitation to Alter Hand

http://nnr.sagepub.com/cgi/content/abstract/21/3/216 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

On behalf of:

American Society of Neurorehabilitation

can be found at:Neurorehabilitation and Neural Repair Additional services and information for

http://nnr.sagepub.com/cgi/alerts Email Alerts:

http://nnr.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://nnr.sagepub.com/cgi/content/abstract/21/3/216#BIBLSAGE Journals Online and HighWire Press platforms):

(this article cites 42 articles hosted on the Citations

© 2007 American Society of Neurorehabilitation. All rights reserved. Not for commercial use or unauthorized distribution. by guest on May 10, 2007 http://nnr.sagepub.comDownloaded from

http://www.asnr.com/

http://nnr.sagepub.com/cgi/alerts

http://nnr.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://nnr.sagepub.com/cgi/content/abstract/21/3/216#BIBL

http://nnr.sagepub.com

216 Copyright © 2007 The American Society of Neurorehabilitation

Comparison of Finger Tracking Versus SimpleMovement Training via Telerehabilitation

to Alter Hand Function and CorticalReorganization After Stroke

James R. Carey, PT, PhD, William K. Durfee, PhD, Ela Bhatt, Ashima Nagpal,Samantha A. Weinstein, MS, Kathleen M. Anderson, PT, MBA, and Scott M. Lewis, MD, PhD

Objective. To compare 2 telerehabilitation training strategies,repetitive tracking movements versus repetitive simple move-ments, to promote brain reorganization and recovery of handfunction. Methods. Twenty subjects with chronic stroke and 10degrees of voluntary finger extension were randomly assignedto receive 1800 telerehabilitation trials over 2 weeks of eithercomputerized tracking training (track group) with theaffected finger and wrist involving temporospatial processingto achieve accuracy or movement training (move group) withno attention to accuracy. Following movement training, themove group crossed over to receive an additional 2 weeks oftracking training. Behavioral changes were measured with theBox and Block test, Jebsen Taylor test, and finger range ofmotion, along with a finger-tracking activation paradigmduring fMRI. Results. The track group showed significantimprovement in all 4 behavioral tests; the move groupimproved in the Box and Block and Jebsen Taylor tests. Theimprovement for the track group in the Box and Block andJebsen Taylor tests did not surpass that for the move group. Aconsistent group pattern of brain reorganization was not evi-dent. The move group, after crossing over, did not show fur-ther significant improvements. Conclusion. Telerehabilitationmay be effective in improving performance in subjects withchronic stroke. Tracking training with reinforcement toenhance learning, however, did not produce a clear advantageover the same amount of practice of random movements. Twoweeks of training may be insufficient to demonstrate a behav-ioral advantage and associated brain reorganization.

Key Words: Stroke—Hand—fMRI—Motor learning—Telemedicine.

An important question in stroke rehabilitation iswhether recovery is influenced more by forceduse versus forced learning.1 Repetitive use of the

paretic limb during simple movements may not opti-mize the conditions for recovery. Research in rodents,2

primates,3 and healthy humans4 suggests that precision-demanding tasks that challenge motor-learning processescreate richer conditions for change.

We questioned whether the improved finger movementand changed brain reorganization demonstrated in anearlier study involving finger-tracking training in stroke5

was due to the repetitive movement or to the cognitiveprocessing engaged to produce accurate movements,as both conditions were convolved. Consequently, theprimary purpose of the present study was to compare 2training strategies in promoting brain reorganization andrecovery of hand function in subjects with stroke: fingermovement tracking that involved temporospatial process-ing to achieve accuracy versus the practice of simple fin-ger movements that required no attention to accuracy.

A 2nd question in the current study concerned thevenue of training. As communication technology hasadvanced, an unconventional method of rehabilitationhas emerged that may allow rehabilitative training tocontinue remotely following discharge from acute care.This method is called telerehabilitation,6-8 defined hereas therapy from a distance directed by a computer andtelecommunication. We questioned whether behavioralimprovements and associated brain reorganizationcould be accomplished in the subject’s own home usinga self-administered, computerized finger-tracking train-ing program with occasional teleconferencing witha remote therapist. Our overall hypothesis was thatsubjects receiving home-based finger-tracking trainingwould experience greater change in finger control andbrain reorganization compared to subjects receivinghome-based simple finger movement training.

From the Program in Physical Therapy, University of Minnesota,Minneapolis (JRC, EB, AN, KMA); Department of MechanicalEngineering, University of Minnesota, Minneapolis (WKD, SAW);Department of Neurology, University of Minnesota Medical School,Minneapolis (SML); and Brain Sciences Center (11B), VA MedicalCenter, Minneapolis (SML).

Address correspondence to James R. Carey, Program in PhysicalTherapy, MMC 388, University of Minnesota, 420 Delaware St. SE,Minneapolis, MN 55455. E-mail: [email protected].

Carey JR, Durfee WK, Bhatt E, Nagpal A, Weinstein SA, Anderson KM,Lewis SM. Comparison of finger tracking versus simple movementtraining via telerehabilitation to alter hand function and cortical reor-ganization after stroke. Neurorehabil Neural Repair 2007;21:216–232.

DOI: 10.1177/1545968306292381



Comparing Telerehabilitation Training Strategies

Neurorehabilitation and Neural Repair 21(3); 2007 217

METHODS

Subjects

Twenty subjects with stroke were randomly assignedto either a track group or move group. Individual char-acteristics are summarized in Table 1. Figure 1 showsthe flow chart summarizing the subject inclusion at thedifferent stages of the study. The mean age (±SD) forthe track group was 65.9 ± 7.4 years, and for the movegroup, it was 67.4 ± 11.8 years (nonsignificant differ-ence). The mean time from stroke onset to entry was42.5 ± 24.3 months in the track group and 35.6 ± 26.1months in the move group (nonsignificant difference).A neurologist identified the location of infarct from MRimages and also classified the infarct as either cortical orsubcortical. The preferred hand prior to stroke wasmeasured by the Edinburgh Inventory.9 All subjects hadsufficient cognition to perform the training and testing.All had Mini-Mental State Examination10 scores of 25 orhigher, except 1 subject with a score of 21, who was stilldeemed appropriate to participate. The maximal scorefor Mini-Mental State Examination is 30, with normalscores ranging from 24 to 30.10

Inclusion criteria were poststroke duration of at least12 months, 30 to 80 years of age, satisfactory correctedvision to recognize the full tracking target and cursormovement, at least 90 deg of passive extension-flexionmovement at the index finger metacarpophalangeal(MP) joint of the paretic hand (no contracture), and atleast 10 deg of active movement at this joint. Exclusioncriteria were indwelling metals or medical implantsincompatible with functional magnetic resonance imag-ing (fMRI), pregnancy, and claustrophobia. Subjectswith subtle kinesthesia deficits, evidenced by impairedability to detect 10-deg passive extension-flexion move-ments at the MP joint with eyes closed, were included inthe study. All subjects were able to detect larger ampli-tude (>10 deg) movements. Subjects were recruited byadvertising in a local newspaper and through announce-ments at local stroke support group meetings. Thisstudy was approved by the institution’s Committee onthe Use of Human Subjects in Research. All subjectssigned a statement of informed consent.

