Cognitive context determines premotor and prefrontal brain activity during hand movement in patients after stroke

Abstract

Background

Stroke patients often have difficulties in simultaneously performing a motor and cognitive task. Functional imaging studies have shown that movement of an affected hand after stroke is associated with increased activity in multiple cortical areas, particularly in the contralesional hemisphere. We hypothesised that patients for whom executing simple movements demands greater selective attention will show greater brain activity during movement.

Methods

8 chronic stroke patients with similarly localised lesions affecting the pyramidal tract performed a behavioural interference test employing a visuo-motor tracking with and without a simultaneous cognitive task. The magnitude of behavioural task decrement under cognitive-motor interference (CMI) conditions was calculated for each subject. Functional magnetic resonance imaging (FMRI) was used to assess brain activity in the same patients during performance of a visuo-motor tracking task alone, correlations between CMI score and movement-related brain activation were then explored.

Results

Movement-related activation in the dorsal precentral gyrus of the contralesional hemisphere correlated positively with CMI score (r² at peak voxel 0.92, p< 0.05). A similar relationship was observed in the ventral precentral gyrus (r² 0.52, p< 0.05) and middle frontal gyrus (r² 0.69, p< 0.05), these however were less robust. There was no independent relationship between hand motor impairment and CMI.

Conclusions

Results suggest that variations in the degree to which a cognitive task interferes with performance of a concurrent motor task explains a substantial proportion of the variations in movement-related brain activity in patients after stroke. The results emphasise the importance of considering cognitive context when interpreting brain activity patterns and provide a rationale for further evaluation of integrated cognitive and movement interventions for rehabilitation in stroke.

Manual dexterity involves grasping and applying muscular forces with the digits on objects ¹, a skill that is important for activities of daily living (ADL) ². Following a stroke a large proportion of patients will suffer from motor problems with the contralateral hand ³ which can affect their performance of ADLs. An added challenge for performing manual dexterity tasks in everyday life arises from the fact that functional tasks are rarely performed alone, but are often executed under distracting conditions that require attention to be divided between multiple tasks ⁴. Cognitive context influences performance even of simple movement tasks; decrements in performance of a manual dexterity task performed at the same time as a demanding cognitive task can be seen even in healthy subjects ¹, but such decrements are typically even greater following stroke ⁵. Such ‘cognitive-motor interference’ (CMI) can cause great difficulties in successful performance of ADLs and limit independence ⁶.

Performance of manual dexterity tasks is associated with altered patterns of motor cortical activation in stroke patients compared to age-matched controls^7-9, with increased activity of motor areas in the contralesional hemisphere most common in patients with the greatest impairment ^{10, 11}. It is possible that this reflects adaptive plasticity mediated by recruitment of intact motor output pathways from non-primary motor cortical areas or from the contralesional hemisphere ^{12, 13}. However, it is also possible that changes in patterns of activation arise from associated differences in cognitive demands of the task for patients and healthy controls. Motor imagery following stroke is associated with greater bilateral activation of primary and premotor cortical areas and with increased coupling between the ipsilesional (contralateral) prefrontal cortex (PFC) and premotor cortex (PMC) ¹⁴ even when individuals have intact motor areas. Differences in attention to movement modulate activity in the primary motor cortex and SMA in healthy subjects ^{15, 16}. Improved motor control of the hemiparetic hand has been reported to be associated with increased activation of the prefrontal areas involved with working memory and visuospatial transformations (Meehan, S, 2010). These results suggest that cognitive aspects of motor control are associated with modulation of activity across widespread motor cortical and association areas.

Given that behavioural evidence suggests that patients require greater attentional resources to execute even simple movements ^{5, 17, 18}, and that attention to movement is associated with modulation of motor cortical activity ^{15, 19}, we hypothesised that some of the brain activity associated with simple movement following stroke could be explained by enhanced attention to movement. To test this hypothesis, we calculated a behavioural measure of the attentional requirement of a simple hand movement using a cognitive-motor interference dual task paradigm. Specifically, stroke patients performed a visuo-motor tracking task with and without a simultaneous cognitive distractor task. We then acquired blood oxygen level dependent (BOLD) functional magnetic resonance imaging (fMRI) data in the same patients while they performed the visuo-motor task alone. We tested whether greater cognitive-motor interference in the behavioural assessment was associated with greater brain activity during simple motor task performance in specific brain regions.

