Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Stroke. Author manuscript; available in PMC 2012 February 1.
Published in final edited form as:
PMCID: PMC3026869

Anatomy of Stroke Injury Predicts Gains from Therapy


Background and Purpose

Many therapies are emerging that aim to improve motor function in people with stroke. Identifying key biological substrates needed for treatment gains would help to predict treatment effects, and to maximize treatment impact. The current study addressed the hypothesis that behavioral gains from therapy targeting distal upper extremity are predicted by the structural integrity of key motor system white matter tracts.


Twenty three subjects with chronic left-sided stroke underwent robotic therapy targeting the distal right upper extremity. MRI was obtained at baseline and used to outline the infarct. For each subject, the degree to which stroke injured each of four descending white matter tracts (from primary motor cortex, supplementary motor area, dorsal premotor cortex, and ventral premotor cortex, respectively) was determined. Correlations between tract-specific injury and behavioral gains from therapy were then examined.


Numerous examples were found whereby tract-specific injury predicted treatment gains. The strongest correlations related to stroke injury to tracts descending from primary motor cortex and dorsal premotor cortex. Infarct volume and baseline behavior were weak predictors of treatment gains.


Extent of injury to specific motor tracts predicts behavioral gains from treatment in subjects with chronic stroke. This supports a role for these tracts in mediating treatment effects, and reinforces the importance of lesion location in stroke. Tract-specific injury was stronger than infarct volume or baseline clinical status at predicting gains, identifies subjects with sufficient biological substrate to improve from therapy, and so might be useful as an entry criterion in repair-based trials.

Keywords: corticospinal tract, stroke, prediction, standardized therapy

Stroke remains a major cause of human disability. Motor deficits are among the most common after stroke and thus a major contributor to this fact1, 2. Involvement of the hand can be particularly disabling given the unique role the hand plays in many daily functions. In response to this, a number of therapies are emerging that aim to improve hand motor function for patients in the chronic phase of stroke. Brain mapping studies3-5 suggest that many of the effects of such therapies are mediated by the surviving motor system of the stroke-affected hemisphere, particularly ipsilesional primary motor cortex (M1), supplementary motor area (SMA), dorsal premotor cortex (PMd), and ventral premotor cortex (PMv). The extent of injury to key motor system structures6 might therefore be a determinant of treatment effect and might also provide insight into the inter-subject variability in treatment response, which is substantial.

Cortical motor regions might exert these effects via direct corticospinal projections that have been described from each of M1, SMA, PMd, PMv7, 8. The primary hypothesis examined in this study is that behavioral gains during a course of standardized therapy targeting distal arm are predicted by the structural integrity of motor system white matter tracts. The corticospinal tract from M1 has the largest number of pyramidal tract neurons9 and is the largest source of corticospinal fibers7, and so this tract was of greatest interest.


Study design and subjects

Twenty-three subjects with hemiparesis and chronic stroke were recruited5, 10, signed informed consent in accordance with the U.C. Irvine Institutional Review Board, received baseline behavioral assessments followed by an MRI scan, underwent 23-24 hours of robotic therapy, then underwent repeat behavioral assessment following completion of therapy. Entry criteria are described in the Supplement.

Robotic hand therapy

Therapy consisted of repeated grasp-release movements by the affected distal upper extremity using a pneumatically actuated robot that has been described previously5.

Subject assessments

Baseline behavioral assessments were performed twice, prior to beginning therapy, 1-3 weeks apart (to insure stable baseline), then averaged for each subject. Baseline exam included assessment of medical history, NIH stroke scale (NIHSS) score, depression11, and apraxia12. MRI included high-resolution (1 mm3 voxel) T1-weighted anatomical image, followed by therapy ~1 week later, then behavioral assessments at the end of therapy. The three main behavioral assessments were arm motor Fugl-Meyer (FM) score, Box and Blocks (B&B) score, and the Action Research Arm Test (ARAT), each of which has established validity and inter-/intra- rater reliability. All assessments were performed by a single examiner.

