OUP user menu

Functional magnetic resonance imaging of chronic dysarthric speech after childhood brain injury: reliance on a left-hemisphere compensatory network

Angela T. Morgan, Richard Masterton, Lauren Pigdon, Alan Connelly, Frédérique J. Liégeois
DOI: http://dx.doi.org/10.1093/brain/aws355 646-657 First published online: 31 January 2013


Severe and persistent speech disorder, dysarthria, may be present for life after brain injury in childhood, yet the neural correlates of this chronic disorder remain elusive. Although abundant literature is available on language reorganization after lesions in childhood, little is known about the capacity of motor speech networks to reorganize after injury. Here, we examine the structural and functional neural correlates associated with chronic dysarthria after childhood-onset traumatic brain injury. Forty-nine participants aged 12 years 3 months to 24 years 11 months were recruited to the study: (i) a group with chronic dysarthria (n = 17); matched for age and sex with two control groups of (ii) healthy control subjects (n = 17); and (iii) individuals without dysarthria after traumatic brain injury (n = 15). A high-resolution 3D T1-weighted whole-brain data set was acquired for voxel-based morphometry analyses of group differences in grey matter. Functional magnetic resonance imaging was used to localize activation associated with speaking single words (baseline: listening to words). Group differences on voxel-based morphometry revealed widespread grey matter reductions in the dysarthric group compared with healthy control subjects, including in numerous speech motor regions bilaterally, such as the cerebellum, the basal ganglia and primary motor cortex representation of the articulators. Relative to the non-dysarthric traumatic brain injury group, individuals with dysarthria showed reduced grey matter bilaterally in the ventral sensorimotor cortex, but this reduction was concomitant with increased functional activation only in the left-hemisphere cluster during speech. Finally, increased recruitment of Broca’s area (Brodmann area 45, pars triangularis) but not its right homologue, correlated with better speech outcome, suggesting that this ‘higher-level’ area may be more critically involved with production when associated motor speech regions are damaged. We suggest that the bilateral morphological abnormalities within cortical speech networks in childhood prevented reorganization of speech function from the left- to right-hemisphere. Rather, functional reorganization involved over-recruitment of left-hemisphere motor regions, a reorganization method that was only partly relatively effective, given the presence of persisting yet mild speech deficits. The bilateral structural abnormalities found to limit functional reorganization here, may also be critical to poor speech prognosis for populations with congenital, degenerative or acquired neurological disorders throughout the lifespan.

  • brain injury
  • traumatic brain injury
  • plasticity
  • speech
  • dysarthria
  • functional MRI


Traumatic brain injury is the most common form of acquired neurological injury in childhood, occurring in ∼5–6 children per 100 000 of the population (Parslow et al., 2005). Dysarthria, a motor speech disorder, is not uncommon in this group, occurring in 20% of children with moderate or severe traumatic brain injury in the acute period (Morgan et al., 2010). There are no proven treatments available in childhood (Morgan and Vogel, 2008), and the disorder often persists into late adolescence and early adulthood (Cahill et al., 2002, 2005). Despite the potentially intractable nature of this impairment, examination of prognostic factors is sparse. In particular, the neural correlates for dysarthria acquired in childhood have not been examined using either structural or functional neuroimaging (for review see Liégeois and Morgan, 2012). The location of structural abnormalities is presumably one key marker for predicting clinical outcome (Liégeois et al., in press), yet a fully informed picture of brain–behaviour relationships also requires consideration of functional activity in grey matter regions that are key to speech production. Success of recovery is also likely to be dependent on the efficacy of repair or reorganization of functional networks associated with speech. Here, we used both structural and functional neuroimaging to examine the underlying neural mechanisms associated with chronic dysarthria acquired in childhood after traumatic brain injury.

Despite functional MRI advances in understanding typical speech control, where speech production in healthy adults is reportedly subserved by a predominantly left-hemisphere network (Riecker et al., 2005; Eickhoff et al., 2009; Price, 2010), few studies have directly examined functional activation associated with disordered, and specifically, dysarthric speech. To our knowledge, no study has examined the neural correlates of adult-acquired dysarthria following traumatic brain injury. However, reports on dysarthric adults after stroke and hereditary ataxia have suggested two different forms of functional neural reorganization or compensation for speech disorder after brain injury. A ‘right shift’ of speech motor control was found in a case of left-hemisphere acute stroke with full speech recovery (Riecker et al., 2002). By contrast, a profile of left-hemisphere overactivation of speech regions was reported in dysarthric adults with hereditary ataxia (Sidtis et al., 2010) and presumably more bilateral and widespread damage than seen in the case of focal unilateral stroke.

No functional MRI study of dysarthria in childhood has been reported as far as we are aware. However, reorganization patterns during language functional MRI tasks seen in children with focal injury (e.g. epilepsy, Liégeois et al., 2004; Gaillard et al., 2007; and stroke, Guzzetta et al., 2008; Tillema et al., 2008; Fair et al., 2010) may include (i) recruitment of regions around the lesions within the left language-dominant hemisphere (intra-hemispheric reorganization) or (ii) recruitment of right-hemisphere homologous regions (interhemispheric reorganization). A third pattern of organization has been reported in an inherited form of childhood apraxia of speech (KE family), where under activation of typical language regions (Broca’s area and putamen) was accompanied by recruitment of regions not traditionally involved in language, such as posterior parietal regions (Liégeois et al., 2003, 2011), showing an instance of ‘alternative network’ organization. A similar alternative pattern of reorganization has been reported in a study examining children post traumatic brain injury, where both overactivation of Broca’s area and additional reliance on the dorsolateral prefrontal cortex was observed relative to a control group with orthopaedic injuries (Karunanayaka et al., 2007). To our knowledge, however, functional imaging profiles of dysarthria associated with childhood traumatic brain injury have not been examined.

