Movement preparation and execution: differential functional activation patterns after traumatic brain injury

Abstract

Years following the insult, patients with traumatic brain injury often experience persistent motor control problems, including bimanual coordination deficits. Previous studies revealed that such deficits are related to brain structural white and grey matter abnormalities. Here, we assessed, for the first time, cerebral functional activation patterns during bimanual movement preparation and performance in patients with traumatic brain injury, using functional magnetic resonance imaging. Eighteen patients with moderate-to-severe traumatic brain injury (10 females; aged 26.3 years, standard deviation = 5.2; age range: 18.4–34.6 years) and 26 healthy young adults (15 females; aged 23.6 years, standard deviation = 3.8; age range: 19.5–33 years) performed a complex bimanual tracking task, divided into a preparation (2 s) and execution (9 s) phase, and executed either in the presence or absence of augmented visual feedback. Performance on the bimanual tracking task, expressed as the average target error, was impaired for patients as compared to controls (P < 0.001) and for trials in the absence as compared to the presence of augmented visual feedback (P < 0.001). At the cerebral level, movement preparation was characterized by reduced neural activation in the patient group relative to the control group in frontal (bilateral superior frontal gyrus, right dorsolateral prefrontal cortex), parietal (left inferior parietal lobe) and occipital (right striate and extrastriate visual cortex) areas (P’s < 0.05). During the execution phase, however, the opposite pattern emerged, i.e. traumatic brain injury patients showed enhanced activations compared with controls in frontal (left dorsolateral prefrontal cortex, left lateral anterior prefrontal cortex, and left orbitofrontal cortex), parietal (bilateral inferior parietal lobe, bilateral superior parietal lobe, right precuneus, right primary somatosensory cortex), occipital (right striate and extrastriate visual cortices), and subcortical (left cerebellum crus II) areas (P’s < 0.05). Moreover, a significant interaction effect between Feedback Condition and Group in the primary motor area (bilaterally) (P < 0.001), the cerebellum (left) (P < 0.001) and caudate (left) (P < 0.05), revealed that controls showed less overlap of activation patterns accompanying the two feedback conditions than patients with traumatic brain injury (i.e. decreased neural differentiation). In sum, our findings point towards poorer predictive control in traumatic brain injury patients in comparison to controls. Moreover, irrespective of the feedback condition, overactivations were observed in traumatically brain injured patients during movement execution, pointing to more controlled processing of motor task performance.

Bimanual coordination deficits may persist for years after traumatic brain injury (TBI). Gooijers et al. compare functional activation patterns during the preparation and execution of a complex bimanual movement in patients with TBI versus controls. Those with TBI show reduced activity during movement preparation and increased activity during movement execution.

fMRI, traumatic brain injury, bimanual motor control, movement preparation, visual feedback

Introduction

Traumatic brain injury (TBI) can result in diverse short- and/or long-term deficits in sensorimotor, cognitive and behavioural functions (Irimia and Van Horn, 2015). Although the degree of recovery varies across individual patients, patients with moderate-to-severe TBI often demonstrate persisting deficits for years following injury (Bales et al., 2009). With respect to motor-related problems, slower reaction times, decreased upper-limb motor speed, dexterity and coordination as well as postural deficits have repeatedly been reported (Chaplin et al., 1993; Gray et al., 1998; Kuhtz-Buschbeck et al., 2003a, b; Caeyenberghs et al., 2009b, 2010b). Evidently, such persistent motor deficits have been associated with post-injury neural alterations. Particularly, TBI-induced white matter microstructural changes, as measured by diffusion tensor imaging, correlate with a great variety of TBI-induced motor deficits (Ewing-Cobbs et al., 2008; Caeyenberghs et al., 2010a, b, 2011a, b, 2012; Jang, 2011; Choi et al., 2012; Shin et al., 2012; Moen et al., 2014). Moreover, TBI-induced grey matter volume alterations in subcortical structures correlate with bimanual impairments (Gooijers et al., 2016). Neuroimaging studies in TBI patients, exploring functional activations during motor performance, however, remain scarce. The few existing studies tend to generally agree on the occurrence of TBI-induced cerebral functional activation changes during task performance (Prigatano et al., 2004; Lotze et al., 2006; Wiese et al., 2006; Mani et al., 2007; Rasmussen et al., 2008; Caeyenberghs et al., 2009b; Lima et al., 2011; Sinopoli et al., 2014). According to Sinopoli et al. (2014), such alterations can be accounted for by (i) permanent rewiring of the brain (i.e. reorganization); (ii) compensatory activation (i.e. additional recruitment of brain areas to facilitate performance); or (iii) poorer cognitive efficiency (i.e. need for increased cognitive control). In many instances, TBI patients as compared to controls, are in need of increased attentional guidance, especially during the performance of more complex (or dual) tasks (Ghajar and Ivry, 2008; Rasmussen et al., 2008; Sinopoli et al., 2014). Additionally, as suggested by previous EEG research, patients with TBI might experience movement preparation deficits. For example, altered readiness potentials during self-initiated movements (Wiese et al., 2004; Di Russo et al., 2005), and a reduced contingent negative variation during a cued button press (Segalowitz et al., 1997), have been reported in TBI patients relative to controls. However, it is worth noting that these studies focused on simple finger movement tasks with limited preparation processes. With regard to imaging evidence, only one study so far claimed the presence of altered preparatory networks post-injury, based on increased activation maps in the premotor cortex during voluntary (unilateral) finger abductions (Wiese et al., 2006). Nevertheless, only patients with focal prefrontal cortex contusions were considered in the latter study.

The aims of the present study were 3-fold. The first aim was to examine the functional neural correlates of movement preparation and execution in healthy young adults and chronic TBI patients, demonstrating widespread diffuse and/or focal lesions. The second aim was to investigate in more detail the effect of TBI on higher order movement planning processes required to perform a complex bimanual coordination task. Finally, the third aim was to compare the effect of internal (naturally available information from sensory systems during movement) versus external (augmented feedback supporting performance) movement guidance on behaviour and neural activations as well as the degree of neural differentiation between these conditions in TBI patients versus controls.

Although some have suggested that internally-guided movements (without augmented feedback), may be more vulnerable to disruption in TBI (Heitger et al., 2009), behavioural and imaging evidence is lacking. A recent study including healthy participants reported that internally—as compared to externally—guided movements, resulted in greater activations in somatosensory areas, cerebellum and the lingual gyrus. On the contrary, frontal and occipital regions showed greater activations during externally- than internally-guided movements (Beets et al., 2015).