Data analysis

The main independent measure in this study was the structural integrity of motor system white matter tracts. Because fiber tracking with diffusion tensor imaging (DTI) can be problematic in brain regions affected by stroke, tract injury for each patient was determined by examining the extent to which the infarct overlapped with the normal white matter tract13, 14. Areas of infarct were outlined by hand on each subject’s MRI scan, from which binary stroke masks were generated. These images were then transformed into MNI (Montreal Neurological Institute) standard stereotaxic space using FSL (FMRIB Software Library).

The normal white matter tract descending from each of four motor-related cortical areas was generated from in 12 healthy right-handed healthy subjects using diffusion weighted images obtained at 3T, as described elsewhere14. Each tract was generated using a unilateral seed region that covered either M1, SMA, PMd, or PMv, respectively. For each, tractographic methods were used to generate probability pathways connecting the motor cortex area to the cerebral peduncle. These tracts were then binarized, transformed into MNI space, summed to generate group probability maps, and then thresholded to include only voxels that were common to the tracts of ≥8 subjects.

Ideally, we would like to measure the percentage of axons in each tract that was injured by stroke, but MRI resolution does not permit this. Towards this approach, we divided each tract into longitudinal subsections, aiming to model the trajectory of groups of axons. The goal was to classify each subsection as injured or not; a subsection only needed to be damaged by stroke once along its length for it to be classified as injured. Therefore, for primary analyses, each tract was divided into 16 separate descending subsections (further methodological details in Supplement).

The extent of injury to each subsection of each tract was determined by measuring the volume of overlap between that subsection and the stroke mask. For each subsection we applied an empirically defined conservative threshold of 5% overlap to designate the subsection as injured by stroke. For each subject, extent of injury to each descending tract was then determined by measuring the percentage of subsections that were injured by stroke.

To address the primary study hypothesis, for each tract, the relationship between extent of injury (% of its 16 longitudinal subsections injured by stroke, at 5% injury overlap threshold) and change in behavior across the period of therapy was then examined, all using parametric statistics. Behavior was examined in three ways (FM, B&B, ARAT), and so for each tract, significance was corrected for multiple comparisons (p < 0.0167). The relationship between tract injury and change in score across treatment was examined using age and baseline score as covariates. To gain additional insight, the relationship between tract injury and baseline score was also examined.

Secondary analyses examined the effect of adding three clinical covariates (time post-stroke, Geriatric Depression Scale score, and a measure of apraxia) or the effect of instead combining all four measures of tract injury as predictors.

In order to examine the robustness of the methods, additional secondary analyses were performed after assumptions made in data processing were varied in two ways. First, the definition of injury was varied: the degree of overlap between the stroke mask and a tract subsection used to label that subsection as injured was varied from the primary method of 5% (above) to two more stringent definitions, 10% and 20%, and injury/behavior relationships were recalculated. Second, the number of subsections into which each tract was subdivided was changed from the primary method of 16 (above) to three alternatives: 64 subdivisions/tract (whereby each of the 16 subdivisions was quartered), 4 subdivisions/tract, and zero subdivisions/tract. For the zero subdivision approach, injury was reported as the percentage of each tract that overlapped with the stroke mask.


Table 1 presents patient characteristics.

Table 1
Characteristics of subjects with stroke

The tracts and their injury by stroke

Each of the four tracts extended from white matter immediately subjacent to the motor cortex region of interest to the cerebral peduncle14. Some tracts showed limited overlap, and some cortical regions that were nearer to each other, such as PMd and PMv, showed higher overlap (Table 2).

Table 2
Degree to which motor tracts overlap in healthy controls

Across the 23 stroke subjects, the range of tract injury was wide, with the extent of overlap between infarct and descending tracts ranging from 0 - 100% for M1, PMd, and PMv; and from 0 - 93.8% for SMA. The mean (± SEM) overlap between infarct and each motor tract was similar: for the tract descending from M1, an average of 71.2 ± 6.3% (i.e., approximately 11 of 16 subsections) of the tract was injured by stroke; for PMd, 71.7 ± 7.4%; for PMv, 66.1 ± 8.1%; and for SMA, 62.5 ± 6.4%.