Whether chronic dysarthric speech is associated with (i) intra-hemispheric reorganization; (ii) inter-hemispheric reorganization; or (iii) recruitment of new areas in either hemisphere, remains unknown. Stimulation of regions crucial to good recovery, or of those associated with maladaptive compensation, is a growing field to assist in the recovery of language post stroke in adults. Specifically, identification of effective regions may guide the design of more efficient non-invasive brain stimulation treatments (Turkeltaub et al., 2011). For instance, the identification of Broca’s area as associated with effective anomia recovery has lead to the hypothesis that transcranial direct stimulation of this region in the acute stage can favour functional outcome (Holland et al., 2011). Similarly, in the future, it is conceivable that behavioural or brain stimulation techniques aimed at altering reorganization patterns could also be developed to optimize speech outcome in children.

The overall aim of the present study was to examine the grey matter structural and functional correlates of chronic dysarthria after traumatic brain injury sustained in childhood. Because traumatic brain injury can result in widespread and heterogeneous damage, we chose to use whole-brain voxel-wise analyses (voxel-based morphometry) to identify structural grey matter disruption in our cohort. We hypothesized that participants with chronic dysarthria would have grey matter morphological abnormalities (i.e. increased or reduced grey matter) bilaterally, as chronic dysarthria subsequent to unilateral lesion is rarely reported in the literature (Liégeois and Morgan, 2012). We expected abnormalities in one or several regions of the ‘motor speech network’, based on these regions being associated with childhood dysarthria (Liégeois and Morgan, 2012). In addition, participants with dysarthria were hypothesized to show atypical functional MRI activation (e.g. under/overactivation) within this production network, reflecting reorganization in response to brain injury. We refer to ‘motor speech networks’ as regions related to the motoric aspect of speech, that is, primary sensorimotor cortex, basal ganglia and cerebellum, with largely bilateral involvement. These networks exclude any regions involved in phonological or semantic processing or retrieval, or ‘language’ networks. It is without a doubt that the language and speech networks interact highly during everyday communication, as evidenced by most current models of speech production that include both predictions and error monitoring of motor output (see Price et al., 2011 for an overview). Yet, dysarthria is a speech motor and not a language disorder, and dissociations do exist. For that reason, our functional MRI study was also designed to control for language-related processes (as single words were presented during both baseline and task periods to control for phonological and semantic processing) and instead focus on the motoric aspects of speech.

Finally, it was hypothesized there would be a correlation between functional task activation and speech performance, age at injury and time since injury for the groups with traumatic brain injury. There is currently a debate regarding whether early brain injury results in better or worse neuropsychological outcome (early plasticity versus early vulnerability debate, see Anderson et al., 2011 for review; Lidzba et al., 2009 for comment). Specifically, there is evidence to suggest that children with early diffuse injury, such as traumatic brain injury, are at greater risk of long-term neuropsychological deficits than those with later injuries (Anderson et al., 2010). For these reasons, we hypothesized that functional MRI activation in some neural regions would be correlated with age at injury. In relation to time since injury, adult stroke literature indicates that patterns of reorganization for language function change over time (see Saur and Hartwigsen, 2012 for review). Increased reliance of the right-hemisphere homologue to Broca’s area has been reported in the acute phase, followed by return to left-hemisphere activation in the chronic phase (4–12 months), associated with a reduction in aphasic symptoms. In children, although little neuroimaging evidence is available, it is similarly plausible that functional MRI activation patterns change with time post-injury, given that both lateralization and focus changes have been noted for language production over time in the healthy developing brain (Lidzba et al., 2011). There is a non-existent literature base for brain–behaviour relationships with regard to long-term dysarthria outcome in children. Here, we aimed to explore whether there was a correlation between neural activation and time post injury in participants with traumatic brain injury.