The following hypotheses were proposed: (i) behaviourally, we expected that the patient compared to the control group would show deficits in bimanual performance, and that performance with augmented feedback (FB) would exceed the performance without augmented feedback (NFB), with a more pronounced effect in the TBI population; and (ii) at the neural level, TBI patients as compared to controls, were expected to demonstrate increased activations during movement execution, while failing to activate the appropriate brain areas during movement preparation. Moreover, given the TBI-induced expected general pattern of neural over-recruitment during task execution, less differentiation was hypothesized to occur between the activation patterns accompanying both feedback conditions in the patient relative to the control group.

Materials and methods

Subjects

Twenty-six right-handed (Oldfield, 1971) young adults [15 females, aged 23.6, standard deviation (SD) = 3.8 years, range 19.5–33 years of age], and 18 right-handed TBI patients (10 females, aged 26.3, SD = 5.2 years, range 18.4–34.6 years of age) with normal or corrected-to-normal vision were recruited. The majority of the healthy controls were included in another study that examined the cerebral activation changes after bimanual coordination training (Beets et al., 2015), whereas a small part of the TBI sample was included in studies regarding executive functioning (Caeyenberghs et al., 2013, 2014; Leunissen et al., 2014a, b). The groups did not differ significantly with respect to age [t(42) = 1.98, P = 0.054] or gender [χ²(1) = 0.02, P = 0.888]. Table 1 lists the clinical characteristics and demographics of the TBI group. All patients were in the chronic stage of TBI [at least 6 months post-injury; which can be considered chronic according to, for example, Kraus et al. (2007); on average 4.7, SD = 3.5 years post-injury]. Their age at injury was on average 21.6, SD = 7.3 years.

Table 1

Summary of demographic and clinical characteristics of the TBI sample.

Gender/age (years)	Age at injury (years)	Time since injury (years)	Cause of trauma	Focal lesion (lesion size in cm3) a	LOC, GCS, Coma, PTA, RTA	Acute scan within 24 h after injury	MRI scan at examination/ location of injury
Female/24.8	21.3	3.5	Traffic accident	-	Coma: 3 weeks, PTA: 1 week, RTA: 1 day, GCS: 7–8	Right frontal lobe haemorrhage, left frontal/temporal/ parietal lobe, right orbito-frontal lobe contusion	Drain tract right, haemosiderin deposits right parietal lobe and right orbito-frontal cortex
Female/18.4	8.2	10.2	Fall	-	LOC, GCS: 8/9, with coma, 2 days artificial respiration	Right frontal lobe subdural haematoma, frontal oedema	-
Female/28.9	23.9	5.0	Traffic accident	-	Not available	Frontal lobe injuries	Drain tract left, parietal and occipital lobe and frontal lobe and right temporal lobe shearing injuries, slightly enlarged ventricles
Male/19.2	14.5	4.7	Fall	-	LOC: temporary, (artificial) coma: 3 weeks	Enlarged right lateral ventricle, right haematoma occipital horn lateral ventricle, hyperdensity left thalamus, parietal and temporal lobe, left hemisphere shearing injuries	Drain tract right, left corpus callosum, thalamus, frontal lobe, right parietal and temporal lobe shearing injuries, occipital horn lateral ventricle asymmetrically enlarged
Female/29.8	25.5	4.3	Work accident	+ (11.36)	(artificial) Coma: 3 months, PTA: 4 or 5 weeks, RTA: 1 month	Bilateral frontal/temporal lobe injuries, right frontobasal injuries, limited subarachnoid haemorrhage parietal/occipital lobe, diffuse axonal injury	Wide sulci, global cerebral and cerebellar atrophy, haemosiderin deposits corpus callosum, left superior frontal lobe and periventricular, parenchyma loss right orbito-frontal lobe and right anterior temporal pole, drain tract right
Male/24.4	19.1	5.3	Traffic accident	-	Coma: 4 days, PTA: several days, RTA: 24 hours	Contusion and diffuse axonal injury (location not specified in available records)	Left thalamus and left temporal lobe and left orbito-frontal cortex and bilateral frontal lobe and central sulcus shearing injuries
Male/26.2	21.8	4.4	Traffic accident	+ (9.71)	RTA, coma and PTA: multiple weeks	Left frontal lobe haematoma, frontal lobe parenchymal bleeding, subarachnoidal bleeding	Drain tract right, orbito-frontal cortex and right cerebellum contusion
Male/20.4	19.5	0.9	Fall	+ (43.82)	Not available	Right frontal lobe haemorrhage, left frontal lobe, bilateral temporal lobe limited parenchymal bleeding, narrowed horns of right lateral ventricle, midline shift, skull fracture left occipital lobe, extracranial haematoma left occipital lobe	Wide extended parenchyma loss right orbito-frontal lobe and anterior frontal, drain tract left frontal lobe, haemosiderosis right (result of subdural haematoma), haemosiderin deposits left cerebellar peduncle, anterior cingulate gyrus and post central gyrus (hand knob)
Male/19.6	9	10.6	Traffic accident	-	GCS: 8	Left temporal/frontal lobe punctiform, right mesencephalon contusion, left frontal lobe, left thalamus haemorrhagic injuries	Orbito-frontal cortex contusion, enlarged ventricles
Male/29.2	28.4	0.8	Violence	+ (13.17)	(artificial) Coma: 2.5 weeks	-	Bleeding right pallidum, left temporal lobe contusion, drain tract left, haemosiderin deposits (bilateral) (result of subdural haematoma) → indication of diffuse axonal injury, small shaped parenchyma loss left medial frontal gyrus
Male/34.6	29.6	5.0	Traffic accident	+ (31.87)	Coma + PTA: 2 weeks, RTA: ≥ 1 month	Left fronto-temporal haemorrhage with diffuse axonal injury. Shearing injuries in occipital/ temporal lobe	Right orbitofrontal, inferior frontal lobe contusion, corpus callosum degeneration, left parietal lobe shearing injuries
Female/30.9	18.6	12.3	Traffic accident	+ (3.35)	Coma: 1 week, PTA: 7 or 8 weeks	Diffuse haemorrhagic contusion	Global atrophy (left > right), contusion zone: left temporal lobe, drain tract right, atrophy in parietal corpus callosum (Wallerian degeneration), widespread haemosiderin deposits (left > right), right parietal lobe injury
Female/23.1	15.8	7.3	Traffic accident	+ (14.25)	Coma + PTA: 1 month, RTA: 2.5 months	Left parietal/occipital lobe contusion, subdural haematoma	Left parietal lobe contusion, corpus callosum and occipital lobe shearing injuries, occipital horn lateral ventricle asymmetrically enlarged
Female/33.7	33	0.7	Traffic accident	-	(artificial) Coma: 2 days	-	Haemosiderin deposits superior frontal gyrus right
Male/29.3	28.5	0.8	Work accident	-	Short LOC	Diffuse axonal injury	Haemosiderin deposits inferior frontal gyrus right, superior temporal gyrus right, superior frontal gyrus left, pericallosal (bilateral) → indication of widespread diffuse axonal injury, bleed over tentorium cerebellum, old blood in sulci
Female/31.9	31.3	0.6	Traffic accident	-	Not available	-	Rest bleed on tentorium cerebellum, haemosiderin deposits (bilateral) pericallosal and body caudate nucleus → indication of diffuse axonal injury
Female/20.3	15.6	4.8	Traffic accident	-	(artificial) Coma: 5 days	Increased epidural haematoma temporal/ frontal lobe, subdural haemorrhage frontobasal, contusion temporal lobe, fracture line fronto-temporal	Injury from drain tract right extending into striatum (head + body caudate nucleus) > enlarged ventricle, haemosiderin deposits insular cortex right
Female/28.9	25	3.9	Accident	-	Not available	-	Slightly enlarged ventricles, rest bleed on cerebellum, haemosiderin deposits pericallosal right

GCS = Glasgow Coma Scale; LOC = loss of consciousness; PTA = post traumatic amnesia; RTA = retrograde amnesia.