Training: Track Group

Subjects in the track group received finger and wristtracking training in their own homes. This training wasdone independent of any direct supervision by the ther-apist. The training equipment consisted of a laptop com-puter (Dell Latitude 600, Dell, Round Rock, TX) withcustomized tracking software. At the completion of thepretest (below), the subject watched a video showing the

training setup procedures. The subject practiced thesetup once under the supervision of the therapist, andthen the subject performed 5 or 6 trials to become famil-iar with the training. After this orientation, the subjecttook the equipment home.

The subject began a training session by securingcustom-made electrogoniometer braces to each hand.One size was used for all subjects, and the fit wasadjustable with Velcro hand straps and moldablemetallic bands that secured around each person’s wrist(Fig. 2). Each brace included 2 potentiometers (ETISystems Inc., Carlsbad, CA): one signaling extension/flexion movement at the index finger MP joint and theother signaling extension/flexion movement at thewrist. The voltage signal representing joint motion wasdirected to the training computer through an ana-logue-to-digital converter that sampled the signal at100 Hz. These devices were designed so that accurateplacement by the subject at home was not required,that is, there was no need to align anatomical jointswith the potentiometers. The device output was repeat-able, with electrical noise below 1 pixel on the displayscreen. As demonstrated through mathematical calcu-lations using the link geometry, the device outputapproximated, but was not linearly related to, jointangle, particularly at extreme angles. However, thefunction of the devices used for the training sessionsdid not require a linear joint angle measure, only ameans to control the screen cursor with joint motion.For training, the subject sat in front of the computerscreen with forearms resting on the chair’s armrests. Ateach training session, subjects followed brief promptson the screen to record the active range of motion ateach joint, which were then used to set theextension/flexion limits for the tracking waveformsthat followed. Figure 2 shows the training setup.

The program required the subject to perform 180tracking trials per day for 10 days. The 180 trials weredivided into 60 different blocks with 3 consecutive trialsper block that were completed over 2 to 8 h dependingon rest breaks determined by the subject. To make thetelerehabilitation training flexible to each person’sschedule and energy level, and thereby promote compli-ance, we did not control the distribution of the restbreaks.

The 60 training blocks came from a host set of 100blocks. A given block did not repeat until the full 100blocks had been experienced. Possible waveformsincluded square, sawtooth left, sawtooth right, triangle,or variable sine. Frequencies were 0.2, 0.4, or 0.6 Hz.Trial durations were 5, 10, or 15 s. For example, a 0.4 Hzfrequency during a 10-s trial duration involved 4 cyclesof extension-flexion movements, each lasting 2.5 s. Thepeak flexion amplitude of a given target waveform wasset at either 0%, 15%, or 30% of the subject’s full range



218 Neurorehabilitation and Neural Repair 21(3); 2007

Tabl

e 1.

Subj

ect

Ch

arac

teri

stic

s

Age

Stro

ke D

ura

tion

Pare

tic

Pre

ferr

ed H

and

Subj

ect

Gro

up

Sex

(yea

rs)

(mon

ths)

Infa

rct

Cla

ssif

icat

ion

(C

orti

cal o

r Su

bcor

tica

l) a

nd

Loca

tion

Side

Pre

stro

keM

MSE

1Tr

ack

M62

69Su

bcor

tica

l – le

ft c

oron

a ra

diat

a,in

tern

al c

apsu

le,s

tria

tum

RR

252

Trac

kM

7917

Subc

orti

cal –

left

cor

ona

radi

ata,

puta

men

RR

293

Trac

kM

5719

Cor

tica

l – r

igh

t fr

onto

-par

ieta

l cor

tex,

coro

na

radi

ata,

inte

rnal

cap

sule

LR

284

Trac

kM

6236

Subc

orti

cal –

rig

ht

coro

na

radi

ata,

inte

rnal

cap

sule

,th

alam

us

LR

295

Trac

kM

6735

Subc

orti

cal –

rig

ht

coro

na

radi

ata

LR

286

Trac

kM

6538

Cor

tica

l – le

ft f

ron

to-p

arie

tal c

orte

x,co

ron

a ra

diat

aR

R21

7Tr

ack

M74

73Su

bcor

tica

l – le

ft c

oron

a ra

diat

a,in

tern

al c

apsu

leR

R26

8Tr

ack

M70

18C

orti

cal –

left

pos

teri

or f

ron

tal c

orte

x,co

ron

a ra

diat

a,in

tern

al c

apsu

le

RR

309

Trac

kF

5536

Cor

tica

l – r

igh

t fr

onto

-par

ieta

l cor

tex

LR

2810

Trac

kM

6884

Cor

tica

l – r

igh

t fr

onto

-par

ieta

l cor

tex,

coro

na

radi

ata,

basa

l gan

glia

LR

27

Trac

k G

rou

p Su

mm

ary

M:F

Cor

tica

l:Su

bcor

tica

lR

:LR

:L(F

requ

ency

/Mea

n ±

SD

)9:

165

.9 ±

7.4

42.5

± 2

4.3

5:5

5:5

10:0

27.1

± 2

.6

11M

ove

M78

58Su

bcor

tica

l – r

igh

t in

tern

al c

apsu

leL

R28

12M

ove

M55

16Su

bcor

tica

l – le

ft c

oron

a ra

diat

aR

R27

13M

ove

M72

70Su

bcor

tica

l – r

igh

t co

ron

a ra

diat

a L

R26

14M

ove

M65

13C

orti

cal –

rig

ht

fron

to-p

arie

tal c

orte

x,co

ron

a ra

diat

aL

R30

15M

ove

M85

14C

orti

cal –

left

pos

teri

or f

ron

tal c

orte

x,co

ron

a ra

diat

aR

R30

16M

ove

F77

14Su

bcor

tica

l – r

igh

t in

tern

al c

apsu

le,p

uta

men

LL

3017

Mov

eF

4982

Cor

tica

l – r

igh

t fr

onta

l cor

tex,

coro

na

radi

ata

LR

2718

Mov

eF

6644

Subc

orti

cal –

rig

ht

inte

rnal

cap

sule

LR

3019

Mov

eM

7328

Subc

orti

cal –

rig

ht

coro

na

radi

ata,

inte

rnal

cap

sule

,th

alam

us

LR

2820

Mov

eF

5417

Subc

orti

cal –

rig

ht

coro

na

radi

ata,

extr

eme

caps

ule

,cla

ust

rum

,in

sula

LR

30

Mov

e G

rou

p Su

mm

ary

M:F

Cor

tica

l:Su

bcor

tica

lR

:LR

:L(F

requ

ency

/Mea

n ±

SD

)6:

467

.4 ±

11.

835

.6 ±

26.

13:

72:

89:

128

.6 ±

1.6

Stro

ke d

ura

tion

= t

ime

from

str

oke

onse

t to

en

tran

ce in

to s

tudy

;MM

SE =

Min

i-M

enta

l Sta

te E

xam

inat

ion

(n

o u

nit

s);M

= M

ale;

F =

Fem

ale;

R =

Rig

ht;

L =

Lef

t.