The relation between injury and behavioral status

Primary analysis evaluated behavior in relation to tract injury, defined with each tract divided into 16 subsections, and requiring that a patient’s infarct have at least 5% overlap with the tract subsection to designate that subsection as injured.

Injury and behavior at baseline

The baseline value for the three motor outcome assessments did not show a significant correlation with injury to any of the four tracts or with infarct volume. However, review of the data (Figure 1a) suggests that this lack of linearity between injury and baseline behavior is explained by the presence of several subgroups. The first subgroup of subjects had mild tract injury to M1 and mild-moderate baseline motor deficits (and maximum therapy gains, see below); among subjects with more severe tract injury, a second subgroup had severe baseline deficits (FM score ≤ 35) while a third subgroup had mild-moderate baseline deficits (FM score ≥ 44).

Figure 1
Relationships between injury and behavior

Injury and change in behavior across the period of therapy

Baseline scores alone did not predict change in scores, with a single exception (Table 3); nor did NIHSS score or demographic measures. However, with tract-specific injury added to the prediction model, numerous examples were found whereby change in motor status was significantly predicted (Figures 1b and and2,2, Table 3). Note that adding infarct volume instead did not predict change in behavior (Figure 1c, Table 3). There were some similarities in injury effects: extent of injury to tracts descending from M1, PMd, and SMA each correlated significantly with change in FM score as well as with change in B&B score, while injury to the tract descending from PMv correlated only with change in B&B score. For no tract did injury correlate with change in ARAT score.

Figure 2
Examples of stroke injury to the tract descending from M1
Table 3
Prediction of motor change across the period of therapy

In secondary analyses, the effect of adding covariates to the model was examined, focusing on the relationship between M1 tract injury and change in FM score. The initial model (Table 3) had r2 of 0.42, with partial correlation coefficient between M1 injury and change in FM score of -0.65. Adding three clinical measures (time post-stroke plus measures of depression and apraxia) only increased r2 to 0.44 with no effect on the partial correlation coefficient. Adding instead the 3 other tract injury measures increased r2 to 0.51, but partial correlation coefficients were each ≤0.35, consistent with the partial degree of tract overlap (Table 2).

Robustness of results across methodological assumptions

Injury to the M1 tract was further studied to examine the robustness of the methods. Varying the extent to which stroke had to overlap with a subsection in order to label that subsection as injured had little effect. Changing the number of tract partitions from 16 to either 4 or 64 subdivisions/tract reduced injury/behavior correlations for B&B and had little effect on FM and ARAT; use of 0 subdivisions/tract consistently reduced injury/behavior correlations (see Table 4, Supplement).


The main finding from this study is that the extent of injury to specific motor tracts is useful for predicting behavioral gains from therapy in subjects with chronic stroke. Tract-specific injury was stronger than infarct volume or baseline behavioral status at predicting gains; note that baseline behavioral status is often an entry criterion in stroke trials, suggesting that measures of tract injury could be useful to some clinical trials, such as for patient stratification. Tract-specific injury is a relatively simple measure to determine, and is useful to identify subjects who have sufficient biological substrate to improve from therapy.

The current results suggest that the tracts studied are important resources for deriving treatment-induced behavioral gains. The findings emphasize the value of white matter survival for achieving such gains, and so complement prior studies that focused on the role of gray matter structures to treatment-induced motor gains after stroke5, 15, 16. Tract-specific injury performed better than infarct volume did at predicting treatment gains, underscoring the importance of lesion location as a determinant of rehabilitation therapy effects. Tract-specific injury also predicted gains better than baseline behavior predicted gains did. A given behavioral phenotype can arise from many different brain states, but only some of these are likely to improve with therapy. The current study suggests that measures of tract-specific injury are useful for identifying treatment responders, beyond that provided by baseline behavior, and so might be useful for patient selection in clinical trials.