^a − = no focal lesion, + = focal lesion. In case of +, lesion size (cm³) was calculated using the fslstats –V function on binarized lesion masks in subject space.

All TBI patients had sustained moderate-to-severe head injury, as defined by the Mayo classification system, which defines patients according to the duration of loss of consciousness, period of post-traumatic amnesia, minimum Glasgow Coma Scale during the first 24 h post injury and MRI or CT images (Malec et al., 2007), assessed by a specialized neuroradiologist. Lesions were typically located in the frontal and temporal regions (Supplementary Fig. 1). Overlaying the focal lesions onto a mask consisting of bilateral primary motor cortices (Harvard-Oxford-cort-maxprob-thr0-2 mm), revealed that, except for one patient, there was no primary motor involvement.

Participants were excluded from the study in case of use of psychoactive medication, history of drug abuse, psychiatric or neurological disease other than TBI, significant multiple trauma, an abbreviated injury score >2 for the upper limbs indicating serious impaired arm function, or any medical history that makes functional MRI impossible. Additionally, patients demonstrating spasticity or biomechanical limitations of the hands or arm that may have interfered with bimanual task performance were excluded. To assess the functionality of bimanual movements in our patient sample, we administered the Purdue Pegboard Test (bimanually place as many metal pins as possible in two parallel rows of holes running vertically down the board, within 30 s; Desrosiers et al., 1995), the TEMPA (Test d’Evaluation de la performance des Membres supérieurs des Personnes Agées; a standardized test for daily life upper limb functions; Desrosiers et al., 1993) and the ABILHAND (questionnaire to assess the functionality of bimanual movements; Penta et al., 1998). Based on their scores on the TEMPA (all patients completed the tasks successfully and without hesitation or difficulty), and the ABILHAND (according to the patients themselves, 98% of the proposed tasks could be performed easily) all TBI patients could perform bimanual daily activities without serious problems. Results on the Purdue Pegboard Test, however, revealed a significant slowing of performance [t(16) = −6.8, P = 0.000 (mean TBI: 11, normal value: 14)]. All participants were fully informed concerning the study and signed a written informed consent prior to participation in accordance with the Declaration of Helsinki. The study was approved by the local ethics committee.

Task description

A bimanual tracking task was used, as previously introduced by our group (Sisti et al., 2011, 2012; Gooijers et al., 2013; Beets et al., 2015). Participants were asked to manipulate two dials (left and right hand) using their thumb and index finger (Fig. 1A). The right dial controlled movement along the horizontal axis; when turned clockwise the cursor moved to the right, when turned counter clockwise the cursor moved to the left. For the left dial, which controlled movement along the vertical axis, clockwise rotation resulted in upward movement, and counter clockwise rotation led to downward movement. Four possible movement patterns were included; rotate the dials towards each other, away from each other, clockwise or counter clockwise. A typical trial started with a movement preparation phase (2000 ms) in which a blue target line was displayed on the screen. Each line slope represented a different movement pattern. No movement was required yet. This movement preparation phase was immediately followed by the execution phase (signalled with an auditory cue) which lasted 9000 ms. During execution, the white-coloured target dot moved along the blue target line. The position of the white target dot had to be matched in space and time as accurately as possible by simultaneous cyclical rotation of the dials. During feedback trials (FB) online visual information was provided by means of a red cursor displaying the actual tracking performance of the subject. In case of a no feedback trial (NFB), no augmented visual feedback was available, i.e. the red cursor was removed. This implies that proprioceptive input was prominent during the NFB condition whereas augmented visual FB was present in association with proprioceptive information during the FB condition. The upcoming condition (FB or NFB) was indicated during the movement preparation phase (Fig. 1C).

Figure 1

Bimanual Tracking Task. (A) Bimanual tracking task setup inside the magnetic resonance scanner. (B) Schematic overview of 20 possible target pathways, including four coordination patterns (CW = clockwise, CCW = counterclockwise, IN = inwards and OUT = outwards) and five frequency ratios (1:1, 1:2, 1:3, 2:1, 3:1). Each line slope represents a different coordination pattern. (C) A typical trial of the bimanual tracking task starts with a display of a blue target line together with a cue indicating whether augmented feedback would be available during the trial (FB) or not (NFB; FB crossed out). The cue is either yellow or pink, indicating whether the upcoming trial was a ‘move’ or a ‘no-move’ trial (colour of the cue was counterbalanced across participants). An auditory cue, presented at 2 s (movement preparation phase), indicates the start of movement. At the same time a white target dot starts to move along the target line from the middle of the screen to the end of the blue target line at a constant speed (duration 9 s; execution phase). Participants are required to rotate the dials to trace the white target dot. In FB trials, the actual tracking trajectory of the subject is displayed by means of a red cursor, which is absent in NFB trials.

Participants were required to rotate the dials according to five different inter-hand frequency ratios; 1:1, 1:2, 1:3, 2:1, 3:1 (Fig. 1B). There was an intertrial interval of 3000 ms.

Experimental design

During a first familiarization session, taking place in the Motor Control Laboratory of the KU Leuven, all subjects practiced the bimanual tracking task while lying supine in a dummy scanner. At the end of this session, all subjects were confident that they could perform all task conditions adequately. Within 1 week a functional MRI scanning session was scheduled at the University Hospital Gasthuisberg, Leuven. During the scanning session, participants lay supine with both upper arms supported by cushions. A custom-made functional MRI compatible set-up consisting of two dials (i.e. one for each hand; see Fig. 1A), was placed over the legs at a comfortable distance. The positioning of both dials was adapted for each person to ensure easy handling. Non-ferromagnetic high precision optical shaft encoders (HP, 2048 pulses per revolution, sampling frequency 100 Hz), attached to the movement axes of the dials, were used to register angular displacements. Foam cushions around the head and a bite-bar were used to limit head movements while performing the task. The task and the visual feedback were displayed via a video projector (Barco 6300, 1280 × 1024 pixels), and viewed via a double mirror, attached to the head coil.