MATERIALS AND METHODS

Patients

Eight individuals with chronic stroke took part in the study. All gave written consent prior to participation, in accordance with the declaration of Helsinki (1983). Individuals with stroke were identified by consultant referral with inclusion criteria of; right-handed (Edinburgh Handedness Miniscale −2 to 2); evidence of subcortical ischaemic infarct (greater than 5 mm in diameter) on CT or MRI (T1- or T2-weighted) scan with a localisation consistent with new motor symptoms and signs; grip strength (dynamometer) in impaired hand of >20% and <80% of that in unimpaired hand (based on the best performance in three repeated trials) with sensation to light touch and ability to give informed consent. Individuals were excluded if they had a clinical history of dementia or aphasia significantly limiting communication (NIH Stroke Scale rating >2), a history of previous symptomatic strokes or active neurological disease, a known active, major psychiatric disease or claustrophobia; any other conditions precluding safe MRI (e.g., pacemaker or other metal implant) or other condition having a functional impairment that would interfere with assessment of motor performance. Hand motor ability was assessed using the Fugl-Meyer (FM) motor assessment ²⁰., global cognitive ability was assessed using RBANS (Randloph). Demographic data can be found in Table 1.

Table 1.

Subject characteristics

Subject	Sex	TPO (months)	Age	FM score	CMI	CMI (norm)
1	M	10	50	64	0.72	0.15
2	M	23	56	50	0.17	−0.41
3	M	50	70	59	0.00	−0.58
4	M	9	44	64	0.01	−0.57
5	M	19	85	45	0.42	−0.15
6	M	31	70	51	2.77	2.2
7	M	42	71	61	−0.09	−0.66
8	M	30	62	42	0.61	0.03

M= male, TPO=time post stroke, FM, Fugl-meyer score (scale range from 0 to a max of 66), CMI=cognitive motor interference

Cognitive-motor interference behavioural testing

After a period of task familiarisation, the participants were asked to perform, in a random order, the following two tasks:

• A visuo-motor tracking task using the impaired (right) hand while seated: Participants viewed a computer screen displaying two moving bars: a computer-controlled target bar and a participant-controlled bar. The participant’s task was to match the height of the participant-controlled bar with the height of the target bar by altering grip force applied to a plastic isometric pressure-sensing device held in their right hand. An increase in applied pressure by the participant caused an increase in the participant-controlled bar height. Stimuli were presented and movements recorded using custom software developed on a Labview platform (National instrucments, Austin, TX) with a sampling rate of 100 Hz. Prior to beginning the task, the maximum amplitude of the target bar was calibrated to approximately 80% the individual’s maximum grip strength. During CMI trials the tracking sequence was a predictable repeated sine-wave to limit the required visuo-spaital processing requirements. Participants performed the task in 3 blocks of 60 seconds each.

• The ‘clock-faces’ cognitive task: A visuo-spatial decision task ¹⁷ in which participants heard a time of day such as “one fifteen”, every 5 seconds, and responded verbally whether the hands of the clock would be on the same or opposite side of the clock face. Participants performed the task in 3 blocks of 60 seconds.

Following single task performance of each task, subjects performed the two tasks simultaneously for a further three 60 second blocks. Both tasks were initiated simultaneously and subjects were instructed to give equal emphasis to the motor and cognitive task and thus perform both tasks to the best of their ability.

Functional Magnetic Resonance Imaging (fMRI) paradigm

The paradigm was a block design with 12 blocks of 30 seconds duration interspersed with 30 sec rest for each condition. During tracking blocks, participants performed a variant on the visuo-motor tracking task described above using the same handheld device, but instead of regular movements, the target moved in a pseudo-random, unpredictable fashion, (constrained so that the movements were always smooth) to maintain participants’ attention to the task throughout the scan session. During rest blocks, participants passively viewed both bars moving but made no responses. A third condition was included (in which a specific, more complex, sequence of movements was repeated), but data for that condition are not considered here. Vision was corrected using corrective lenses as necessary.

fMRI Data acquisition

Imaging data were acquired using a 3.0-T Varian (Siemens, Erlangen, Germany) MRI system. A total of 256 T2*-weighted blood-oxygen-level-dependent (BOLD) sensitive EPI images were acquired during task performance (40 axial slices per volume, TR=3500 ms, TE=30 ms, matrix=64 × 64, in-plane resolution=3 × 3 mm², flip angle=90°). A T₁-weighted anatomical image was also acquired (TR= 30 ms, TE= 5 ms, TI (inversion time) = 500 ms, flip angle 15°, FOV 256 × 256, matrix = 256 × 256).