Materials and methods


Participants with traumatic brain injury were identified for recruitment based on an extensive medical chart review (Morgan et al., 2010) at the main paediatric tertiary care centre (the Royal Children’s Hospital) for acute and rehabilitative management of children with traumatic brain injury in Melbourne Australia (see Table 1 for participant information). Individuals with traumatic brain injury were required to have sustained their brain injury before 18 years of age and were to be >1 year post-injury at study commencement to avoid effects of swelling on magnetic resonance data analysis. Participants with a history of pre-injury developmental or neurological disorder were excluded (e.g. speech and language impairment, movement disorder or autism spectrum disorder). Participants with traumatic brain injury and dysarthria had to fulfil two further criteria: (i) a diagnosis of dysarthria made by a speech and language pathologist at first admission post injury; and (ii) confirmed diagnosis of chronic/persistent dysarthria at the time of recruitment by a speech and language pathologist rating a 5-min conversational sample, using the Mayo clinic rating system (Liegeois et al., 2010; Morgan et al., 2011). Acute dysarthria severity for the traumatic brain injury with dysarthria group is presented in Table 1 and was largely moderate to severe across the group (severe n = 8; moderate–severe n = 3; moderate n = 3; mild–moderate n = 2; unknown n = 1). No participant with traumatic brain injury with dysarthria had apraxia of speech. For participants with traumatic brain injury without dysarthria, a diagnosis of speech disorder in the acute stage was an additional exclusion criterion. Both groups with and without dysarthria post-traumatic brain injury were also matched for injury severity (mild, moderate, severe denoted by the Mayo clinic traumatic brain injury classification system; Malec et al., 2007), cause of injury (e.g. fall, blunt trauma, motor vehicle accident) and time post-injury wherever possible (Table 1). Forty-nine participants agreed to take part: n = 17 in the traumatic brain injury with dysarthria group [nine female subjects, mean age 17.8 years, standard deviation (SD) = 4.05 years, range 11.5–25 years]; n = 15 in the traumatic brain injury without dysarthria group (nine female subjects, mean age 18.73, SD 4.29 years, range 12.25–24.67 years) and n = 17 typically developing control participants (nine female subjects, average age 18.31 years, SD 4.87 years, range 11.7–24.7 years). The two-tailed t-test assuming unequal variances confirmed that average time post-injury was not statistically significantly different (P = 0.83) for the traumatic brain injury with dysarthria (8 years 3 months, SD = 5 years 1 month) and traumatic brain injury without dysarthria groups (8 years, SD = 2 years). A one-way ANOVA revealed statistically significant differences between IQ scores (assessed using the Wechsler Abbreviated Scales of Intelligence, Wechsler, 1999) across the groups at the time of study (F = 4.8, P = 0.012; traumatic brain injury with dysarthria: mean = 90.6, SD = 12.02; traumatic brain injury without dysarthria: mean = 99.6, SD = 12.65; typically developing: mean = 103.7, SD = 13.19). Post hoc Tukey’s Honestly Significant Difference test revealed that the only statistically significant difference in IQ scores existed between typically developing and traumatic brain injury with dysarthria groups (P < 0.05). Further, one-way ANOVAs revealed no significant difference between groups on performance IQ (F = 2.47; P = 0.096), and a difference between groups on verbal IQ (F = 6.13, P < 0.005) was found for the traumatic brain injury with dysarthria versus typically developing groups comparison only. Statistically significant differences between groups were revealed by analysis of co-variance with IQ as a covariate, on a standardized test of language function, the Clinical Evaluation of Language Fundamentals (CELF-IV)—core-language (F = 3.5, P = 0.039; traumatic brain injury with dysarthria: mean = 88.6, SD = 12.23; traumatic brain injury without dysarthria: mean = 96.5, SD = 11.65; typically developing: mean = 104.9, SD = 11). A post hoc comparison (Bonferroni corrected) demonstrated differences between the traumatic brain injury with dysarthria and typically developing groups only (P = 0.042). Four participants from the traumatic brain injury with dysarthria group and one from the traumatic brain injury without dysarthria group had language impairment, defined as a score of >1.5 SD below the mean on the CELF-IV.

View this table:
Table 1

Demographic and clinical data for participants across three groups (traumatic brain injury with dysarthria, traumatic brain injury without dysarthria and typically developing brains)

Traumatic brain injury with dysarthriaTraumatic brain injury without dysarthriaTypically developing
IDSexAge at injury (years)Time post-injury (years)Injury severityInjuryDysarthria severity (acute)Dysarthria type and severity (current)SexAge at injury (years)Time post-injuryInjury severityInjurySexAge (years)
1M7.174.33Mod–SevFallSevMild mixed spastic–ataxicM11.7
2F11.425.91Mod–SevMVA PassengerMod–SevMild mixed hypokinetic–ataxicF8.008.58Mod–SevBlunt traumaF17.5
3F8.173.91MildFallModMild spasticF7.835.09MildMotorcycle accidentF12.7
4F12.675.08Mod–SevMotorcycle accidentMod–SevMild mixed spastic–flaccid–ataxicF9.507.25Mod–SevMotorcycle accidentF17.9
5M3.0014.33Mod–SevMVA PedestrianModMild mixed spastic–ataxicM9.758.50Mod–SevMVA PedestrianM25.1
6F3.5813.92Mod–SevMVA PedestrianSevMild mixed spastic–ataxicF12.254.42Mod–SevMVA PedestrianF16.5
7M8.084.42Mod–SevMVA PedestrianMild–modMild mixed spastic–ataxicM4.088.17Mod–SevMVA PassengerM12.6
8F11.582.25Mod–SevFallSevMild-mod mixed spastic–flaccidF5.928.58Mod–SevMVA PedestrianF13.9
9F14.676.41Mod–SevFallMod–SevMild spasticF16.175.00Mod–SevMVA PedestrianF21.11
10M7.5015.42Mod–SevMVA PedestrianSevMild spasticM10.9212.00Mod–SevMVA PedestrianM23.8
11F7.3316.17Mod–SevMVA PassengerSevMild–mod mixed spastic–flaccid–ataxicF15.927.58Mod–SevMVA PassengerF24.1
12M3.5016.67Mod–SevMVA PedestrianMild–modMild mixed spastic–flaccid–ataxicM12.259.67Mod–SevMVA PassengerM21
13M3.679.33Mod–SevMVA PedestrianUnknownMild–mod spasticM5.927.58Mod–SevFallM11.9
14M14.082.67Mod–SevMVA PassengerSevMild spasticM13.179.91Mod–SevFallM15.11
15F16.508.50Mod–SevAssaultedSevSev spasticF15.589.09Mod–SevMVA PedestrianF24.7
16M12.923.66Mod–SevFallSevMild spasticM18.0
17F13.427.16Mod–SevFallModMild mixed spastic–hyperkinetic–ataxicF15.427.83Mod–SevFallF23.4
  • – = No TBI-matching participants could be identified.

  • Mod = moderate, MVA = motor vehicle accident, Sev = severe.