The functional MRI experiment consisted of six scanning runs, each consisting of 24 trials containing four different scanning conditions. The four scanning conditions were: (i) performing the bimanual tracking task with augmented visual feedback (move FB; 48 trials); (ii) performing the bimanual tracking task without augmented visual feedback (move NFB; 48 trials); (iii) passively viewing the display presenting the bimanual tracking task with augmented feedback (example of a trial), not requiring active movement (no-move FB; 24 trials); (iv) passively viewing the display representing the bimanual tracking task without augmented visual feedback, not requiring active movement (no-move NFB; 24 trials). The ‘no-move’ trials were included to measure the baseline blood oxygen level-dependent response. The cue, presented at the start of the movement preparation phase, indicated which of these four conditions applied to the upcoming trial (Fig. 1C). The division of FB/NFB trials in the ‘move’ and ‘no-move’ scanning conditions was 50/50. The required frequency ratio was semi-randomly distributed across the experiment, such that one-third of the trials was performed according to a 1:1 ratio, one-third according to a 1:2/2:1 ratio, and one-third according to a 1:3/3:1 ratio. Computer programming for this task was done using Laboratory Virtual Instrumentation Engineering Workbench, version 8.5 (National Instruments).

Image acquisition

MRI data were obtained on a Siemens 3 T Magnetom Trio MRI scanner using a 12-channel head coil. First, a high-resolution T₁-weighted structural image, lasting 8 min, was acquired using magnetization prepared rapid gradient echo (repetition time/echo time = 2300/2.98 ms, 1 × 1 × 1.1 mm voxels, field of view: 240 × 256, 160 sagittal slices). Second, a fieldmap with an effective echoplanar imaging echo spacing of 0.71 ms and an echo time of 30 ms was obtained for B0 unwarping. Subsequently, functional images were acquired with a descending gradient echoplanar imaging pulse sequence for T₂*-weighted functional images (repetition time/echo time = 3000/30 ms, flip angle = 90°, 50 oblique axial slices, slice thickness = 2.8 mm, interslice gap = 0.28 mm, in-plane resolution = 2.5 × 2.5 mm, 80 × 80 matrix, 116 volumes). To ensure T₁ equilibration, the first three volumes from each subject’s run were removed.

Data analysis

Kinematic analysis

The data of the bimanual tracking task were analysed using both MATLAB R2008a and Microsoft Excel 2007. On each trial, the x and y positions of the target and the cursor were sampled at 100 Hz. Accuracy of bimanual performance, indicated as the average target error, was measured by calculating the Euclidian distance between the position of the target and the cursor at each time point and then averaged per trial. Accordingly, a lower score indicates a better performance. No-move trials were discarded from the functional MRI analyses in case the participant did turn the dials for more than one movement cycle (1.3% for control subjects and 2.4% for TBI patients). To correct for the non-normal distribution of the behavioural data (Shapiro-Wilk test: P < 0.05), a log transformation was performed. After the log transformation 9 of 10 variables [five (Frequency ratios) × two (Feedback condition)] were normally distributed for the control group, and 7 of 10 for the TBI patients. This was an improvement relative to no transformation.

Statistical analyses

For average target error, a 2 group (TBI, controls) × 2 feedback condition (FB, NFB) × 5 frequency ratio (1:1, 1:2, 1:3, 2:1, 3:1) repeated measures ANOVA was performed on the data obtained during scanning. The directional patterns (Clockwise, Counter clockwise, Inwards, Outwards) were collapsed as these were not of interest for the present analyses. Tukey post hoc tests were used to further explore the significant main and interaction effects. All statistical analyses were performed using Statistica 12, Statsoft, Tulsa, Oklahoma, and thresholded by P < 0.05. All tests were two-tailed.

Imaging analysis

The functional MRI of the Brain Software Library (v 5.0.5) was used to preprocess and statistically analyse the imaging data. Preprocessing consisted of the following steps: removing non-brain tissue using the Brain Extraction Tool, realignment of the echoplanar images using linear motion corrections (MCFLIRT), B0 unwarping using a fieldmap, and slice time correction was applied together with spatial smoothing using a full-width at half-maximum of 5 mm. One patient and one control were excluded from further analyses because of an incorrect field of view and excessive motion, respectively. Brain registration was performed according to a two-stage process: firstly, echoplanar images were coregistered to the structural T₁ image using a linear transformation with six degrees of freedom and secondly registered to the Montreal Neurological Institute brain image using a non-linear affine transformation with 12 degrees of freedom (FNIRT). For a small number of patients (n = 3) the non-linear registration did not work properly, and was replaced by a linear transformation (FLIRT). In addition, cost-function masking was applied to patients with a focal lesion (n = 7; Table 1) to refine registration and avoid problems of stretching normal brain tissue (Brett et al., 2001). Lesion masks were manually drawn in FSLView 3.2.0 on each slice of the T₁-weighted structural image. Next, the lesion mask was binarized and inverted (resulting in zeros in the lesion and ones outside the lesion), and used as an input-weighting mask when the structural T₁ image was coregistered with the Montreal Neurological Institute (MNI) template. The output of the last step was then used to achieve functional to standard registration (Brownsett et al., 2014). For the statistical analyses, the functional MRI data were analysed in serial steps accounting for fixed and random effects, respectively. The analyses used a general linear model (GLM) that included, for each run, responses to the movement preparation (2000 ms) and execution (9000 ms) as well as their temporal derivatives for the following conditions: move FB, move NFB, no-move FB, and no-move NFB. For each trial type, a given item was modelled as an event representing the trial onset. The ensuing vector was convolved with the canonical haemodynamic response function, and used as a regressor in the individual design matrix. Discarded no-move trials were added to the model as regressors of no interest. Furthermore, movement parameters derived from realignment of the functional volumes were also included as covariates of no interest. Finally, motion outliers, created during the preprocessing using Fsl_motion_outlier, were also entered into the model, for each run, as an additional confounding experimental variable. This was done to remove the effects of time points that were corrupted by large motion and/or distortions, without adverse effects on further statistical analyses. The threshold was set to the 75th percentile + 1.5 times the interquartile range (used when creating boxplots). The no move trials were used as baseline and were therefore subtracted from all the contrasts listed below. In particular, contrasts tested, across runs, referred to the main effect of movement condition (move versus no move) irrespective of the feedback condition (FB + NFB trials) as well as the main of effect of feedback (FB trials versus NFB trials) separately for movement preparation and execution. Movement preparation and execution were never statistically compared since their onsets were dependent. The resulting contrast images were then entered into a higher level analysis (random effects model) to study the main effect of group with a two-sample t-test. A supplemental one-sample t-test was performed to investigate the main effect of feedback across groups. On this level we also added a grey matter mask, using feat_gm_prepare in FSL, to add structural grey matter data as a confound covariate. For all functional MRI analyses corrections for multiple comparisons were performed using Gaussian Random Field-based cluster inference with a threshold of Z < 2.3 and a cluster probability threshold of P < 0.05. Together with the activation peak per cluster, local maxima will be reported. Moreover, as we were only interested in activations in grey matter, we used a grey matter template and included all voxels with a grey matter intensity level above 0.3. For visualization purposes and to calculate regression coefficients, per cent signal changes were determined using fslmaths and fslmeants (Mumford, 2014). Finally, to investigate the relationship between functional activation changes and behavioural performance, average target error scores (averaged across feedback conditions) were correlated (Pearson) with per cent signal changes in regions of interest across and within groups (two-tailed tests).