TR

AC

K G

RO

UP

(n

= 12

)

Lo

st a

t p

ost

test

(n

= 2)

Tw

o sh

owed

exc

essi

ve

m

otio

n du

ring

MR

I

Lo

st a

t fo

llow

-up

(n

= 2

)

One

mov

ed o

ut o

f tow

n

One

tech

nica

l fai

lure

durin

g M

RI

TR

AC

K G

RO

UP

(n

= 10

)

An

alyz

ed a

t p

ost

test

(n

= 10

)

An

alyz

ed a

t fo

llow

-up

(n

= 8)

MO

VE

GR

OU

P (

n =

10)

An

alyz

ed a

t p

ost

test

(n

= 10

)

An

alyz

ed a

t cr

oss

ove

r (n

= 9

)

TR

AC

K G

RO

UP

(n

= 13

)

Exc

lud

ed (

n =

1)

O

ne d

id n

ot m

eet

in

clus

ion

crite

ria a

fter

co

nsen

t due

to

cl

aust

roph

obia

MO

VE

GR

OU

P (

n =

12)

Exc

lud

ed (

n =

2)

O

ne d

id n

ot m

eet

in

clus

ion

crite

ria a

fter

co

nsen

t due

to n

orm

al

m

otio

n

O

ne r

efus

ed

to

par

ticip

ate

MO

VE

GR

OU

P (

n =

10)

Lo

st a

t p

ost

test

(n

= 0)

Lo

st a

t cr

oss

ove

r (n

= 1

) A

dmitt

ed to

hos

pita

l due

to

med

ical

rea

sons

un

rela

ted

to s

tudy

Exc

lud

ed (

n =

142)

N

ot m

eetin

g in

clus

ion

crite

ria (

n =

71)

R

efus

ed to

par

ticip

ate

(n =

6)

O

ther

rea

sons

(pa

rtic

ipat

ing

in a

noth

er

st

udy

or n

o re

turn

cal

l) (n

= 6

5)

Ass

esse

d f

or

elig

ibili

ty

(n =

167

)

Enr

olle

d w

ithsi

gned

cons

ent

(n =

25)

Crossover/Follow-up

Randomization

Allocation

Analysis

Figu

re 1

.Fl

ow c

har

t sh

owin

g th

e n

um

ber

ofsu

bjec

ts a

t ea

ch s

tage

of

the

stu

dy.



Carey et al


(0% = full flexion), and the peak extension amplitudewas set at 70%, 85%, or 125% of full range (100% = fullextension). Thus, if the subject could extend the joint 60deg from the fully flexed position, and the designatedtracking protocol involved flexion/extension limits of15%/85% of the available range, respectively, the flexionlimit of the tracking target would be 9 deg above thefully flexed position and the extension limit would be9 deg below the fully extended position. The conditionof 125% was included, even though it exceeded thesubject’s active extension range, to stimulate the subjectto strive for further extension range.

The paretic hand was used for 90% of the blocks andthe nonparetic hand for 10%. For a given hand, the jointused during tracking was either the index finger MPjoint (50%) or the wrist (50%).

Hand position was varied to create either stimulus-response compatibility11,12 or incompatibility in thetask. Finger or wrist extension always produced upwardmovement of the cursor on the screen, but the position

of the hand was varied between pronation, neutral, andsupination. The pronated position was the compatibleposition, as upward finger or wrist extension producedupward cursor movement on the computer screen. Withthe hand neutral between pronation and supination,finger or wrist extension occurred in the horizontalplane but produced vertical cursor movement. With thehand supinated, downward finger or wrist extensionproduced upward cursor movement. Incompatibilitywas included because spatial processing has been shownto be a potent factor influencing excitability of theprimary motor area in primates.13

During a trial, the cursor swept automatically fromleft to right across the screen while the subject extendedand flexed the proper joint to track the cursor as accu-rately as possible along the target waveform. Next, thecomputer displayed a pause screen for a time equal tothe duration of the preceding trial. After the 3rd trial ofa given block, the computer showed the icon promptsfor the next block of 3 trials.

Figure 2. Training setup for a subject in the track group. The computer screen shows a “sawtooth right” target waveform and track-ing response. Cartoons at bottom prompt which hand (right or left), joint (finger or wrist), and position (pronated, supinated, or neu-tral) to use. A flashing light-emitting diode on the electrogoniometer also assists subject in recognizing which joint to use (in this case,the right index finger metacarpophalangeal joint). The Web camera above the screen is turned on with a switch on the white box toshow an image of the subject to the remote therapist simultaneous with the subject seeing an image of the therapist on screen. A cel-lular phone with speaker adaptation allows verbal communication with the therapist while video conferencing over the regular tele-phone line. With the keyboard covered for simplicity, pressing the red button on the black box starts, pauses, or ends a training session.



Knowledge of results (KR)14 was provided during thepause at the end of each trial with a computer-calculatedaccuracy score.15 Although Anderson and others16 foundbetter retention of motor skill in healthy subjects with aless frequent schedule of KR, we chose every trial becausewe believed it helped to keep these subjects with strokemotivated. Knowledge of performance (KP)14 was pre-sented less frequently. KP consisted of a computer-generated text comment describing a feature to correct inthe tracking behavior. Algorithms were written to detectsuch behaviors as undershooting the target, lagging/lead-ing the target, or a declining range of extension move-ment during the last quarter of the trial compared to thefirst quarter. If values exceeded a certain threshold, basedon values derived from healthy subjects during pilot test-ing, a comment was displayed to the subject identifyingthe error along with a suggestion for improvement. Forgiven tracking protocols, the threshold for eliciting KPcomments was held constant across training sessions andacross subjects. The schedule of KP, however, was faded,occurring at the end of every 2nd block for the 1st dayand declining to every 20th block by the end of the 10thday. This schedule was based on the principle that exces-sive extrinsic feedback interferes with one’s own intrinsicerror detection capability, which can disrupt learning.17

When KP did occur, it was for only the preceding trial.Included with the KP was a brief motivational commentthat encouraged continued effort.

Teleconferencing was used to reinforce a human inter-action and therapeutic relationship between the subjectand the therapist, albeit from a distance. This technologyconsisted of a cellular phone and a Web camera thatoperated over the Internet connection using the subject’stelephone line. This equipment was issued to each subjectalong with the training computer. The therapist con-tacted the subject about 5 times in total over the 10 train-ing sessions using a cellular phone and Web camera. Also,the therapist had a pager and was available to answer spe-cific questions if the subject called. When the subject acti-vated the Internet connection by turning on the switchon the white box, the subject and the therapist would seeeach other on their respective computer screens. Thetherapist and patient could see each other on the com-puter screen in a 128 × 96 pixel window. Video imagingwas transmitted at 3 frames/s. This system did not showthe actual tracking effort of the subject, nor did it trans-mit the performance records to the therapist. Theserecords were retrieved from the computer when thesubject returned to the laboratory for the posttest.

Training: Move Group

Subjects in the move group received finger and wristmovement training in their own homes. The setup and

procedures for determining the range of motion foreach joint were the same as for the track group. Thenumber of blocks, trials per block, duration of eachtrial, and rest between trials were the same as for thetrack group. During a movement trial, the screenshowed a sweeping cursor but it did not show a target orresponse. The subject was instructed to move the properjoint into extension and flexion using a range thatapproximated their full limits and a rate that approxi-mated 0.4 Hz, as instructed at orientation. Neither KRnor KP was provided to the subject. Motivational com-ments were provided by the computer with the samefrequency given to the track group but were not basedon prior performance. Also, teleconferencing occurredwith the same frequency. At the completion of the 10days (1800 move trials), the posttest was given, afterwhich move subjects crossed over to receive 10 addi-tional days of tracking training.