During the acute and early rehabilitation phase, children in the traumatic brain injury with dysarthria group received standard speech and language pathology services of assessment and monitoring. No child received any specific treatment for their dysarthria either in the acute phase or at any time after their injury. This is not surprising, given there are no proven effective treatments for childhood acquired dysarthria (see Cochrane database of systematic reviews: Morgan and Vogel, 2008; Pennington et al., 2009). Perhaps because of the lack of an evidence base, an international survey of paediatric brain injury speech rehabilitation revealed that no centre had set treatment protocols nor used intensive dysarthria treatments (Morgan and Skeat, 2011). Participants in both traumatic brain injury groups were managed by other multi-disciplinary professionals (e.g. neuropsychologists, physiotherapists and occupational therapists) where necessary as part of their regular inpatient and outpatient care. Ethics approval was obtained from the Royal Children’s Hospital Human Research Ethics Committee (HREC), approval number HREC 27083.

Speech assessment

Perceptual speech (dysarthria) ratings were made for all three groups (traumatic brain injury with dysarthria, traumatic brain injury without dysarthria and typically developing) based on a recorded 5-min conversation and standard paragraph reading. Connected speech tasks such as these are more ecologically valid than structured assessments because they better approximate the individual’s functional level of communication (Bunton et al., 2007). Two speech and language pathologists blinded to the participants’ medical histories performed the speech ratings (one with 8 years and one with 12 years of experience working with adults and children with dysarthria). Speech samples were rated to determine the presence, severity and type of dysarthria. The following criteria were used to diagnose dysarthria: presence of an isolated speech deficit or a cluster of speech deficits at any level of the speech subsystem (i.e. pitch, vocal quality, loudness, resonance, respiration or prosody) that impacted on intelligibility and/or naturalness of speech as rated by the listener (adapted from Darley et al., 1975). Dysarthria was rated using a modified version of the Mayo dysarthria classification (Duffy, 2005, p. 90), considered to be the most commonly used approach both clinically (Simmons and Mayo, 1997) and for research (Duffy and Kent, 2001; Kent et al., 2001), and as frequently applied in the field of acquired childhood dysarthria (e.g. Cahill et al., 2005; Morgan et al., 2007; Liégeois et al., 2010; Morgan and Liégeois, 2010; Morgan et al., 2011). This modified version of the dysarthria rating scale assesses speech dimensions related to the domains of pitch, loudness, voice quality, resonance, respiration, prosody and articulation (Table 2) and were rated using a scale from 0 (normal), 1 (sub-clinical), 2 (mild), 3 (moderate) and 4 (severe). An overall dysarthria severity rating, indicative of intelligibility and naturalness of the child’s everyday conversational speech, was also rated on a scale from 0 (normal), 1 (mild), 2 (mild-moderate), 3 (moderate) and 4 (severe). Raters provided consensus ratings where the judgements were discordant. Consensus ratings and the average number of deviant features for each of the groups (calculated by summing the number of deviant features demonstrated by all individuals in a group, divided by the total number of participants in a group) are reported in Table 2.

View this table:
Table 2

Frequency of perceptually detected speech deviations in individuals with traumatic brain injury and dysarthria (TBI+, n = 17), with traumatic brain injury without dysarthria (TBI−, n = 15) and typically developing control participants (TD, n = 17)

Speech domainFeatureTBI+TBI−Typically developing
Frequency of deviation (%)Frequency of deviation (%)Frequency of deviation (%)
PitchAltered (increased/decreased)921
Reduced overall710
Voice qualityHarsh798
Audible inspiration100
ProsodyReduced rate1323
Short phrases200
Prolonged intervals300
Inappropriate silences100
ArticulationImprecise consonants1621
Prolonged phonemes1511
Distorted vowels1100
X (SD) number of deviant features6.9 (5.5)1.8 (2.7)1.1 (1.8)
VMPAC SC subtest (SD)73.0 (26.7)96.1 (10.2)98.3 (7.0)
  • VMPAC SC = Verbal Motor Production Assessment for Children, Speech Characteristics.

Ratings by the two speech and language pathologists were compared using Cohen’s κ. A high degree of inter-rater reliability was determined between the raters (Cohen’s κ was 0.65; M% agreement = 84%; range = 64–94.5%). The majority of ratings clustered at the mild end point of the scale, consistent with our previous paediatric reports of mild dysarthria associated with other forms of acquired brain injury (Liégeois et al., 2010; Morgan et al., 2011).

In addition to the Mayo clinic defined perceptual dysarthria analyses, a standardized measure of speech performance was taken using the speech characteristics subtest of the Verbal Motor Production Assessment for Children (Hayden and Square, 1999). This subtest consists of seven items, including evaluation of a child’s speech motor movements in familiar and unfamiliar sequences, and whether an individual has pitch, resonance, voice quality, volume intensity, prosody/intonation and rate of speech within normal limits for their age. As such, this subtest acts as a representative and standardized measure of the child’s dysarthria profile. A higher score indicates better speech function. Although children aged 7–12 years with average performance will reach ceiling on this subtest, it is sensitive enough to show differences between individuals with and without dysarthria (Table 2).

Magnetic resonance imaging data acquisition

T1-weighted magnetisation-prepared rapid gradient echo (MP-RAGE) MRI data sets were acquired on a Siemens 3 T Tim Trio MRI scanner for voxel-based morphometry analyses (isotropic voxel size = 0.9 mm, repetition time = 1900 ms, echo time = 2.6 ms, flip angle = 9°).