The current results are in general agreement with a prior study that found tract-specific injury to be a useful predictor of treatment gains17. In that study, Stinear et al17 found best prediction by measuring a combination of white matter integrity via DTI and neurophysiology via transcranial magnetic stimulation. Such multimodal assessment allows for more detailed classification of tract injury. However, the current approach also has its advantages, including assessment of injury to several specific tracts and simplicity of data analysis methodology. A prior study18 that used the tract/injury overlap approach to study gait found a limited relationship between tract injury and therapy-related gains. The reason for this disparity with current results might reflect fundamental differences in cerebral organization between hand movements (Table 3) and gait18, or that these authors examined corticospinal tract with only zero subdivisions, which was found to perform less well than with the current approach of using 16 subdivisions (Table 4, Supplement).

The constellation of findings (Table 3) provides some insight into the relative contributions that the four motor tracts make to motor gains during post-stroke motor therapy. Injury to the tract descending from PMv predicted treatment gains less often than injury to the other descending tracts did, suggesting a smaller role of this tract in supporting gains from a robotic therapy targeting the distal arm. Injury to the tracts descending from M1 and from PMd each had approximately equal predictive power, which was of some surprise given the predominance of M1 in issuing pyramidal and corticospinal fibers, though possibly related to the fact that the majority of fibers overlapped between these two tracts (Table 2). Together, the results do suggest that all four tracts play a role in mediating gains from motor therapy in chronic stroke, though their inter-relationships makes challenging a precise estimate of each tract’s role. The reason for the incomplete concordance of findings across the three behavioral scales, in particular the lack of prediction of ARAT score gains, is unclear but might reflect differences in extent to which content of each scale corresponded to content of the therapy. Further insight might be gained in future studies that use of assessments more aligned with the specific behaviors supported by each motor area, such as assessment of internal movement triggering for SMA and of external movement triggering for PMd19.

The current analysis did not find a threshold whereby a certain level of tract injury was clearly associated with total loss of distal arm motor function at baseline, or with inability to derive gains from therapy. For example, some subjects with >90% tract injury had good baseline function and good treatment gains, whereas some had poor baseline function/gains (Figure 1). This current finding regarding white matter tract injury contrasts with a prior study that examined motor cortex injury, which did find such a threshold; in that study, injury to >37% of the hand motor map was associated with total loss of hand motor function20. With severe injury to motor cortex, recovery does not occur, but with severe injury to the white matter tract descending from motor cortex, recovery occurs in some patients. Together, these findings suggest that behavioral effects of large white matter injuries might be reduced by forms of plasticity that are not possible with large cortical injuries.

Strengths of the study included that the findings were robust across varied methodological assumptions such as the threshold used to define injury. All infarcts were in left brain, eliminating variance due to stroke side. There were also limits with the current approach. Clearly, numerous factors besides tract injury, such as injury to other brain structures21, extent to which brain plasticity is exhausted at baseline22, or psychosocial factors23, influence behavioral response to therapy. The stroke masks obtained in the current chronic stroke cohort do not likely reflect the full nature of the acute injury; using the tract/injury overlap approach in the acute phase of stroke might find stronger relationships. The focus on left hemisphere injury does not provide specific information on right brain injury. The reason that the tract descending from M1 had lower volume than the tracts descending from PMd or SMA, despite contributing a larger proportion of axons to the corticospinal tract7, is unclear but might reflect tract fiber density or topography6, 24, or perhaps DTI-specific issues such as coherence. Finally, while measures of tract-specific injury correlated with change in behavior across therapy, these measures of injury showed a limited relationship with baseline behavior, in contrast to prior studies25-27. This is likely due to the pattern of deficits among enrollees--those with moderate-severe tract injury in the current study had a bimodal distribution, with some having mild-moderate and some having severe behavioral deficits (Figure 1a), and so the strength of the behavior/injury correlation varies with enrollee characteristics.