Testing

For track and move subjects, testing occurred before(pretest) and after (posttest) training. Additionally, tracksubjects received a follow-up test 3 months after theposttest to check for retention and move subjects receiveda crossover test after 10 days of tracking training. All testswere identical and consisted of a Box and Block test,Jebsen Taylor test, finger range of motion test, and finallya finger tracking test simultaneous with fMRI. As thepoststroke duration was at least 12 months for all subjects,we did not establish a baseline of stable motor recoverywith multiple pretests prior to initiation of treatment.

Box and Block test. For this test, subjects grasped 2.54 cm3

blocks from one side of a divided box with the tip of theindex finger and tip of the thumb of the paretic handand released the block on the opposite side of the box.Subjects performed three 60-s trials of grasping andreleasing as many blocks as possible, one at a time.18

Jebsen Taylor test. For each testing point, subjects per-formed 1 trial of the following components of theJebsen Taylor test19: 1) turning over cards, 2) picking upsmall objects (e.g., pennies, paper clips), 3) stackingcheckers, 4) picking up beans with a spoon, 5) turningover large empty cans, and 6) turning over largeweighted cans. We did not use the handwriting compo-nent of this test, because not all subjects used theirparetic hand for handwriting before the stroke. Eachtest was timed separately, and the dependent measurewas the total time to complete all tests. Although theoriginal Jebsen Taylor test did not include a time limit,we set a maximal allowable time for completing anygiven component at 180 s.


Neurorehabilitation and Neural Repair 21(3); 2007 221 © 2007 American Society of Neurorehabilitation. All rights reserved. Not for commercial use or unauthorized distribution.

by guest on May 10, 2007 http://nnr.sagepub.comDownloaded from


Carey et al


Finger range of motion test. An electrogoniometer wasattached to the paretic hand with the potentiometercentered at the MP joint of the index finger. The voltagesignal from this potentiometer was directed to the testcomputer through an analogue-to-digital converter(Interactive Structures, Inc., Bala-Cynwyd, PA) thatsampled the signal at 60 Hz. To determine the range ofmotion at the index finger, subjects were instructed tomake a fist followed by maximum extension of theindex finger. The subject held the finger at the peak ofeach motion for approximately 3 s, and the investigatorpressed a key on the computer to automatically recordthe voltage signal, which was then converted into anangular value (degrees).

Finger movement tracking test. Accuracy of trackingmovements of the index finger was measured duringfMRI. The electrogoniometer remained attached to thesubject’s paretic hand. A 2nd electrogoniometer wasplaced on the nonparetic hand but was used only tomonitor for mirror movements.20 Subjects were posi-tioned supine inside a 3 Tesla magnet (see below). Headmovements were minimized through stabilization padsinside the head coil. A mirror was mounted to the headcoil to allow subjects to view a target waveform that thecomputer displayed on a projection screen. The pareticarm was adducted to the chest. The elbow was flexed to45 deg, and the hand was propped in the pronated posi-tion near the hip, neutral between extension and flexionof the wrist, using foam pads. With this position, fingerextension and flexion occurred in the vertical plane.Subjects were instructed to move only the paretic fin-gers (long, ring, and little fingers were allowed to movewith the tracking index finger) and nothing else.Subjects wore earplugs and headphones to dampen themagnet noises and allow for communication with theinvestigators.

The tracking test consisted of six 1-min phases alter-nating between Rest and Track conditions: Rest1, Track1,Rest2, Track2, Rest3, Track3. For all phases, the computerdisplayed a sinewave target (0.4 Hz) along with the cor-responding prompt, “Rest” or “Track,” at the bottom ofthe screen. The peak-to-peak amplitude was the same atpretest and posttest. The lower (flexion) peaks were setat 15% of the subject’s available finger range of motion,with the subject’s full flexion defined as 0% of the rangeand the subject’s full extension defined as 100%. Theupper (extension) peaks were set at 150%. Thus, atpretest, the extension peaks exceeded the subject’s avail-able range but allowed for improved tracking accuracyat posttest if the subject’s extension range of motionimproved following training. For each Track phase, thecursor swept from left to right across the screen and thesubject attempted to track the target as accurately as

possible with careful extension and flexion movements.For each Rest phase, the subject watched the cursorsweep across the screen but executed no finger move-ments. Two 1-min practice trials occurred before enter-ing the magnet, and a 10-s practice trial occurred againinside the magnet. If necessary, vision was correctedusing appropriate lenses inserted into plastic frames.Investigators visually monitored for extraneous move-ments in the subject throughout the tracking test, plusany subtle mirror movements in the nonparetic indexfinger were detected with the 2nd electrogoniometeridentified above. Subjects who showed such movementswere not included in the fMRI analysis, and thisoccurred for two subjects in the track group.

Tracking accuracy was quantified with an accuracyindex15 defined as

Accuracy index = 100(P-E)/P

Where E is the root-mean-square (rms) error betweenthe target line and the response line and P is the magni-tude of the subject’s target pattern, calculated as the rmsdifference between the sinewave and the midline sepa-rating the upper and lower halves of the sinewave. Themaximum possible score is 100%. Negative scores occurwhen the response line falls on the opposite side of themidline from the target.

fMRI

Anatomical and functional images were acquired usinga whole-body 3 Tesla magnet (Magnetom Trio, Siemens,Munich, Germany) equipped with a standard head coil. Ahigh-resolution (1 mm3), T1-weighted, 3D anatomicalimage dataset (3D FLASH, TR = 20 ms, FA = 30 deg, totalacquisition time = 10:44 min) was acquired over theentire brain to identify appropriate landmarks and serveas a template upon which functional images would beoverlaid. Functional images were obtained while thesubject performed the tracking test and consisted of T2

*-weighted functional MR images of the BOLD (bloodoxygen–level dependent) signal with a slice thickness of 3mm, which were obtained in the transverse plane using agradient echo EPI sequence (TE = 30 ms, TR = 3000 ms,FA = 80 deg, FOV = 192 × 192 mm with a matrix size of64 × 64 leading to a resolution of 3 × 3 × 3 mm). The totalimaged volume extended from the superior pole of thecortex to a depth of 108 mm in 36 interleaved slices. Ablock fMRI design was used whereby 145 MR scans wereacquired for a total scan time of 7:15 min, which coveredthe time for all the alternating rest/track phases plus the3 to 4 s between each condition needed to toggle thecomputer display from one phase to the next.





Analysis

fMRI analysis. Brain Voyager (Brain Innovation B.V.,Maastricht, the Netherlands) software was used forfMRI data preprocessing and analysis. Functionalimages were preprocessed to correct for head motionartifacts, differences in slice scan time acquisition, andtemporal linear trends. The 3D functional volume wasaligned with the corresponding 3D anatomical volume,and both were normalized to standard Talairach space.21

The change in BOLD signal intensity (percentage)between the track and rest conditions was analyzed sep-arately for every subject using a general linear model(GLM) with 7 predictors. One predictor was the trackcondition. Three predictors accounted for translationalmovement of the head in sagittal, coronal, and trans-verse planes, and the remaining 3 predictors accountedfor rotational movement in the same 3 planes. These last6 predictors were entered as covariates in the model andserved to exclude the effect of any movement artifact inthe variability of BOLD signal.