Functional MRI data were acquired using a multi-slice echo-planar imaging sequence with whole-brain coverage (repetition time = 3000 ms, echo time = 30 ms, flip angle 85°, 44 slices, matrix size = 72 × 72, 3-mm isotropic voxels). Two consecutive runs of 120 volumes were collected for each participant, with the same functional MRI speech paradigm (described below) repeated in each run. Echo-planar imaging data were collected continuously throughout each run.

Since dysarthria is a disorder of neuromuscular execution and/or control, the functional MRI task was designed to measure the blood–oxygen level-dependent response associated with motor production and not processing of speech. The functional MRI speech paradigm consisted of repeating single words (‘active’ condition) contrasted with only listening to the stimuli (‘baseline’). Stimuli were presented auditorily to the participants preceded by the phrase ‘speak now’ (active) or ‘listen now’ (baseline condition). The word lists for both conditions were randomly generated from a standardized assessment, the Children’s Speech Intelligibility Measure (CSIM, Wilcox and Morris, 1999). The paradigm comprised five active task/baseline cycles, with six stimuli (three active, three baseline) of 12-s duration each per cycle, giving a total duration of 6 min. Each run consisted of 30 stimuli in total (i.e. 15 rest/15 active) with 60 stimuli delivered across the two runs. The 12-s period for each stimulus consisted of 2.5 s for stimulus presentation (i.e. ‘speak now car’) followed by the participant response within 2.0 s (i.e. ‘car’) and a further 7.5 s of scanning time without stimulus presentation to capture the peak of the haemodynamic response. This relatively slow event presentation rate ensured that head motion during overt speech did not impact on the measurement of the haemodynamic response to a preceding event.

Participant’s compliance with the task in the scanner was monitored online to ensure that children were speaking or not speaking appropriately dependent on the particular stimuli.

Functional magnetic resonance imaging analyses

Data sets were preprocessed and analysed using Statistical Parametric Mapping 8 software (http://www.fil.ion.ucl.ac.uk/spm/software/spm8). Data preprocessing steps included a correction for timing delays because of slice acquisition differences, spatial realignment to correct for subject motion, spatial normalization to Montreal Neurological Institute (MNI) template space and spatial smoothing using an 8-mm full-width at half-maximum Gaussian smoothing kernel.

The functional MRI data were processed using a two-level random-effects analysis. At the first-level, an event-related analysis was performed for each subject with a contrast of interest comparing the active condition (speaking single word) against the baseline condition (listening to single word). The first-level contrasts were then analysed at the second-level to generate group activation maps and comparisons. One sample t-tests were used to test for common areas of activation within each group, and two-sample t-tests to test for areas of over- or under-activation in the traumatic brain injury with dysarthria group relative to the two control groups (traumatic brain injury without dysarthria and typically developing).

As it was hypothesized that there would be a correlation between functional task activation and (i) speech performance; (ii) age at injury; and (iii) time since injury for the groups with traumatic brain injury, three separate exploratory correlation analyses were conducted to address each of these questions. A partial correlation approach was conducted for the speech performance analyses with age as a covariate. The speech performance rating was based on each traumatic brain injury participant’s score on the speech characteristics subtest of the Verbal Motor Production Assessment for Children (Hayden and Square, 1999).

All functional MRI results were reported at a threshold of P = 0.001, uncorrected for multiple comparisons across the whole brain, with an extent threshold of 10 voxels, except for the correlation analysis where we had no a priori hypothesis (P = 0.05, family wise error correction).

Voxel-based morphometry analyses

T1-weighted data sets were segmented into grey matter, white matter and CSF using the New Segmentation toolbox implemented in Statistical Parametric Mapping 8. The algorithm is similar to that of the Unified Segmentation (described in Ashburner and Friston, 2005), but with, amongst others, an improved registration model and more robust initial affine registration (see http://www.fil.ion.ucl.ac.uk/spm/software/spm8/ for details). The data were then smoothed using an 8-mm full-width at half-maximum Gaussian kernel. This degree of smoothing was applied because of the a priori hypotheses anticipating bilateral morphological abnormalities of relatively small brain structures, i.e. the putamen and thalamus, in individuals with chronic dysarthria. After review of segmented images, no participants were excluded on the basis of inadequate preprocessing. The normalized modulated images were used for subsequent analyses.

Between-group comparisons of regional differences were conducted in SPM 8, co-varying for age, gender and total intracranial volume (defined as grey matter + white matter + CSF volumes extracted using the Easyvolume tool implemented in Statistical Parametric Mapping 8, http://www.sbirc.ed.ac.uk/LCL/LCL_M1.html). Group comparisons were performed using ANCOVAs, with a statistical threshold of P < 0.005 (uncorrected for multiple comparisons), and a minimum cluster size threshold of 10 voxels. Because we aimed to identify group differences within the speech network, groups comparisons between the two traumatic brain injury groups were restricted to speech relevant regions using an inclusive mask created from the speak > listen contrast (typically developing group, at P = 0.05 uncorrected). Unmasked results can be found in Supplementary Fig. 2.


Speech assessment

One-way ANCOVA, with age as a covariate, revealed a group effect for the speech characteristics subtest of the Verbal Motor Production Assessment for Children (F = 10.55, P < 0.0002; see Table 2 for average scores). The traumatic brain injury with dysarthria group scored significantly lower than the traumatic brain injury without dysarthria (P < 0.001) and typically developing groups (P < 0.002). Perceptual ratings revealed a generally mild dysarthria severity profile for the traumatic brain injury with dysarthria group (13 mild, three mild–moderate, one severe; Table 1), characterized by features of mono-pitch, mono-loudness, breathy voice, hypernasality, reduced speech rate, imprecise consonants and vowels and prolonged phonemes (Table 2). The most dominant dysarthria profile was spastic dysarthria, followed by mixed dysarthrias (seven spastic, four mixed spastic–ataxic, three mixed spastic–flaccid–ataxic, one mixed spastic–hyperkinetic–ataxic, one mixed spastic–flaccid and one mixed hypokinetic–ataxic) in-line with previous reports of dysarthria associated with adult traumatic brain injury (Duffy, 2005).