Many therapies are emerging that aim to improve motor function in people with stroke. Measuring the key biological substrate needed to achieve treatment gains will be useful to best match therapies with the right patients. The current report provides a robust and relatively simple method for measuring extent of injury to specific motor tracts, reinforces the importance of lesion location in stroke, and emphasizes the greater predictive value of tract-specific measures as compared to global measures of injury such as infarct volume.

Supplementary Material



Sources of Funding This study was supported by funds provided by the National Center of Research Resources, 5M011 RR-00827-29, US Public Health Service, and by R24 HD050821 and R01 NS059909.

Disclosures Steve Cramer has received grant support from GlaxoSmithKline and Stem Cell Therapeutics, lecture honoraria from Grupo Ferrer and Genentech, and has served as a paid consultant to GlaxoSmithKline, Phoththera, Allergan, Pfizer/Cogstate, and Johnson & Johnson. Lucy Der Yeghiaian and Jill See have served as paid consultants to Pfizer/Cogstate.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


1. Gresham G, Duncan P, Stason W, Adams H, Adelman A, Alexander D, Bishop D, Diller L, Donaldson N, Granger C, Holland A, Kelly-Hayes M, McDowell F, Myers L, Phipps M, Roth E, Siebens H, Tarvin G, Trombly C. Post-stroke rehabilitation. Rockville, MD: U.S Department of Health and Human Services Public Health Service, Agency for Health Care Policy and Research; 1995.
2. Rathore S, Hinn A, Cooper L, Tyroler H, Rosamond W. Characterization of incident stroke signs and symptoms: Findings from the atherosclerosis risk in communities study. Stroke. 2002;33:2718–2721. [PubMed]
3. Liepert J, Miltner W, Bauder H, Sommer M, Dettmers C, Taub E, Weiller C. Motor cortex plasticity during constraint-induced movement therapy in stroke patients. Neurosci Lett. 1998;250:5–8. [PubMed]
4. Sawaki L, Butler AJ, Xiaoyan L, Wassenaar PA, Mohammad YM, Blanton S, Sathian K, Nichols-Larsen DS, Wolf SL, Good DC, Wittenberg GF. Constraint-induced movement therapy results in increased motor map area in subjects 3 to 9 months after stroke. Neurorehabil Neural Repair. 2008;22:505–513. [PMC free article] [PubMed]
5. Takahashi CD, Der-Yeghiaian L, Le V, Motiwala RR, Cramer SC. Robot-based hand motor therapy after stroke. Brain. 2008;131:425–437. [PubMed]
6. Fries W, Danek A, Scheidtmann K, Hamburger C. Motor recovery following capsular stroke. Role of descending pathways from multiple motor areas. Brain. 1993;116:369–382. [PubMed]
7. Galea M, Darian-Smith I. Multiple corticospinal neuron populations in the macaque monkey are specified by their unique cortical origins, spinal terminations, and connections. Cereb Cortex. 1994;4:166–194. [PubMed]
8. Boudrias MH, McPherson RL, Frost SB, Cheney PD. Output properties and organization of the forelimb representation of motor areas on the lateral aspect of the hemisphere in rhesus macaques. Cereb Cortex. 2010;20:169–186. [PMC free article] [PubMed]
9. Donoghue J, Sanes J. Motor areas of the cerebral cortex. J Clin Neurophysiol. 1994;11:382–396. [PubMed]
10. Der-Yeghiaian L, Sharp K, See J, Abidi N, Mai K, Le V, Cramer S. Robotic therapy after stroke and the influence of baseline motor status. International Stroke Conference. 2009:e169.
11. Agrell B, Dehlin O. Comparison of six depression rating scales in geriatric stroke patients. Stroke. 1989;20:1190–1194. [PubMed]