Results

Kinematics

Bimanual motor performance (average target error) was assessed using a 2 group (TBI, Controls) × 2 feedback condition (FB, NFB) × 5 frequency ratio (1:1, 1:2, 1:3, 2:1, 3:1) mixed repeated measures ANOVA. First, a significant main effect of Group was observed [F(1,42) = 16.30, P < 0.001], reflecting higher error scores in the TBI group as compared to the controls. Second, a poorer performance in the absence of augmented visual feedback was reflected by a significant main effect of Feedback Condition [F(1,42) = 200.56, P < 0.001]. Furthermore, a main effect of Frequency Ratio was present [F(4,168) = 32.77, P < 0.001]. Post hoc Tukey analyses indicated that all non-1:1 frequency ratios were more difficult to perform than the 1:1 frequency ratio (all P < 0.005) and that 1:2 was significantly easier than 3:1 (P < 0.005). In addition, a significant interaction effect between feedback Condition and frequency ratio was found [F(4,168) = 3.39, P = 0.011], revealing that in the presence of augmented visual feedback, ratio 1:1 is significantly easier than all non-1:1 ratios, whereas in the absence of augmented visual feedback, also ratio 1:2 is easier than ratios 1:3 (P = 0.002), 2:1 (P = 0.037) and 3:1 (P < 0.001). The Group × Feedback Condition [F(1,42) = 0.97, P > 0.05], the Group × Frequency Ratio [F(4,168) = 1.84, P > 0.05], and the three-way [F(4,168) = 1.08, P > 0.05] interactions did not reach significance. Behavioural results are displayed in Fig. 2.

Figure 2

Behavioural results. Average target error for each frequency ratio (3:1, 2:1, 1:1, 1:2, 1:3). We refer to the left hand first and the right hand second (L:R). Left: Average target error during FB trials. Right: Average target error during NFB trials. Black squares with full line represent healthy controls, white circles with dashed line represent TBI patients. Non-log transformed data are presented. The error bars indicate the standard error of the mean. Images were created using Microsoft Excel 2007.

Functional MRI

Between Group effects

During the movement preparation phase, reduced activations for TBI patients compared with controls were found in the (i) right superior frontal gyrus (with extensions into left superior frontal gyrus, and right dorsolateral prefrontal cortex); (ii) right primary visual cortex [also including extrastriate visual areas (bilateral V2)]; and (iii) left inferior parietal lobe (Table 2 and Fig. 3). There were no regions with higher activation levels in TBI compared to controls.

Figure 3

Mean functional activations during movement preparation and execution. TBI patients are depicted in red, healthy controls in green, and common activations are depicted in yellow. Top: Activations for the following contrast: preparation move > preparation no-move (averaged over FB and NFB trials). The magenta colour indicates the brain regions showing higher activation levels in the control group relative to the TBI group during movement preparation. Bottom: Activations for the following contrast: execution move > execution no-move (i.e. passively viewing a video of the task on the screen; averaged over FB and NFB trials). The blue colour indicates the brain regions showing enhanced activity in the TBI group as compared with the control group during movement execution. Activation maps are overlaid on the ch2better template using MRIcron. Activations are significant, using a clusterwise threshold Z > 2.3, P < 0.05.

Table 2

Main effect of Group

Brain region	Side	Activation coordinates			Z-value	P-value	Number of voxels in cluster
		x	y	z
TBI > Controls
Preparation
–	–	–	–	–	–	–
Execution
Dorsolateral prefrontal cortex	L	−50	26	38	4.63	0.000249	1213
Lateral anterior prefrontal cortex	L	−34	60	18	3.98
Orbitofrontal cortex	L	−48	46	0	3.68
Inferior parietal lobe	L	−28	−86	42	3.9	0.00488	801
Superior parietal lobe	L	−8	−66	68	3.45
V1	R	16	−102	12	4.13	0.00683	758
V2	R	28	−98	16	4
V4	R	42	−82	−14	3.6
Superior parietal lobe	R	22	−68	64	3.8	0.00961	715
Inferior parietal lobe	R	50	−76	28	3.6
Precuneus	R	10	−52	46	3.38
Inferior parietal lobe	R	62	−38	28	3.73	0.0149	661
Primary somatosensory cortex	R	48	−26	44	3.31
Cerebellum crus II	L	−30	−88	−40	4.3	0.0397	545
Controls > TBI
Preparation
Superior frontal gyrus	R	20	36	38	4.52	<0.00001	6255
Dorsolateral prefrontal cortex	R	40	12	48	4.31
Superior frontal gyrus	L	−18	46	40	4.3
V1	R	8	−92	10	4.31	0.000125	1546
V2	R	22	−90	16	3.98
V2	L	−2	−86	28	3.63
Inferior parietal lobe	L	−44	−60	38	3.92	0.00749	868
Execution
–	–	–	–	–	–	–

Peak activations are presented in bold type. L = left; R = right.

During the execution phase, no regions were significantly more activated in the control group compared to the TBI patients. However, the TBI greater than control contrast revealed higher levels of activation in the (i) left dorsolateral prefrontal cortex (with local maxima in left lateral anterior prefrontal cortex and left orbitofrontal cortex); (ii) left inferior parietal lobe (also including left superior parietal lobe); (iii) right primary visual cortex [also including extrastriate visual areas (right V2, V4)]; (iv) right superior parietal lobe (also including right inferior parietal lobe and right precuneus); (v) right inferior parietal lobe (extending into right primary somatosensory cortex); and (vi) left cerebellum crus II (Table 2 and Fig. 3).

Effect of feedback condition

In the movement preparation phase (across groups), the FB > NFB contrast indicated significant activations in three main clusters (Table 3) including (i) the right precuneus (peak voxel), extending into (extra)striate visual areas (right V1, V3 and left V2); (ii) right medial anterior prefrontal cortex (peak voxel), also including bilateral dorso- and ventral medial prefrontal cortex and left medial anterior prefrontal cortex; and (iii) the left primary somatosensory cortex (peak voxel), with local maxima in left primary motor area. The NFB > FB contrast did not reveal any significant activations.