A one-way ANCOVA with age as covariate revealed a statistically significant difference in the mean number of deviant speech features rated across groups (F = 14.43, P < 0.0001). Post hoc Tukey’s Honestly Significant Difference test revealed statistically significant differences between the traumatic brain injury with dysarthria and typically developing groups (P < 0.01) and the traumatic brain injury with dysarthria and traumatic brain injury without dysarthria groups (P < 0.01), but no significant difference was revealed between the two control groups. Further, it is important to highlight that ratings of deviant speech features across the control groups (traumatic brain injury without dysarthria and typically developing) were only ever given a subclinical rating on the Mayo Clinic dysarthria classification system (score of 1), whereas ratings for the dysarthric group were almost always mild or greater (score of ≥2). Therefore, whilst deviant speech features were noted across all groups, they were markedly different levels of severity. Finally, the similar frequency of harshness noted across groups (with only subclinical ratings in typically developing and traumatic brain injury without dysarthria groups) was not unexpected given that voice quality metrics can arise from a broad range of aetiologies, such as dehydration (Witt et al., 2011) caffeine (Erickson-Levendoski and Sivasankar, 2011), oedema associated with vocal strain, over-use or laryngitis or other structural vocal fold deficits and hence is also a common finding in non-neurological populations.

Functional magnetic resonance imaging

For the speak > listen contrast analyses, all three groups demonstrated bilateral activation in a typical speech network involving the sensorimotor and superior temporal regions, thalamus, putamen and cerebellum (Fig. 1 and Supplementary Table 1).

Figure 1

Average activation during speaking (versus listening) in the typically developing (TD), traumatic brain injury without dysarthria (TBI−) and traumatic brain injury with dysarthria (TBI+) groups. Results displayed at P < 0.001 uncorrected for whole-brain multiple comparisons and superimposed onto each group’s mean T1-weighted image. Left-hemisphere is on the right.

The group comparison activation maps revealed increased activation for the traumatic brain injury with dysarthria group relative to the traumatic brain injury without dysarthria group in the left sub-gyral white matter of the parietal lobe (MNI coordinates −20, −46, 20; cluster size = 11 voxels; T = 3.61; P < 0.001 uncorrected), which was not hypothesized and is not discussed further. Increased activation was also noted for the traumatic brain injury with dysarthria group in the left sensorimotor cortex (pre- and postcentral gyrus) in comparison with typically developing group [Brodmann area (BA) 6, MNI coordinates −42, −14, 26; cluster size = 16 voxels; T = 3.67; P < 0.001 uncorrected]. No functional differences between groups were observed in the basal ganglia or cerebellum. There was a significant positive correlation for the traumatic brain injury with dysarthria group for speech-related activation in Broca’s area (BA 45, pars triangularis, MNI coordinates −48, +36, +2; cluster size = 2 voxels; T = 8.45; P = 0.032, family wise error corrected; cluster size >100 voxels at a threshold of P = 0.001, uncorrected, Fig. 2A) and performance on the speech characteristics subtest of the Verbal Motor Production Assessment for Children, where higher scores indicate better performance. A similar positive correlation was also found within the same cluster, in the right posterior fusiform gyrus/inferior temporal gyrus (BA 37/20; MNI coordinates 42, −22, −24; cluster size = 1 voxel; T = 8.04; P = 0.049; cluster size > 100 voxels at P = 0.001, uncorrected). There were no statistically significant correlations detected between brain activation and age at injury, or time post-injury.

Figure 2

Inferior frontal gyrus region showing (A) positive correlation between functional MRI activation and speech performance in the traumatic brain injury with dysarthria (TBI+) group, and (B) increased grey matter concentration in the traumatic brain injury with dysarthria group relative to the typically developing (TD) control group.

Global brain volumes

The three groups did not differ significantly with respect to global grey matter or white matter concentration on voxel-based morphometry (P > 0.10 in all cases; see Table 3 for average measures).

View this table:
Table 3

Mean (SD) global white and grey matter volumes for the traumatic brain injury group with dysarthria (TBI+, n = 17), traumatic brain injury group without dysarthria (TBI−, n = 15) and typically developing group (TD, n = 17)

White matter volume138 880.2 (17 282.2)141 513.4 (8170.6)148 621.2 (12 066.5)
Grey matter volume209 956.8 (23 279.8)212 063.8 (15 527.2)212 997.2 (15 111.7)

Voxel-based morphometry

Numerous regions of grey matter concentration reduction (Supplementary Fig. 1) were identified in the traumatic brain injury with dysarthria group relative to the typically developing group using voxel-based morphometry analysis. Relevant to the speech execution system, reductions were found bilaterally in the putamen, the right thalamus and also in the left precentral gyrus (likely to correspond to primary motor cortex representations of the lip/larynx and tongue, Takai et al., 2010). A reduction in the right-hemisphere, more ventrally, was identified at a lower threshold (cluster size = 28; peak coordinates = 41, −21, 42; P = 0.002). When compared with the traumatic brain injury without dysarthria group, grey matter reductions were found in two virtually homotopic areas located just ventrally to the hand knob in each hemisphere (Table 4 and Fig. 3). In view of the relationship between speech performance and functional MRI activation in BA45, an exploratory analysis examined regions of increased grey matter in the traumatic brain injury with dysarthria group relative to the typically developing control group. Indeed, increased grey matter was found in a nearby anatomical region (BA 45, pars triangularis, peak MNI coordinates = −44, +23, +10; cluster size = 11 voxels; T = 3.24; P = 0.001; Fig. 2B). Relative to the dysarthric group, the traumatic brain injury group without dysarthria showed grey matter reduction only in the right medial superior frontal gyrus (peak coordinate = 15, 5, 52, T = 3.53, P < 0.001).