12. Alexander MP, Baker E, Naeser MA, Kaplan E, Palumbo C. Neuropsychological and neuroanatomical dimensions of ideomotor apraxia. Brain. 1992;115(Pt 1):87–107. [PubMed]
13. Kunimatsu A, Itoh D, Nakata Y, Kunimatsu N, Aoki S, Masutani Y, Abe O, Yoshida M, Minami M, Ohtomo K. Utilization of diffusion tensor tractography in combination with spatial normalization to assess involvement of the corticospinal tract in capsular/pericapsular stroke: Feasibility and clinical implications. J Magn Reson Imaging. 2007;26:1399–1404. [PubMed]
14. Newton JM, Ward NS, Parker GJ, Deichmann R, Alexander DC, Friston KJ, Frackowiak RS. Non-invasive mapping of corticofugal fibres from multiple motor areas--relevance to stroke recovery. Brain. 2006;129:1844–1858. [PubMed]
15. Tardy J, Pariente J, Leger A, Dechaumont-Palacin S, Gerdelat A, Guiraud V, Conchou F, Albucher JF, Marque P, Franceries X, Cognard C, Rascol O, Chollet F, Loubinoux I. Methylphenidate modulates cerebral post-stroke reorganization. Neuroimage. 2006;33:913–922. [PubMed]
16. Gauthier LV, Taub E, Perkins C, Ortmann M, Mark VW, Uswatte G. Remodeling the brain: Plastic structural brain changes produced by different motor therapies after stroke. Stroke. 2008;39:1520–1525. [PMC free article] [PubMed]
17. Stinear CM, Barber PA, Smale PR, Coxon JP, Fleming MK, Byblow WD. Functional potential in chronic stroke patients depends on corticospinal tract integrity. Brain. 2007;130:170–180. [PubMed]
18. Dawes H, Enzinger C, Johansen-Berg H, Bogdanovic M, Guy C, Collett J, Izadi H, Stagg C, Wade D, Matthews PM. Walking performance and its recovery in chronic stroke in relation to extent of lesion overlap with the descending motor tract. Exp Brain Res. 2008;186:325–333. [PubMed]
19. Passingham R. The frontal lobes and voluntary action. Oxford: Oxford University Press; 1993.
20. Crafton K, Mark A, Cramer S. Improved understanding of cortical injury by incorporating measures of functional anatomy. Brain. 2003;126:1650–1659. [PubMed]
21. Lie C, Hirsch JG, Rossmanith C, Hennerici MG, Gass A. Clinicotopographical correlation of corticospinal tract stroke: A color-coded diffusion tensor imaging study. Stroke. 2004;35:86–92. [PubMed]
22. Schaechter JD, Perdue KL, Wang R. Structural damage to the corticospinal tract correlates with bilateral sensorimotor cortex reorganization in stroke patients. Neuroimage. 2008;39:1370–1382. [PMC free article] [PubMed]
23. Wolf SL. Revisiting constraint-induced movement therapy: Are we too smitten with the mitten? Is all nonuse “Learned”? And other quandaries. Phys Ther. 2007;87:1212–1223. [PubMed]
24. He S, Dum R, Strick P. Topographic organization of corticospinal projections from the frontal lobe: Motor areas on the lateral surface of the hemisphere. J Neurosci. 1993;13:952–980. [PubMed]
25. Thomalla G, Glauche V, Koch MA, Beaulieu C, Weiller C, Rother J. Diffusion tensor imaging detects early wallerian degeneration of the pyramidal tract after ischemic stroke. Neuroimage. 2004;22:1767–1774. [PubMed]
26. Lindenberg R, Renga V, Zhu LL, Betzler F, Alsop D, Schlaug G. Structural integrity of corticospinal motor fibers predicts motor impairment in chronic stroke. Neurology. 2010;74:280–287. [PMC free article] [PubMed]
27. Nelles M, Gieseke J, Flacke S, Lachenmayer L, Schild HH, Urbach H. Diffusion tensor pyramidal tractography in patients with anterior choroidal artery infarcts. AJNR Am J Neuroradiol. 2008;29:488–493. [PubMed]