For each subject, regions of interest encompassing theprimary motor area (M1), supplementary motor area(SMA), and premotor cortex (PMC) in each hemi-sphere were drawn manually according to landmarksdescribed earlier.22,23 The primary somatosensory area(S1) was also drawn and included the gray matter com-prising the entire postcentral gyrus.

Once the regions of interest were drawn, statisticalanalyses were conducted to identify voxels with signif-icantly greater BOLD signal intensity during the trackphases compared to rest. For this, a GLM analysis wascarried out to compare the boxcar model of alternat-ing rest-track phases (including the projected hemo-dynamic response) with the observed BOLD signaldata. The voxels were deemed active if they surpassedthe threshold of a false determination rate of less than0.01.24 The raw volume of activation for each region ofinterest was quantified as the total number of activevoxels. However, raw voxel count is prone to physio-logical variations in BOLD signal; thus, voxel countfor each region of interest was normalized in 2 ways.First, within each hemisphere, we calculated the rela-tive volume of activation for each region of interest asfollows:

Second, volumes of activation for a given region ofinterest were compared between hemispheres by usingthe laterality index,20 calculated as:

where ipsilesional refers to the stroke hemisphere con-tralateral to the performing paretic hand and contrale-sional refers to the nonstroke hemisphere ipsilateral tothe performing paretic hand.

Besides exploring the volumes of activation duringtrack phases, we were also interested in studying thechange in BOLD signal intensity during the track phasesversus rest. Thus, for each cluster of active voxels, wecalculated an intensity index as the percentage increasein signal intensity during Track above Rest:

where IntensityTrack is the average BOLD signal intensityduring Track phases and IntensityRest is the averageBOLD signal intensity during Rest phases.

Statistical analysis. Behavioral and fMRI data at pretestand posttest were analyzed statistically with a 2-factor(group × test) analysis of variance (ANOVA) withrepeated measures. Post hoc analysis examining forbaseline differences between groups was done with2-sample t tests. Examination for change from pretest toposttest was done with preplanned paired t tests. Wecompared results from posttest to follow-up within thetrack group and from posttest to crossover within themove group using paired t tests. Statistical significancewas set at P < 0.05.

RESULTS

Examination of the tracking records retrieved fromthe training computers confirmed that each subject didexert effort on each trial, evidenced by the responsetrace. Only on rare trials (<1%) did subjects fail to exe-cute movement throughout the entire trial. We observedthat the move subjects generally selected a higher rate ofmovement than the track subjects. To estimate the dif-ference, we visually counted the movement repetitionsfor all subjects on all trials of 2 training sessions (total =360 trials). The mean (±SD) of the total number ofmovement repetitions for the move group was 13.0(±4.9), compared to 5.5 (±0.8) for the track group(Mann-Whitney U, P = 0.002). Figure 3 illustrates thisdifference in movement repetitions for 1 subject in themove group doing movement training (beforecrossover) and tracking training (after crossover).

Behavioral

For all measures below, there was no significant differ-ence between groups at pretest. For the Box and Block

Intensity index = (IntensityTrack − IntensityRest) × 100

(IntensityRest),

Laterality Index = (Voxel countIpsilesional) − (Voxel countContralesional)

(Voxel countIpsilesional) + (Voxel countContralesional)

,

Relative Volume = (Voxel countspecif ied region) × 100

(Voxel countM1+S1+PMC+SMA).



Carey et al


measure, the ANOVA revealed significant effects for test(P < 0.001) and interaction (P = 0.02). From pretest toposttest, as shown in Figure 4A, the track group improvedfrom a mean (±SE) of 23.9 (±4.4) to 25.9 (±5.5) blocks(P = 0.03) and the move group showed greater improve-ment from 31.7 (±4.9) to 36.6 (±5.9) blocks (P < 0.001).For the track group, 8 subjects returned for a follow-uptest and their performance was unchanged at 28.9 (±5.2)blocks, indicating that the benefit from pretest to posttestwas retained. For the move group, 9 subjects crossedover to receive tracking training. Their performance atcrossover was unchanged at 37.9 (±7.3), indicating thatthe tracking training following their movement trainingdid not result in additional improvement.

For the Jebsen Taylor test, a significant test effect (P =0.002) was found but there was no evidence of differ-ences between groups. From pretest to posttest (Fig.4B), the track group improved from 218.6 (±49.1) to154.3 (±31.5) s (P = 0.01) and the move group

improved from 144.8 (±31.5) to 92.2 (±21.9) s (P =0.03). Performance in the track group at follow-up was154.9 (±42.2) s, and in the move group at crossover, itwas 98.6 (±34.4) s, which were not significantly differ-ent from their respective posttests.

For finger range of motion, significant effects werefound for test (P = 0.02) and interaction (P = 0.03). Frompretest to posttest (Fig. 4C), the track group improvedfrom 64.5 (±10.8) to 86.5 (±8.4) deg (P = 0.004), whereasthe move group showed a nonsignificant change from76.5 (±12.9) to 77.7 (±5.4) deg. Performance in the trackgroup at follow-up was 82.8 (±10.0) deg, and in the movegroup at crossover, it was 78.9 (±6.3) deg, which was notdifferent from their respective posttests.

For finger tracking, a trend toward significance wasfound for interaction (P = 0.07). From pretest toposttest (Fig. 4D), the track group improved from–30.3% (±5.2) to –6.9% (±6.6) (P = 0.02), whereas themove group showed a nonsignificant change from

Figure 3. Finger movement responses of 1 subject from the move group at trial #M1179 of movement training (A) and in the samesubject, following crossover, at trial #T1179 of tracking training (B). During (A), the subject saw only a blank screen and produced26 extension-flexion movements. During B, the subject saw the 8 cycles of the target square wave and the response trace and pro-duced 8 extension-flexion movements. (On the y axis, 0% represents full flexion and 100% represents full extension for each subject.)





–27.6% (±5.2) to –29.1% (±5.5). Performance in thetrack group at follow-up was –1.65%, which was notdifferent from their posttest. For the move group atcrossover, however, tracking accuracy increased to–8.14% (±9.2) (P = 0.06).

Imaging

For the relative volume of activation in the ipsile-sional hemisphere, M1 showed group (P = 0.05) andinteraction (P = 0.05) effects, PMC showed test (P =0.04) and interaction (P = 0.04) effects, and SMAshowed interaction (P = 0.03). From pretest to posttest(Fig. 5), M1 for the track group showed a trend towardincreased activation (P = 0.07), whereas SMA showed asignificant decrease (P = 0.008). For the move group,PMC showed a trend toward increased activation (P =0.08). In the contralesional hemisphere, there were nosignificant findings for any region.

For the laterality index data during the finger trackingtest, interaction was significant in the M1 (P = 0.04) and

approached significance in the S1 (P = 0.06). Figure 6shows that the laterality index in the M1 decreasedsignificantly from pretest to posttest for the move group(P = 0.04), indicating a shift away from ipsilesionaltoward contralesional activation, whereas it increasedfor the track group, although not significantly. Similarly,in the S1, the decrease for the move group approachedsignificance (P = 0.06), whereas the increase for thetrack group was not significant.