Data Analysis

Behavioural data

Motor performance was evaluated by calculated the average root mean squared error (RMSE) of the difference between the target sequence pattern and the participants movement. The mean absolute tracking error for the middle 30sec of each 60sec block was calculated and averaged over the three blocks. This was to eliminate any asynchrony between the initiation and finish of the motor and cognitive task. Motor task data satisfied tests for normality therefore the difference in motor task performance from single to dual task conditions across the group was compared using a paired t-test. To quantify the interference effects of the cognitive task on the visuo-motor tracking task, a CMI index score was calculated for each participant as (error_single-error_dual)/ error_single. Each subject’s CMI score was normalised (giving unit variance and zero mean) and then analysed for relationship to hand motor ability (FM score) using 2-tailed Spearman’s correlation.

Individual level fMRI analysis

Data from each subject were initially analysed separately using tools from the FMRIB software library (www.fmrib.ox.ac.uk/fsl) ²¹. The following pre-statistical processing was applied: Motion correction using MCFLIRT ²²; non-brain removal using BET ²³), spatial smoothing using a Gaussian kernel of 5 mm full-width half maximum; mean based intensity normalisation; non-linear high-pass temporal filtering (Gaussian-weighted, least squares straight line fitting with sigma = 50.0 seconds). Individual data sets were checked visually to ensure no excessive head motion. Statistical analysis of the images was performed in two stages. The first level (within subject) analysis was carried out using FILM with local autocorrelation correction ²⁴ to produce effect size images of movement related activity versus rest for each subject that were carried up to the group level analysis.

Group level analysis

The first level results were registered into standard space (based on the Montral Neurological Institute (MNI) 152 template), via each subject’s T₁-weighted anatomical scan, using FNIRT (FMRIB’s non-linear image registration tool). Higher level analyses on the resulting aligned images used mixed effects with outlier de-weighting ²⁵ to calculate group average patterns of movement-related activity and to test for correlations between movement-related activity and CMI. As we were interested in correlations with CMI over and above any correlations with motor ability, Fugl-Meyer score was included in the model and shared variance between CMI and FM was assigned to the FM regressor by orthoganalising the CMI regressor with respect to the FM score. Group Z-statistic images were thresholded at Z > 2.3 with a (corrected) cluster significance threshold of P=0.05. Clusters showing a correlation with CMI were masked by the group average movement-related activity to identify movement-related areas where increased activity correlated with increased cognitive-motor interference. All images are shown in radiological convention in which the left side of the image is the right side of the brain.

RESULTS

Behavioural results

Analysis of error scores revealed a tendency for greater tracking error under dual-task conditions compared to single-task conditions (t=−1.86, p=0.053), although the difference did not reach significance. CMI scores quantifying the dual task effects were generally positive (reflecting greater error in the dual task condition) (range, −0.09 to 2.77) (Table 1). No significant relationships were identified between motor task error and FM score or between CMI score and FM score.

FMRI results

Movement related brain activity

The visuo-motor tracking activated the expected sensorimotor and visual network (Figure 1) including sensorimotor, posterior parietal, prefrontal and visual cortices and the cerebellum.

Figure 1.

Thresholded images (Z > 2.3) of group mean activation for movement versus rest. Significant clusters defined according to extent (at corrected P < 0.05).

Neural correlates of cognitive motor interference score

Correlations between increased FMRI signal change and CMI score were found in the right (contralesional) dorsal and ventral pre-motor cortices (PMd and PMv respectively) and in the right middle frontal gyrus (DLPFC) (Figure 2, Table 2).

Figure 2.

Influence of CMI score on brain activation during movement. Images show significant clusters on coronal, sagittal and axial slice views from left to right. Images have been thresholded at Z > 2.3 and significant clusters defined according to extent (at corrected P < 0.05). Colour bar indicates the voxel-wise z-statistic from red (Z = 2.3) to yellow (Z > 3.5).

Table 2.

Extent, magnitude and location of fMRI clusters for correlation between movement and CMI score

Cluster size	Max Z score	MNI co-ordinates of max Z score			Anatomical region
		x	y	z
267	3.62	40	36	34	R Middle frontal gyrus
273	3.50	28	−14	44	R Pre-central gyrus (dorsal)
292	3.24	60	4	22	R Pre-central gyrus (ventral)

Anatomical regions derived from the Harvard-Oxford cortical structural atlas. All clusters were significant at p<0.05, corrected.

To estimate how much of the variance in brain activity at each location can be explained by CMI, we calculated r² values between CMI scores and the percent signal change at the voxel of maximal correlation in each cluster. CMI score explained 52% (P<0.05), 92% (P<0.01) and 69% (P<0.05) of the variance in FMRI activity for the contralesional PMv, PMd and DLPFC, respectively. Consideration of individual subject data highlighted a participant with extreme values for both CMI and movement-related activity. Although the impact of this subject should be limited by the automatic outlier de-weighting included in our analysis ²⁵, to more fully assess the dependence of our results on this single subject, correlations at the peak voxels were re-assessed without the subject. There was still a positive relationship between CMI and activation in the dorsal pre-motor area (r²=0.45, P=0.05), although no meaningful relationship remained for the PMv or DLPFC.