Table 3

Main effect of Feedback Condition

Brain region	Side	Activation coordinates			Z-value	P-value	Number of voxels in cluster
		x	y	z
FB > NFB
Preparation
Precuneus	R	2	−76	50	5	<0.00001	12282
V3	R	8	−88	36	4.96
V1	R	8	−96	12	4.77
V2	L	−14	−98	10	4.67
Medial anterior prefrontal cortex	R	12	64	22	5.54	<0.00001	6072
Ventromedial prefrontal cortex	L	−2	54	16	4.86
Ventromedial prefrontal cortex	R	4	54	14	4.83
Dorsomedial prefrontal cortex	R	6	46	38	4.83
Dorsomedial prefrontal cortex	L	−6	64	26	4.73
Medial anterior prefrontal cortex	L	−14	64	26	4.76
Primary somatosensory cortex	L	−22	−42	72	3.54	0.00673	563
Primary motor cortex	L	−24	−26	64	3.24
Execution
Human middle temporal/V5+	L	−40	−68	6	9.26	0	40444
Human middle temporal/V5+	R	48	−62	−4	9.19
V4	R	32	−82	0	8.86
Insula	R	32	18	6	5.95	<0.00001	2049
Brainstem	R	6	−30	−4	4.74
Caudate	R	20	12	14	4.57
Brainstem	L	−6	−28	−2	4.33
Dorsolateral prefrontal cortex	R	32	38	28	5.79	<0.00001	1337
Insula	L	−28	20	10	5.09	0.00421	560
Putamen	L	−20	18	2	3.77
NFB > FB
Preparation
–	–	–	–	–	–	–
Execution
V2	L	−12	−96	16	7.82	<0.00001	8277
V2	R	12	−92	18	7.68
V1	L	−8	−96	6	7.16
V1	R	2	−82	0	6.92
V1		0	−84	−4	6.84
Dorsolateral prefrontal cortex	L	−40	16	46	4.55	<0.00001	2443
Inferior frontal gyrus	L	−42	26	20	4.46
Lateral orbitofrontal cortex	L	−42	40	−4	4.15
Primary motor cortex	L	−42	−18	54	5.73	0.00081	705
Primary somatosensory cortex	L	−50	−16	50	3.97
Primary somatosensory cortex	R	44	−20	54	4.72	0.00361	573
Inferior parietal lobe	L	−34	−64	46	4.06	0.00756	511
Inferior parietal sulcus	L	−34	−54	36	3.11
Secondary somatosensory cortex	R	44	−14	12	4.93	0.0113	478
Heschl’s gyrus	R	38	−26	6	3.83

Peak activations are presented in bold type. L = left; R = right.

In the execution phase (across groups), for FB > NFB, three main clusters revealed significant activations (Table 3) including a cluster within (i) the left human middle temporal cortex/V5+, with an extension into the right human middle temporal cortex/V5+ and right V4; (ii) insula (extending into bilateral brainstem and right caudate); (iii) right dorsolateral prefrontal cortex; and (iv) the left insula (peak voxel), extending into left putamen. The NFB > FB contrast revealed significant activations in six main clusters: (i) left V2 (including parts of bilateral V1 and right V2); (ii) left dorsolateral prefrontal cortex (extending into left inferior frontal gyrus and left lateral orbitofrontal cortex); (iii) left primary motor cortex (also left somatosensory cortex); (iv) right primary somatosensory cortex; (v) left inferior parietal lobe (also left inferior parietal sulcus); and (vi) right secondary somatosensory cortex (including parts of right Heschl’s gyrus).

Group × Feedback interaction

The interaction between Group and Feedback Condition revealed no significant findings in the movement preparation phase. However, in the execution phase, the Controls_{NFB > FB} > TBI_{NFB > FB} contrast demonstrated significant activations in three clusters: (i) left cerebellum lobe V (also including left temporal occipital fusiform cortex, right cerebellum lobe V and left parahippocampal gyrus); (ii) left primary motor cortex (extending into left superior parietal lobe and left primary somatosensory cortex); and (iii) right primary motor cortex (also including right primary somatosensory cortex). The TBI_{NFB > FB} > Controls_{NFB > FB} revealed one significant cluster: left caudate (extending into left putamen) (Table 4 and Fig. 4). To characterize these interaction patterns, ANOVAs and post hoc region of interest analyses were performed on the per cent signal changes extracted from the significant clusters (left cerebellum lobe V, bilateral primary motor cortices and left caudate). Confirming the previously mentioned results using the general linear model, these analyses revealed a significant Group × Feedback Condition interaction [left cerebellum lobe V: F(1,40) = 12.64, P = 0.001; left primary motor cortex: F(1,40) = 30.56, P = 0.000; right primary motor cortex: F(1,40) = 15.34, P = 0.000; left caudate: F(1,40) = 5.40, P = 0.025]. Tukey post hoc tests with respect to Controls_{NFB > FB} > TBI_{NFB > FB} contrast revealed, among the healthy controls, greater activations for the NFB than for the FB condition [left cerebellum lobe V (P = 0.034), left primary motor cortex (P = 0.000), right primary motor cortex (P = 0.007)], whereas no differences in activity levels were found between the Feedback Conditions among TBI patients (P > 0.05). With respect to the opposite contrast (TBI_{NFB > FB} > Controls_{NFB > FB}), healthy controls demonstrated greater activations for the FB than for the NFB condition in the left caudate (P = 0.027), whereas the TBI patients showed no differences in activity level between the Feedback Conditions (P > 0.05).

Figure 4

Brain regions demonstrating a significant Group × Feedback Condition interaction during movement execution. Per cent signal changes in peak coordinates of the cluster are presented. Activation maps are overlaid on ch2better template using MRIcron. Activation corresponding to the TBI group is represented in black (squares + full line), and to the control group in white (circles + dashed line). All activations have been thresholded with FSL clusterwise correction: Z > 2.3, P < 0.05. The error bars indicate the standard error of the mean. Line graphs were created using Microsoft Excel 2007. L = left, R = right.

Table 4

Interaction effects between Group and Feedback Condition.

Brain Region	Side	Activation coordinates			Z-value	P-value	Number of voxels in cluster
		x	y	z
TBINFB>FB > ControlsNFB>FB
Preparation
–	–	–	–	–	–	–	–
Execution
Caudate	L	−20	18	8	4.08	0.0448	371
Putamen	L	−18	6	8	3.87
ControlsNFB>FB > TBI NFB>FB
Preparation
–	–	–	–	–	–	–	–
Execution
Cerebellum lobe V	L	−2	−62	−12	4.04	<0.00001	1998
Temporal occipital fusiform cortex	L	−30	−54	−10	3.9
Cerebellum lobe V	R	2	−62	−10	3.77
Parahippocampal gyrus	L	−22	−40	−16	3.45
Primary motor cortex	L	−44	−14	56	4.81	<0.00001	1435
Primary somatosensory cortex	L	−50	−16	48	4.39
Superior parietal lobe	L	−18	−40	72	3.75
Primary motor cortex	R	30	−22	70	4.04	0.000205	834
Primary somatosensory cortex	R	50	−14	38	3.8

Peak activations are presented in bold type. L = left, R = right.