For the intensity index in the ipsilesional hemisphere,the M1 showed test (P = 0.004) and interaction (P = 0.05)effects, whereas only test effects were found for the S1 (P= 0.03) and PMC (P = 0.04), with no evidence of groupdifferences. From pretest to posttest (Fig. 7), the decreasein intensity for the M1 in the move group was significant(P = 0.03), whereas the decrease for the track group wasnot. PMC showed a trend toward a significant decrease (P= 0.06) for the track group but not for the move group.However, after crossover, the move group showed anincreased intensity (P = 0.004). In the contralesionalhemisphere, trends toward a test effect were found in theM1 (P = 0.07) and PMC (P = 0.07). Figure 7 shows that

Figure 4. Means (± standard error) for track and move groups on Box and Block test (A), Jebsen Taylor test (B), finger range of motiontest (C), and finger tracking test (D). a = significant increase from pretest (P = 0.03); b = significant increase from pretest (P < 0.001);c = significant decrease from pretest (P = 0.01); d = significant decrease from pretest (P = 0.03); e = significant increase from pretest(P = 0.004); f = significant difference from Track pretest (P = 0.02); g = trend toward significant change from posttest (P = 0.06).



Carey et al


the decrease for the move group was significant (P = 0.05)but not for the track group. The changes from posttest tofollow-up and posttest to crossover were nonsignificantexcept for the contralesional M1 in the move group, whichshowed a significant increase (P = 0.05) in intensity.

Figure 8 shows for 1 subject in the track group theimprovement in tracking responses and correspondingMR scans from pretest to posttest, with a shift towardgreater ipsilesional activation.

Figure 9 shows for 1 subject in the move group thetracking responses and MR scans at pretest, posttest,and post-crossover. In contradistinction to the results ofFigure 8, the laterality index for the M1 showed strongipsilesional (0.85) activation at pretest, shifting to con-tralesional (–0.39) activation at posttest, and evengreater contralesional activation (–0.44) at crossover,despite the improved tracking accuracy.

DISCUSSION

This study examined in subjects with stroke whethertelerehabilitation emphasizing temporospatial process-ing during movement training would be more effectivein improving hand function and brain reorganizationthan movement training without such processing. The

difference in training was the mental effort required toperform the assigned task. Both groups performedrepetitive manual movements, but the track group pre-sumably had to engage greater motor and cognitiveresources to produce accurate movements. Although wedid not quantify the difference in effort, subjects in thetrack group consistently reported that the task wasfatiguing, whereas subjects in the move group did not.The results showed that telerehabilitation for bothgroups was effective in producing functional gains, buta clear advantage for those who performed more tem-porospatial processing (track group) did not material-ize. The track group did show greater improvements intracking accuracy and finger range of motion, butbecause this group trained with tracking and alsotrained to increase their range of motion (using track-ing protocols with amplitudes at 125% of their range),these gains over the move group were expected. Intransferring skill to more functional tasks (Box andBlock, Jebsen Taylor tests), the improvements for thetrack group did not surpass those for the move group.In the Box and Block test, the improvement for the trackgroup was small (2 blocks), and although statisticallysignificant, this change may not be clinically significant.The improvement was actually greater for the movegroup. Despite the behavioral improvements in both

Figure 5. Means (± standard error) of relative volume of activation of the ipsilesional and contralesional primary motor area(M1), primary somatosensory area (S1), premotor cortex (PMC), and supplementary motor area (SMA) for track and move groupsacross tests. a = trend toward significant increase from pretest (P = 0.07); b = significant decrease from pretest (P = 0.008); c = trendtoward significant increase from pretest (P = 0.08).





groups, a consistent pattern of brain reorganizationlinked to these gains was not evident.

We based our hypothesis of greater gains in the trackgroup on studies that had shown a pattern of more brainreorganization following motor tasks with active cogni-tive processing compared to motor tasks without suchprocessing.1-4 The mechanisms underlying these changeswere not determined but may be related to trophic fac-tors in the brain responding to both physical and cogni-tive activity.25-28 Such studies point to the possibility ofneurochemical mechanisms predicated as much or moreon motor learning as on physical activity per se to pro-mote neuroplasticity associated with skills learning.

Two reasons may explain why we did not find an advan-tage with motor learning training in the track group overthe more simple training in the move group. First,although we “yoked” our track subjects and move subjectsso that they had the same number of training trials andtotal training time, differences within training trialsoccurred. The number of extension/flexion movementsproduced by move subjects during any given trial was notcontrolled, and this was important to remove temporalprocessing from the task in the move group. Instead, movesubjects were instructed at orientation to perform move-ments at a “comfortable” rate that we demonstrated.Nonetheless, move subjects produced extension and flex-ion movements at a significantly higher frequency thanthe track group (Fig. 3). Repetitive movement training byitself has been shown to improve manual performance instroke,29 and the possibility exists that the greater number

of extension/flexion finger movements in the move groupcould have offset the temporospatial processing advantagein the track group.

Second, the treatment dosage (total trials), duration(total time), and distribution of rest in the track groupmay not have been adequate to promote the necessarybrain reorganization for consolidation of learning. Thetotal trials devoted to finger tracking training in this studywas 900, whereas our earlier work,5 which did show sig-nificant brain reorganization, involved 1200 trials. Also,the duration of training in that study was 4 weeks, com-pared to 2 weeks in the present study. Karni and others30

showed that brain reorganization during motor learn-ing is dependent on the duration of training in normalsubjects. Finally, the distribution of rest time betweenpractice was more compressed in the current studycompared to our earlier work. Ofen-Noy and others31

suggested that the amount of practice may not be asimportant to learning as the “passage of time” betweenpractice, and Savion-Lemieux and Penhune32 empha-sized that spacing of rest between practice is importantto allow time to process and encode ongoing task infor-mation. Thus, as we took advantage of the convenienceof telerehabilitation training in the subject’s own homeby intensifying the training into a shorter durationwith less time between training sessions, we may haveviolated as yet unknown biologic rules important forskills learning after stroke. As a result, the observedfunctional gains in the track group may not havereached their full potential.

Figure 6. Means (± standard error) of laterality index for the primary motor area (M1), primary somatosensory area (S1), pre-motor cortex (PMC), and supplementary motor area (SMA) for track and move groups across tests. a = significant decrease frompretest to posttest (P = 0.04); b = trend toward significant decrease from pretest (P = 0.06).



Carey et al


In assessing fMRI activity during paretic hand func-tion following training, Feydy and others33 described“focusing” as the gradual restriction of neural activationto the ipsilesional hemisphere and “recruitment” as theextension of activation to the contralesional hemi-sphere. Ward and others34 studied patients with strokelongitudinally and found that patients with lesser recov-ery were more likely to recruit a greater number of brainregions, including contralesionally. They suggested thatthe more expanded and intense activation in patientswith poorer recovery was due to their need to pay closerattention to the motor task.