Figure 3

Regions of reduced grey matter in the group with traumatic brain injury with dysarthria relative to the traumatic brain injury group without dysarthria controlling for age, gender and intracranial volume. Results are displayed at a threshold of P = 0.005, uncorrected for multiple comparison and overlayed onto the average grey matter map of the 49 participants. Inset: Sagittal and coronal views of the homotopic precentral gyrus clusters. L = left hemisphere; R = right hemisphere. See Table 4 for complete list of regions and coordinates.

View this table:
Table 4

Regions of reduced grey matter in the traumatic brain injury with dysarthria group relative to the traumatic brain injury without dysarthria group within the speech system

Anatomical regionPeak coordinatesCluster sizeUncorrected P-value, cluster levelPeak level TUncorrected P-value, peak level
Right pre/postcentral gyrus, dorsal42, −21, 423110.064.22<0.001
Right superior frontal gyrus, medial15, −4, 58240.603.96<0.001
Right parahippocampal gyrus36, −39, −5370.513.84<0.001
Left precentral gyrus, adjacent to hand knob−36, −15, 52460.463.74<0.001
Left postcentral gyrus, dorsal−27, −40, 42250.593.420.001
Right cingulate gyrus8, −10, 371630.163.420.001
Left supplementary motor area11, −21, 55290.563.380.001
Right amygdala24, −10, −9310.553.200.001
Right parahippocampal gyrus15, −36, 1460.463.100.002
Left superior parietal gyrus−15, −63, 43120.732.980.002
Left parietooccipital fissure−12, −76, 22100.752.890.003
  • Results at a threshold of P = 0.005, uncorrected for multiple comparisons (extent = 10 voxels) using a mask of functional MRI activation (see main text). No voxel survived P = 0.05 family-wise error. See Fig. 3 for illustration.


Here, we provide the first biomarkers of chronic dysarthria in young people who have sustained traumatic brain injury in childhood. In line with our hypotheses, voxel-based morphometry data revealed bilateral grey matter abnormalities in regions of the primary motor cortex corresponding to the representation of the articulators relative to the non-dysarthric group. These morphological abnormalities were associated with significantly increased recruitment of the left sensorimotor cortex. Unlike what was hypothesized; however, no functional changes were observed in the basal ganglia or cerebellum. This could be explained by the fact that these structures were highly abnormal. In addition, recruitment of Broca’s area was positively correlated with better speech outcome in the dysarthric group. There was no evidence of increased activation or ‘reliance’ on the right-hemisphere network for individuals with dysarthria.

Effectiveness of functional reorganization: limited by structural damage?

Increased left-hemisphere sensorimotor activation during speaking in participants with chronic dysarthria relative to typically developing controls was seen here. A possible explanation for this finding is that whilst grey matter damage was bilateral and widespread, lesions to speech motor regions were not of a sufficient degree to require intra- or inter-hemispheric reorganization to other less affected regions, and rather compensation for speech function was supported by increasing recruitment or reliance on the existing left-dominant speech networks. This profile of increased reliance on the left-hemisphere was arguably relatively effective here in that the speech disorder in our traumatic brain injury with dysarthria group improved from a moderate-to-severe profile acutely post-injury to a largely mild presentation. However, in cases of unilateral injury, being able to ‘shift’ function to an unaffected hemisphere may arguably lead to better functional outcomes, as in the adult dysarthric case with a left-sided capsular infarction who demonstrated ‘right shift’ of speech motor function and made full speech recovery just 9 days post-onset (Riecker et al., 2002). Similarly, dysarthria subsequent to unilateral stroke in adults can result in a return to normal speech ‘within weeks’ (in 40% of the group reported by Urban et al., 2006). Whilst future large-scale studies are required to better understand prognostic factors, such as lesion size and location on dysarthria outcome, preliminary evidence suggests functional reorganization may be less likely to lead to full recovery in cases with bilateral structural damage.

Critical regions of the speech execution network after injury in childhood

Comparison between the traumatic brain injury group with dysarthria and typically developing control subjects indicated a generally bilateral profile of grey matter abnormalities in chronic dysarthria. It is important to emphasize that differences were observed between these groups in both structural grey matter and functional MRI in the region immediately ventral to the hand motor cortex representation in the left-hemisphere. This region, reported to correspond to the lips/larynx motor representation (Takai et al., 2010), was morphologically abnormal yet showed functional MRI over-activation in dysarthric participants. Our results from the voxel-based morphometry and functional MRI approaches converge to suggest that the left speech-related sensorimotor cortex may play a crucial role in supporting normal speech production during development.

Grey matter concentration increases were also seen in BA 45 for the traumatic brain injury with dysarthria group relative to control subjects, the same region where functional MRI activation was positively correlated with better speech performance. Given the lack of literature explaining the potential biological bases of concentration increases on voxel-based morphometry, we can only speculate that this finding could be associated with neuronal reorganization or compensation, given that studies of adult stroke recovery of motor function have also shown correlations between increased grey matter concentration and better outcomes (Gauthier et al., 2008). It is compelling that both voxel-based morphometry and functional MRI data highlighted a difference between structure and function of these particular regions in individuals with dysarthria.