DISCUSSION

We found an association between cognitive-motor inference and patterns of movement-related brain activity in chronic stroke patients. Patients who showed greater interference effects of a distracting cognitive task on motor performance had increased activation in specific premotor and dorsolateral prefrontal areas during performance of a motor task alone. These associations could not be explained by differences in motor impairment: no relationship was found between CMI and motor impairment (FM score) and the correlations between CMI and brain activity were over and above any relationships between FM and brain activity.

Previous functional imaging studies of simple hand movement following stroke have reported increased activity in contralesional areas, including premotor and prefrontal cortex ^{8, 11, 26-30}, but the functional role played by these areas is unclear. The current results suggest that increased activity in contralesional PMd and possibly also the PMv and DLPFC may in part reflect increased cognitive demands of even simple movements for patients ³¹. Performance of a concurrent cognitive task disrupts motor performance in patients to a greater extent than in healthy controls ^{5, 18} although evidence is mixed for upper limb movements ¹⁷. Functional imaging studies in healthy participants show that movement-related activity in premotor and prefrontal cortices depends on cognitive context: activity in motor and prefrontal areas increases with increased attention to simple movements ^{15, 16, 19}; PMd activity is associated with increased action selection demands ³², and DLPFC is involved in maintenance of task performance in the presence of a competing cognitive task ^{33, 34}. Findings from a study that compared performance on a learned repeated sequence visuo-motor tracking task to a random sequence performance demonstrated that during early learning of a motor task, individuals with stroke appear to rely upon increased DLPFC and PMd activation to perform the task regardless of sequence and state of learning (Meehan, 2010)

Our findings are consistent with the literature regarding the involvement of PMd in simple movement in stroke and thus underline the importance of considering cognitive demands when designing rehabilitation interventions. Movements are typically made in the context of multiple distracting task demands such as talking while walking. Our behavioural data show that patients with a relatively homogeneous clinical profile (FM scores ranging from 42 to 64) and stroke pathology (left hemisphere subcortical stroke) vary widely in the degree to which a distracting cognitive task impairs motor performance. In conjunction with previous data, our observation that variation in these distraction effects explains some of the variability in FMRI activity during a movement task alone suggests that the cognitive context of a movement is an important determinant of brain activity during motor task performance. If PMd, for example, is required even to perform a simple motor task then this will place limits on how well a patient can perform everyday activities in which additional activation demands are placed on the PMd by competing simultaneous cognitive tasks. Our results support a rationale for further development of rehabilitation strategies integrating cognitive and motor training. The potential effects of cognitive control training are suggested by a recent report that such an intervention can shift the pattern of brain activity associated with performance of a cognitive task from activation of a widespread network of brain regions including the lateral PFC to a more focal pattern relying on the lateral PFC ³⁵.

Nonetheless, there are several limitations to our study. Only a small number of patients were studied and all of the patients had left hemisphere subcortical stroke with mild to moderate impairments. Although use of a relatively homogeneous patient group allows for clearer interpretation of voxel-wise analysis of brain imaging data, it limits the degree to which our findings can be immediately generalised to the wider stroke population. One patient had a particularly high CMI score and, although the correlations found in PMd were robust to removal of this extreme subject, results were no longer significant in DLPFC or PMv; a larger follow-up study would help to clarify whether or not the DLPFC and PMv effects found here are meaningful. In addition, we did not study healthy controls so do not know whether the patients studied showed greater CMI effects than controls. However, previous studies have shown that CMI effects are typically greater in stroke patients compared to controls ^{18, 36}. We also did not assess behavioural performance of the cognitive task as we were primarily interested in changes in motor task performance but a more complete picture of interference effects could be gained from monitoring performance in both tasks. Finally, we were primarily interested in the dual task paradigm as a behavioural means to assess the attentional demands of simple movements and so dual tasks were not performed during fMRI scanning. Nevertheless, further research could directly contrast brain activity patterns during single and dual task performance to shed more light on the neural correlates of cognitive-motor interference in chronic stroke.

SUMMARY

In conclusion, our results suggest that variation in the degree to which a competing cognitive task interferes with performance of a concurrent motor task explains variation in movement-related brain activity, particularly in PMd, in patients after stroke. These associations are found over and above any associations with motor impairment and suggest that the cognitive demands of even simple movement execution should be considered when interpreting brain activity patterns in stroke patients. Additional work is needed to further evaluate the cognitive neuroscientific basis for integrated cognitive and movement therapy in rehabilitation of patients after stroke.