Our fMRI results show evidence of training-inducedfocusing. For intensity index (Fig. 7), a significant decrease

occurred in the ipsilesional M1, S1, and PMC from pretestto posttest, but this decrease was not specific to eithergroup. However, in the relative volume of activation forthe ipsilesional M1, PMC, and SMA, we did find signifi-cant interaction between groups from pretest to posttest.In particular, the trend with tracking training toward asignificant increase in the M1 proportion of total activa-tion in this hemisphere, combined with the significantdecrease in the SMA proportion, may represent an earlyexpression of training-induced focusing of cortical activa-tion. But longer training time with more trials or morerest between training sessions may be needed to reveal amore causal link between brain reorganization and train-ing-induced improvements in behavior.

Figure 7. Means (± standard error) of intensity index in the ipsilesional and contralesional primary motor area (M1), primarysomatosensory area (S1), premotor cortex (PMC), and supplementary motor area (SMA) for track and move groups across tests.a = trend toward significant decrease from pretest (P = 0.06); b = significant decrease from pretest (P = 0.03); c = significantincrease from posttest to crossover (P = 0.004); d = significant decrease from pretest (P = 0.05); e = significant increase fromposttest to crossover (P = 0.05).





A further expression of training-induced focusing ofcortical activation is suggested by the change in lateralityindex in the M1 and S1. Significant interaction betweengroups from pretest to posttest was found in both regions,but this difference in response was due primarily to theprecipitous decrease in volume of ipsilesional activationin the move group. Although the volume of ipsilesionalactivation in the M1 and S1 did increase in the trackgroup, it was not significant, as it was in our earlier study,5

and the possible reasons may again relate to differencesin dosage, duration, or distribution of rest. The movegroup’s precipitous change from ipsilesional activationtoward contralesional activation in the M1 and S1 was

conspicuous. Different patterns of brain reorganizationexist following stroke,35 and the possibility exists thattraining with simple movements may promote predomi-nantly contralesional M1 and S1 activation, whereas moredifficult tracking training may promote predominantlyipsilesional activation. A number of studies have shownthat with evoked or voluntary activity of the paretichand, greater contralesional cortical activation (i.e.,ipsilateral to the paretic hand) is associated with poorerfunction20,35-39 and that with improved recovery, there is ashift toward ipsilesional activation.34,37 Such contrale-sional activation, through exaggerated interhemisphericinhibition, may actually interfere with excitability of the

Figure 8. Sample tracking responses for 1 subject from the track group tracking with paretic left index finger (0 deg = maximumflexion). Accuracy improved from –20.07% at pretest to 45.01% at posttest. Representative brain activation maps (1 slice) showmainly contralesional activation in the primary motor (M1) and primary somatosensory (S1) areas (green arrows mark central sul-cus) at pretest (M1 laterality index = –0.50 and S1 laterality index = –0.53), shifting in the direction toward more ipsilesional acti-vation at posttest (M1 laterality index = –0.03 and S1 laterality index = –0.18).



Carey et al


Figure 9. Sample tracking responses for 1 subject from move group tracking with paretic right index finger (0 deg = maximumflexion). Tracking accuracy was –21.39% at pretest and –33.01% at posttest and improved to 26.4% at crossover. Brain activationmaps show mainly ipsilesional activation in the primary motor area (M1) (green arrows mark central sulcus) at pretest (M1 later-ality index = 0.85), shifting to contralesional activation at posttest (M1 laterality index = –0.39) with further progression to con-tralesional activation at crossover (M1 laterality index = –0.44).

ipsilesional hemisphere,40,41 and studies have shown thatdisruption of the contralesional M1 with repetitive tran-scranial magnetic stimulation can improve paretic handfunction.42,43 Thus, further studies are needed to deter-mine whether certain training procedures are more apt topromote brain reorganization in the ipsilesional hemi-sphere versus the contralesional hemisphere and whethersuch orchestration influences recovery.

A further conspicuous feature in the move group wastheir absence of change in Box and Block, Jebsen Taylor,and finger range of motion performance (Fig. 4A-C)following crossover. In our previous study,5 the controlgroup showed significant functional improvement aftercrossing over to receive tracking training. The notionthat move subjects here had reached a ceiling effect aftermovement training seems doubtful, as scores were still





far below normal. Mathiowetz and others18 reported forhealthy males and females 65 to 69 years of age that themean (±SD) Box and Block scores were 68.4 (±7.1) and72.0 (±6.2) blocks, respectively, for the right hand and67.4 (±7.8) and 71.3 (±7.7) blocks, respectively, for theleft hand.

The possibility exists that the movement training mayactually have prevented further functional gains fromthe tracking training that followed closely. For example,several clinical studies44,45 suggested that bidirectionalsynaptic plasticity and thresholds for long-term potenti-ation and depression can be affected by evoking corticalmotor responses and by voluntary movements that pre-cede an action. Perhaps, then, the absence of improve-ment in the move group following crossover resultedfrom its recent history of intensive practice prior to thetracking training. Clearly, the conditioning methods andthe time course between conditioning and measurementdiffered between the present study and experimentsdesigned to test the limits of synaptic plasticity.44,46 Thepossibility, however, that overly intensive conditioning(training) blocks synaptic plasticity and motor learningremains unexplored in stroke recovery studies.

We believe that telerehabilitation can offer a valuablemethod for promoting further recovery from stroke, inpart because of the convenience of practice at home at atime and intensity that befits a person’s own poststrokemedical and psychological conditions. As more detailsare discovered in stroke on such factors as forced useversus forced learning, optimal schedules of practicewith rest, contextual interference, and reinforcement, aswell as best dose-response intensity curves for train-ing,46 the adaptability and flexibility of training availedthrough telerehabilitation may lessen impairments anddisability.

ACKNOWLEDGMENTS

This work was supported by the National Institute onDisability and Rehabilitation Research (U.S. Departmentof Education #H133G020145-04) and the NationalInstitutes of Health (National Center for ResearchResources #M01-RR00400 and Biomedical TechnologyResearch Resources #P41-R008079).

REFERENCES

1. Plautz EJ, Milliken GW, Nudo RJ. Effects of repetitive motortraining on movement representations in adult squirrel mon-keys: role of use versus learning. Neurobiol Learn Mem 2000;74:27-55.

2. Kleim JA, Barbay S, Cooper NR, et al. Motor learning-dependentsynaptogenesis is localized to functionally reorganized motorcortex. Neurobiol Learn Mem 2002;77:63-77.

3. Nudo R, Milliken G, Jenkins W, et al. Use-dependent alterationsof movement representations in primary motor cortex of adultsquirrel monkeys. J Neurosci 1996;16(2):785-807.

4. Pascual-Leone A, Nguyen KT, Cohen AD, et al. Modulation ofmuscle responses evoked by transcranial magnetic stimulationduring the acquisition of new fine motor skills. J Neurophysiol1995;74:1037-45.

5. Carey JR, Kimberley TJ, Lewis SM, et al. Analysis of fMRI andfinger tracking training in subjects with chronic stroke. Brain2002;125:773-88.

6. Piron L, Tonin P, Atzori AM, et al. Virtual environment systemfor motor tele-rehabilitation. Stud Health Technol Informat 2002;85:355-61.

7. Reinkensmeyer DJ, Pang CT, Nessler JA, et al. Web-based telere-habilitation for the upper extremity after stroke. IEEE TransNeural Syst Rehabil Eng 2002;10(2):102-8.