To support the idea of increased differentiation of activation patterns in response to different feedback conditions in the control group relative to the TBI group, we performed a whole-brain conjunction analysis (FB ∩ NFB). The common activation for FB and NFB conditions with respect to the total number of active voxels was calculated per participant (=percentage overlap). The larger the number of voxels that show common activation, the more similar the neural patterns are for the FB and NFB condition. During the execution phase, revealing significant Group × Feedback interactions, the percentage overlap between both feedback conditions was significantly larger in the patient (mean = 46.97, SD = 10.27) relative to the control group (mean = 28.57, SD = 9.84), [t(40) = −5.85, P = 0.000]. Note that this increase in common activation (common number of voxels) is a relative number compared with the total amount of activation (total number of voxels) (see Supplementary Table 1 for absolute numbers).

Relation between functional MRI activation and behaviour

Given the significant group differences in brain activation, region of interest-based correlations were performed to explore whether these between-group differences were related to distinct behavioural outcomes. During movement preparation, although TBI patients showed reduced activation as compared to the controls, a positive correlation was found between the activity pattern in the left inferior parietal lobe, and the subsequent performance (averaged across feedback conditions) (r = 0.65, P = 0.005), within the patient group. No significant across group correlations were found during the preparation phase. During the execution phase, significant positive correlations were found, across groups, between per cent signal changes in the left dorsolateral prefrontal cortex (r = 0.42, P = 0.006), left inferior parietal lobe (r = 0.39, P = 0.011), right inferior parietal lobe (r = 0.53, P = 0.000) and average target error (averaged across feedback conditions). Even though these regions were on average more activated in the control group than in the TBI group, a significant correlation was observed between activity within these regions and performance across participants of both groups. No significant within group correlations were observed during the execution phase. These results indicate that higher error scores on the bimanual tracking task were associated with higher percent signal changes within these brain areas (Supplementary Fig. 2).

Discussion

Here, we provide the first evidence that TBI patients activate frontal, parietal and occipital areas to a lesser extent than healthy controls during the movement preparation phase of a complex bimanual motor task. Importantly, the set of coordination tasks used represented a broad spectrum of the tasks pace of which some approximated those performed in everyday life more closely than others. Conversely, (sub)cortical overactivation was observed in TBI patients during the subsequent execution phase. Such TBI-induced functional alterations may underlie the poorer bimanual performance in the patient group across both feedback conditions. Interestingly, the current findings also indicated that TBI patients are less able to differentiate neural activation maps in response to the presence versus absence of augmented visual feedback, particularly in brain areas involved in sensorimotor information processing.

Behavioural performance post-traumatic brain injury

In line with previous research on TBI-induced general motor impairments (Gray et al., 1998; Kuhtz-Buschbeck et al., 2003a, b; Di Russo et al., 2005; Caeyenberghs et al., 2009b, 2010b), patients performed significantly poorer on the present bimanual coordination task than controls across both feedback conditions. Contrary to our expectations, however, the performance difference between groups was not more pronounced in the NFB condition, indicating that the TBI and control group were similarly affected by the removal of augmented visual feedback.

Group effect: imaging data

Brain areas less activated in the patient relative to the control group

During movement preparation, TBI patients showed reduced activations in various clusters compared to controls. TBI patients deactivated (relative to baseline) frontal (bilateral superior frontal gyrus and right dorsolateral prefrontal cortex), parietal (left inferior parietal lobe) and visual areas (right V1, bilateral V2), whereas the control group demonstrated activations within these significant clusters. These hypoactivations in the TBI group may suggest that our patients failed to recruit areas belonging to the anticipatory neural network (i.e. dorsolateral prefrontal cortex and parietal lobe) (Ghajar and Ivry, 2008). Consequently, TBI patients were probably less able to translate the available information on the upcoming movement into an appropriate action plan.

Comparable TBI-induced motor preparation impairments have previously been reported (Di Russo et al., 2005; Wiese et al., 2006; Ghajar and Ivry, 2008; Puopolo et al., 2013). Where Di Russo et al. (2005) showed a reduction and delay of the readiness potential, present in the motor cortex and supplementary motor area (using EEG), Wiese et al. (2006) reported altered activation levels in the caudate and dorsal lateral premotor cortex during movement preparation (using functional MRI). Moreover, a TBI-induced impaired ability for motor imagery (imagining a movement without actual action), an important aspect of motor preparation (Steenbergen et al., 2007), has previously been observed (Caeyenberghs et al., 2009a; Oostra et al., 2012).

Our study adds significantly to the previous work by addressing group differences in brain activation during the movement preparation as well as execution phase. Interestingly, the observed lower activation levels during the movement preparation phase in the TBI versus control group did not extend into the execution phase where higher activation levels emerged in the former as compared to the latter group, as discussed next.

Brain areas more activated in the patient relative to the control group

While no regions demonstrated enhanced activity in TBI patients as compared to controls during the movement preparation phase, hyperactivations for the patient group were present during the execution phase of the task. Increased cortical (mainly fronto-parietal) and subcortical (cerebellar) activations were demonstrated. Specifically, in view of the present motor task, dorsolateral prefrontal cortex (DLPFC) activation may be explained as enhanced online monitoring of motor performance (Pochon et al., 2001; Hoshi and Tanji, 2007), whereas activation in parietal areas may reflect increased sensory processing and integration (e.g. Chen et al., 2008). Moreover, activations in the cerebellum were particularly observed in Crus II with known projections to posterior parietal and prefrontal areas (Li et al., 2013). Even though this recruitment of generic prefrontal areas is less common in a motor context, the present activation of DLPFC and parietal areas during motor task performance is in partial agreement and converges with previous functional imaging work on cognitive tasks in TBI patients. During the performance of working memory as well as dual tasks, TBI patients have repeatedly demonstrated increased prefrontal cortex (PFC) activation (McAllister et al., 1999, 2001; Christodoulou et al., 2001; Perlstein et al., 2004; Hillary et al., 2006, 2008; Maruishi et al., 2007; Scheibel et al., 2007; Rasmussen et al., 2008; Turner and Levine, 2008; Medaglia et al., 2012). Three possible explanations have been put forward to account for this prefrontal recruitment: (i) permanent rewiring of the brain (i.e. reorganization); (ii) compensatory activation (i.e. additional recruitment of brain areas to facilitate performance) (McAllister et al., 1999, 2001; Maruishi et al., 2007; Turner and Levine, 2008); and (iii) the necessity of increased cognitive control and attentional resources (Hillary et al., 2006, 2008; Scheibel et al., 2007; Medaglia et al., 2012). As significant behavioural group differences were apparent in the present study as well as a negative association between fronto-parietal activations and performance, it appears reasonable to assume that the system acted in response to increases in task load (Hillary et al., 2006, 2008). In other words, additional neural resources may have been recruited to process novel/challenging information, and can therefore be considered as a cognitive control mechanism (Hillary et al., 2006). Experimental evidence supporting the hypothesis of increased cognitive control was reported by Medaglia et al. (2012) and Hillary et al. (2011). Both studies revealed that when a working memory task became less novel due to practice, prefrontal activation (Medaglia et al., 2012) and/or anterior (in relation to parietal) connectivity (Hillary et al., 2011) decreased, suggesting that the threshold for recruitment of latent resources was dependent on the level of cerebral challenge (in both patients and controls). Thus, irrespective of the task domain (motor and cognition), an increased need for cognitive control emerged. As such, TBI-induced impaired task automaticity may have resulted in enhanced cognitive control, as reflected by generic frontal and/or parietal recruitment during either cognitive or motor task performance.