Functional reorganization in acquired dysarthria: recruitment of higher-order areas

The left sensorimotor overactivation and positive correlation between speech performance and Broca’s area activation seen here in chronic dysarthria are perhaps not surprising, given the emphasis placed on critical links between ‘higher-’ (i.e. Broca’s area) and ‘lower-level’ motor areas required for successful communication in numerous models (e.g. DIVA, Golfinopoulos et al., 2010; State feedback control, Ventura et al., 2009; Generative model, Price et al., 2011). The finding of activation in Broca’s area (pars triangularis, BA 45) positively correlated with better speech performance is similar to the findings of Sidtis et al. (2010) who noted increased blood flow in Broca’s area associated with faster speech rate (a measure of speech performance) in their longitudinal PET study of adults with hereditary ataxia and dysarthria. These authors suggested that this activation profile may have been compensatory, as speech production had declined somewhat, but not markedly, since initial evaluation. Interestingly, in the recovered dysarthric patient of Riecker et al. (2002), Broca’s area did not present as a region of significant difference between the functional MRI scans conducted at Days 4 and 34 post insult. Taken together, these data on individuals with persistent or recovered dysarthria could imply that Broca’s area is only critically involved as a ‘higher-level’ motor output region in the case of persistent dysarthria, potentially acting to support or compensate for speech dysfunction into the longer-term. Interestingly, Broca’s area has also been proposed as a compensatory region (non-specific to speech or language production) in the motor field, where overactivation of BA 44/45 is seen in elderly versus young participants during fine motor tasks (e.g. Loibl et al., 2011; Heuninckx et al., 2008). Whilst preliminary evidence suggests an association between better speech performance and Broca’s activation in dysarthric speakers (as seen here and in Sidtis et al., 2010) implying that an intact Broca’s area may be important for better outcome, future studies must delineate whether this is truly a form of active compensation or a pathological response to injury affecting Broca’s area’s existing role in speech motor control. It is, however, unlikely that the compensatory mechanism seen on functional MRI is driven by language recovery in the traumatic brain injury group with dysarthria. First, our functional MRI study used real words for both the baseline and task conditions, being designed to control for language-related processes. We are confident that activation patterns are predominantly related to the motoric aspects of speech, as evidenced by the lack of Broca’s activation in the control groups. Secondly, the two traumatic brain injury groups did not perform differently on the standardized language test.

Neuroplasticity and prognostic markers for speech: clinical implications and future directions

A recent view supports that the range of behavioural outcomes following early brain insult represent extremes along a ‘recovery continuum’ dependent on injury factors (severity, nature, age) and environmental influences (family, socio-demographic, interventions) (Anderson et al., 2011). No correlations between functional MRI activation and age at injury or time post-injury were found here, possibly because of the limited sample size, or perhaps grey matter abnormalities are less likely to be mediated by these factors than white matter damage. Future acute and longitudinal functional MRI studies exploring reorganization profiles for speech in recovered versus persistent cases along a spectrum of ages in a larger sample including a range of time post-injury, injury severity and dysarthria severity levels may provide greater understanding of plasticity and prognosis for dysarthria in children with brain injury. The wide age range, broad time post-injury and relatively restricted range of dysarthria severity present in this otherwise carefully selected sample restricted the ability to further explore important prognostic variables here.

Another area in critical need for further investigation is the interaction between behavioural speech treatment and functional reorganization. To our knowledge, no data have been reported on cortical plasticity associated with intensive therapy for motor speech disorder. Children in the present study received speech and language pathology services of assessment and monitoring only, in the absence of any intensive or formalized treatment, in line with care in other international centres (Morgan and Skeat, 2011). Whilst no treatments for acquired dysarthria are currently underpinned by high-level evidence (Morgan and Vogel, 2008), encouraging case-based treatment studies exist in the child (Murdoch et al., 1999; Morgan et al., 2007) and adult traumatic brain injury literature (Wenke et al., 2008), including a case of ‘restoration of intelligible speech’ in a dysarthric adult, >13 years after the original injury (Workinger and Netsell, 1992). Such data suggest that there is propensity for improvement in speech function many years post-injury.

Conclusions and clinical implications

This study has provided empirically based biomarkers for chronic dysarthria associated with acquired brain injury. Specifically that bilateral grey matter structural abnormalities in key motor speech regions may place individuals at risk for long-term persistent speech deficits, inline with the findings of a systematic review in this field (Liégeois and Morgan, 2012). It seems that the profile of bilateral damage triggered attempts at functional compensation, involving overactivation of key left-hemisphere motor and language regions. Importantly, data suggested that this pattern of reorganization was relatively effective, considering the generally severe acute dysarthria presentation of the group and the typically mild chronic speech deficits into the long-term. In relation to clinical practice, overall, our results suggest that (i) individuals post-traumatic brain injury with one intact sensorimotor cortex may have a better prognosis than those with bilateral damage; and (ii) BA 45 may be a suitable candidate for stimulation studies aimed at improving speech outcome in those with bilateral damage, who are otherwise likely at risk of lifelong dysarthria.


This study was supported by grant D131 from the Transport Accident Commission: Victorian Neurotrauma Initiative, awarded to A.T.M., F.J.L, A.C., and J-D.T.; and grants 607315 and 1023043 from the National Health and Medical Research Council awarded to A.T.M.

Supplementary material

Supplementary material is available at Brain online.

Clinical Evaluation of Language Fundamentals


View Abstract