8. Holden MK, Dyar TA, Schwamm L, et al. Virtual-environment-based telerehabilitation in patients with stroke. Presence:Teleoperat Virtual Environ 2005;14(2):214-33.

9. Oldfield R. The assessment and analysis of handedness: theEdinburgh inventory. Neuropsychologia 1971;9(1):97-113.

10. Folstein M, Folstein S, McHugh P. “Mini-mental state”: a practi-cal method for grading the cognitive state of patients for the clin-ician. J Psychiatr Res 1975;12:189-98.

11. Proctor R, Reeve T. Stimulus-response compatibility: an inte-grated perspective. Elsevier (North-Holland): Elsevier Science;1990.

12. Carey J, Bogard C, Youdas J, et al. Stimulus-response compatibil-ity effects in a manual tracking task. Percept Mot Skills 1995;81:1155-70.

13. Georgopoulos AP, Lurito JT, Petrides M, et al. Mental rotationof the neuronal population vector. Science 1989;243(4888):234-6.

14. Schmidt RA, Lee TD. Motor control and learning: a behavioralemphasis. 3rd ed. Champaign (IL): Human Kinetics; 1999.

15. Carey JR. Manual stretch: effect on finger movement control andforce control in stroke subjects with spastic extrinsic finger flexormuscles. Arch Phys Med Rehabil 1990;71:888-94.

16. Anderson DI, Magill RA, Sekiya H. Motor learning as a functionof KR schedule and characteristics of task-intrinsic feedback. JMot Behav 2001;33(1):59-66.

17. Schmidt RA. Frequent augmented feedback can degrade learn-ing: evidence and interpretations. In: Requin J, Stelmach G, eds.Tutorials in motor neuroscience. Dordrecht (the Netherlands):Kluwer Academic; 1991:59-75.

18. Mathiowetz V, Volland G, Kashman N, et al. Adult norms for theBox and Block Test of manual dexterity. Am J Occup Ther1985;39:386-91.

19. Jebsen RH, Taylor N, Trieschmann RB, et al. An objective andstandardized test of hand function. Arch Phys Med Rehabil1969;50(6):311-19.

20. Cramer S, Nelles G, Benson R, et al. A functional MRI studyof subjects recovered from hemiparetic stroke. Stroke 1997;28:2518-27.

21. Talairach J, Tournoux P. Co-planar stereotoxic atlas of the humanbrain. New York: Thieme; 1988.

22. Dassonville P, Lewis SM, Zhu XH, et al. Effects of movementpredictability on cortical motor activation. Neurosci Res 1998;32:65-74.

23. Dassonville P, Lewis SM, Zhu XH, et al. The effect of stimulus-response compatibility on cortical motor activation. Neuroimage2001;13(1):1-14.

24. Genovese CR, Lazar NA, Nichols T. Thresholding of statisticalmaps in functional neuroimaging using the false discovery rate.Neuroimage 2002;15(4):870-8.

25. Gomez-Pinilla F, So V, Kesslak JP. Spatial learning and physicalactivity contribute to the induction of fibroblast growth factor:



Carey et al


neural substrates for increased cognition associated with exer-cise. Neuroscience 1998;85(1):53-61.

26. Klintsova AY, Dickson E, Yoshida R, et al. Altered expression ofBDNF and its high-affinity receptor TrkB in response to complexmotor learning and moderate exercise. Brain Res 2004;1028(1):92-104.

27. Will B, Galani R, Kelche C, et al. Recovery from brain injury inanimals: relative efficacy of environmental enrichment, physicalexercise or formal training (1990-2002). Prog Neurobiol 2004;72(3):167-82.

28. Lu B. Acute and long-term synaptic modulation by neu-rotrophins. Prog Brain Res 2004;146:137-50.

29. Butefisch C, Hummelsheim H, Denzler P, et al. Repetitive trainingof isolated movements improves the outcome of motor rehabilita-tion of the centrally paretic hand. J Neurol Sci 1995;130(1):59-68.

30. Karni A, Meyer G, Jezzard P, et al. Functional MRI evidence foradult motor cortex plasticity during motor skill learning. Nature1995;377:155-8.

31. Ofen-Noy N, Dudai Y, Karni A. Skill learning in mirror reading:how repetition determines acquisition. Cognit Brain Res 2003;17(2):507-21.

32. Savion-Lemieux T, Penhune VB. The effects of practice and delayon motor skill learning and retention. Exp Brain Res 2005;161(4):423-31.

33. Feydy A, Carlier R, Roby-Brami A, et al. Longitudinal study ofmotor recovery after stroke; recruitment and focusing of brainactivation. Stroke 2002;33:1610-17.

34. Ward NS, Brown MM, Thompson AJ, et al. Neural correlates ofmotor recovery after stroke: a longitudinal fMRI study. Brain2003;126:2476-96.

35. Calautti C, Baron JC. Functional neuroimaging studies of motorrecovery after stroke in adults—a review. Stroke 2003;34(6):1553-66.

36. Johansen-Berg H, Dawes H, Guy C, et al. Correlation betweenmotor improvements and altered fMRI activity after rehabilitative

therapy. [Erratum appears in Brain 2003;126(Pt 11):2569]. Brain2002;125(Pt 12):2731-42.

37. Dong Y, Dobkin B, Cen S, et al. Motor cortex activation duringtreatment may predict therapeutic gains in paretic hand functionafter stroke. Stroke 2006;37:1552-55.

38. Traversa R, Cicinelli P, Pasqualetti P, et al. Follow-up of inter-hemispheric differences of motor evoked potentials from the‘affected’ and ‘unaffected’ hemispheres in human stroke. BrainRes 1998;803(1-2):1-8.

39. Netz J, Lammers T, Homberg V. Reorganization of motor outputin the non-affected hemisphere after stroke. Brain 1997;120(Pt9):1579-86.

40. Duque J, Hummel F, Celnik P, et al. Transcallosal inhibition inchronic subcortical stroke. Neuroimage 2005;28(4):940-6.

41. Murase N, Duque J, Mazzocchio R, et al. Influence of interhemi-spheric interactions on motor function in chronic stroke. AnnNeurol 2004;55(3):400-9.

42. Mansur CG, Fregni F, Boggio PS, et al. A sham stimulation-con-trolled trial of rTMS of the unaffected hemisphere in strokepatients. Neurology 2005;64:1802-4.

43. Takeuchi N, Chuma T, Matsuo Y, et al. Repetitive transcranialmagnetic stimulation of contralesional primary motor corteximproves hand function after stroke. Stroke 2005;36(12):2681-6.

44. Siebner HR, Lang N, Rizzo V, et al. Preconditioning of low-frequency repetitive transcranial magnetic stimulation withtranscranial direct current stimulation: evidence for homeostaticplasticity in the human motor cortex. J Neurosci 2004;24(13):3379-85.

45. Ziemann U, Ilic TV, Pauli C, et al. Learning modifies subse-quent induction of long-term potentiation-like and long-termdepression-like plasticity in human motor cortex. J Neurosci2004;24:1666-72.

46. Dobkin B. Rehabilitation and functional neuroimaging dose-response trajectories for clinical trials. Neurorehabil NeuralRepair 2005;19:276-82.



neurorehabilitation and neural repairwkdurfee/publications/nnr_2007.pdfneurorehabilitation and...

Documents