Important to note, the findings were corrected for group differences in grey matter volume by adding a grey matter mask to the general linear model for each subject. Therefore, we tentatively conclude that significant between-group effects were mainly task-based. Nonetheless, all models were run without the grey matter mask as well and revealed principally similar results.

Feedback effect: imaging data

Brain areas more activated during feedback than no-feedback conditions

During movement preparation (across groups), the occipito-parieto-frontal pathway was activated in anticipation for the upcoming augmented visual feedback. Occipito-parietal activations were expected, as these regions are involved in the dorsal visual stream, which is important for visually-guided behaviour (de Haan and Cowey, 2011). The activation differences in the medial anterior prefrontal cortex and primary somatosensory cortex were characterized by more deactivation during preparation for the NFB than the FB trials. Being part of the default mode network, the medial anterior prefrontal cortex is activated at rest, and deactivated during task performance (Greicius et al., 2003). Apparently, there is more deactivation in this brain area in case of upcoming NFB trials.

During movement execution (across groups), the FB > NFB contrast evoked activations in the human middle temporal cortex/V5+ bilaterally, right dorsolateral prefrontal cortex and bilateral insula, with local maxima in the caudate and putamen, consistent with previous research (Beets et al., 2015). Activation in the human middle temporal cortex/V5+, specialized for motion processing (Born and Bradley, 2005), is in agreement with previous studies where a dynamic Lissajous figure (integrating two kinematic signals into a gestalt) was used to guide complex bimanual coordination tasks online (Debaere et al., 2004; Ronsse et al., 2011). Anterior dorsolateral prefrontal cortex activation, in turn, reflects attention to action, which is needed to continuously monitor and adjust ongoing behaviour based on available augmented feedback (Cieslik et al., 2013). Furthermore, integrating visual and somatosensory information during the feedback condition may have resulted in increased insular activity to dissociate internal from external sensory signals (Amedi et al., 2005). Finally, enhanced activations in the caudate and putamen can be explained in light of their involvement in early motor learning (Grafton et al., 1995; Jueptner et al., 1997; Hardwick et al., 2013), particularly when feedback is available (Tricomi et al., 2006; Choi et al., 2013).

Brain areas more activated during no-feedback than feedback conditions

Although there were no brain regions significantly more active during the NFB than the FB condition during movement preparation, various clusters demonstrated higher activity levels for the NFB condition during movement execution. Observed activations in somatosensory brain areas particularly, revealed that participants relied more on somatosensory information to overcome the absence of augmented visual feedback, as expected. Moreover, for this contrast the posterior dorsolateral prefrontal cortex was activated, which is supportive of motor control functions, particularly when a stimulus-induced response and the retrieval of spatial information from working memory is required (Cieslik et al., 2013).

Interaction between Group and Feedback Condition: imaging data

During movement preparation no significant interaction effects were found. With respect to the execution phase, however, our results revealed that activations accompanying FB and NFB conditions vary across groups. More specifically, healthy controls activated cerebellar and sensorimotor areas to a greater extent during the NFB than the FB condition, whereas TBI patients showed no significant functional activation differences between both conditions. Controls may have been able to rely on better-developed internal sensory prediction models in the NFB trials. Consequently, their sensorimotor processing to map incoming sensory input to these models may have been more intensive. Moreover, healthy controls activated the caudate to a greater extent during the FB than the NFB condition, whereas TBI patients showed no significant functional activation differences between both feedback conditions. As previously mentioned the caudate is an important subcortical structure, which is involved in performance feedback (Tricomi et al., 2006). Apparently, healthy controls are better able to adapt the activity level to the requested movement.

The absence of significant activation differences between both FB conditions in the TBI group may reflect a more generic feature of TBI-induced functional brain alteration. Indeed, the conjunction analyses (FB ∩ NFB), demonstrated more overlap in brain activation across both FB conditions in the patient relative to the control group. This may reflect decreased specialization in brain activation in the TBI group, a process of brain dedifferentiation.

Relation between functional MRI activation and behaviour

Correlational analyses between behavioural performance and per cent signal changes in areas demonstrating significant group differences, revealed significant correlations within parietal and frontal areas. In other words, higher per cent signal changes were associated with higher error scores, i.e. poorer performance. More specifically, within the patient group, a higher per cent signal change in the inferior parietal lobe during movement preparation was associated with higher error scores. Regarding the execution phase, similar correlations were found across both groups within the left dorsolateral prefrontal cortex and bilateral inferior parietal cortices. With respect to the execution phase, the scatterplots reveal a continuum between patients and controls, with lower error scores for control subjects and higher error scores for the patient group. Pooling both groups together, it became clear that activating these attention-related brain areas, involved in performance monitoring (Taylor et al., 2007), was accompanied with worse performance. This correlation may reveal less automaticity, reflecting increased cognitive control to perform this complex bimanual coordination task.

Conclusion

In conclusion, our findings suggest that TBI patients demonstrate apparent movement coordination deficits, concomitant with altered neural activation patterns. Reduced activations during movement preparation were followed by widespread hyperactivations during execution, reflecting increased cognitive control over action. The increased neural recruitment during movement execution suggests that performance monitoring is less automatic and this may leave less room for dividing attention to other tasks for multitask purposes.

Funding

This work was supported by the Interuniversity Attraction Poles program of the Belgian federal government (P7/11). Additional support was provided by the Research Fund KU Leuven (C16/15/070) and FWO Vlaanderen (G.0721.12, G0708.14). J.G. is funded by a fellowship of the Research Foundation – Flanders (FWO).

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

Abbreviations
 
• FB

  augmented visual feedback

 
• NFB

  no augmented visual feedback

 
• TBI

  traumatic brain injury