Melody-based treatments for patients with aphasia rely on the notion of preserved musical abilities in the RH, following left hemisphere damage. However, despite evidence for their effectiveness, the role of the RH is still an open question. We measured changes in resting-state functional connectivity following melody-based intervention, to identify lateralization of treatment-related changes. A patient with aphasia due to left frontal and temporal hemorrhages following traumatic brain injuries (TBI) more than three years earlier received 48 sessions of melody-based intervention. Behavioral measures improved and were maintained at the 8-week posttreatment follow-up. Resting-state fMRI data collected before and after treatment showed an increase in connectivity between motor speech control areas (bilateral supplementary motor areas and insulae) and RH language areas (inferior frontal gyrus pars triangularis and pars opercularis). This change, which was specific for the RH, was greater than changes in a baseline interval measured before treatment. No changes in RH connectivity were found in a matched control TBI patient scanned at the same intervals. These results are compatible with a compensatory role for RH language areas following melody-based intervention. They further suggest that this therapy intervenes at the level of the interface between language areas and speech motor control areas necessary for language production.
The goal of the current study was to examine changes in functional connectivity within the right and left hemispheres following melody-based treatment in a patient with aphasia. The long-held notion that recovery from aphasia following LH damage involves compensatory recruitment of the RH [
The use of melody in aphasia treatment is based on the observation that singing and the production of melodic speech are often intact, even when standard speech is impaired in patients with nonfluent aphasia [
Despite the common use of melody-based therapies, and the positive evidence for their effectiveness, the underlying mechanisms and the role of the melodic component in language recovery are still unresolved [
The current evidence, from functional imaging studies, for the involvement of the RH in melody-based treatment is mixed. An early PET study ([
Structural-imaging studies show more consistent evidence for the involvement of the RH in melody-based therapy. 11 patients undergoing MIT showed an increase in RH white matter volume and a correlation of behavioral improvement with changes in the right IFG pars opercularis [
Finally, two transcranial brain stimulation studies showed that excitatory stimulation of the right posterior IFG has improved the effects of MIT in some of the participants [
The current study aims to examine the underlying brain mechanisms associated with melody-based treatment by looking at changes in the functional connectivity in the language network. In contrast to the ambiguous interpretation of local activation changes (in which a decrease in activation in the RH may reflect less reliance on the RH or more efficient processing in the RH), functional connectivity with RH regions is more clearly associated with increased involvement of the RH. We examined the changes in resting-state connectivity in the language network of a patient with chronic nonfluent aphasia associated with melody-based treatment. Resting-state measures do not depend on the patient’s level of language performance, while still depicting connectivity in the language network [
JV was a 48-year-old right-handed female at the time of injury, with 16 years of education. She is a native speaker of Tagalog, and was a fluent speaker of English as a second language. She sustained a moderate-severe traumatic brain injury (TBI) secondary to a fall from a ladder onto a concrete floor. A CT scan performed on the day of the injury indicated left frontal and temporal subdural hematomas (Glasgow Coma Scale (GCS) score not available), which were evacuated in an urgent craniotomy. JV was in a coma for five to six weeks. Unfortunately, she sustained a second TBI three months later, as a result of another fall while in hospital, at which time her language symptoms worsened. MRI at the time of the second TBI showed evidence of a new subarachnoid hemorrhage in the left medial temporal sulcus, in addition to underlying encephalomalacia in the left frontal and left temporal lobes associated with the initial hematomas (see Figure
MR T1-weighted image of the patient’s brain in (a) axial, (b) coronal, and (c) sagittal views. The patient presents with an extensive area of encephalomalacia within the left cerebral hemisphere involving the temporal and frontoparietal lobes with volume loss.
JV’s language assessment 36 months postinjury.
Language assessment | Pretreatment | Posttreatment | ||
---|---|---|---|---|
Raw score | Percentile | Raw score | Percentile | |
BNT—number of spontaneously given correct responses | 2/15 | 30 | 0/15 | 30 |
BNT—number of correct responses following phonemic cue | 4/11 | NA | 4/13 | NA |
BNT—number of correct choices | 3/9 | NA | 6/11 | NA |
I.A. simple social responses | 6/7 | 50 | 6/7 | 50 |
Aphasia Severity Rating Scale | 1/5 | 40 | 2/5 | 50 |
II.A. word comprehension | 8/16 | <10 | 10/16 | 10 |
II.B. commands | 8/10 | 40 | 8/10 | 40 |
II.C. complex ideational material | 3/6 | 30 | 4/6 | 50 |
III.B. automatized sequences | 1/4 | 10 | 1/4 | 10 |
III.B. repetition single words | 4/5 | 60 | 4/5 | 60 |
III.B. repetition sentences | 1/2 | 60 | 1/2 | 60 |
III.C. responsive naming | 0/10 | 10 | 4/10 | 30 |
III.C. naming—screening of special categories | 7/12 | 20 | 7/12 | 20 |
IV.C. oral word reading | 3/15 | <20 | 3/15 | <20 |
IV.C. oral sentence reading | 0/5 | 30 | 0/5 | 30 |
IV.C. sentence comprehension | 0/3 | 10 | 0/3 | 10 |
IV.D. reading comprehension—sentences and paragraphs | 2/4 | 40 | 1/4 | 10 |
40/52 | NA | DNT | ||
22/24 | NA | 22/24 | NA |
BDAE = Boston Diagnostic Aphasia Examination [
JV neuropsychological assessment 25 months postinjury.
Domain/test | Raw score | Standard score | Classification |
---|---|---|---|
Grip strength (dom) | 25 kg | Low average | |
Grip strength (non-dom) | 22.5 kg | Average | |
Grooved pegboard (dom) | 119 sec (0 drops) | Severely impaired | |
Grooved pegboard (non-dom) | 80 sec (0 drops) | Low average | |
Visual span forwards | 8 | Average | |
Symbol Digit Modalities Test—W | 42 items (0 errors) | Mildly impaired | |
Trail Making Test A | 40 sec | Mildly impaired | |
Trail Making Test B | 179 sec (2 errors) | Moderately impaired | |
MAE Token Test | 9/44 | Very defective | |
Peabody Picture Vocabulary Test-III | 126 | 1st %ile | Impaired (age equivalent = 9 years, 9 months) |
Visual Form Discrimination | 32/32 | Intact | |
WAIS—Block Design | 24/68 | Borderline impaired | |
RVDLT—copy | 32/36 | 6–10%ile | Mildly impaired |
RVDLT—time to complete (copy) | 361 sec | 2-5th %ile | Mildly impaired |
RVDLT—copy organizational quality | 1/5 | Extremely piecemeal; drawn rotated 90 degrees | |
RVDLT—total trials 1–5 | 46 Figures (26 intrusions) | Average | |
RVDLT—immediate recall | 14.5/36 | Mildly impaired | |
RVDLT—highest number of figures recalled | 12/15 | ||
RVDLT—delayed recall | 11/15 (4 intrusions) | ||
RVDLT—delayed recognition | 10/15 (2 false alarms) | Severely impaired | |
BVMT Total Immediate Recall | 17/36 | Mildly impaired | |
BVMT Total—Delayed Recall | 8/12 | Average | |
BVMT Recognition | 6/6 (0 false alarms) | >16th %ile | Intact |
Visual span backwards | 7 | Average | |
WAIS—matrix reasoning | 11/26 | Average | |
WCST—total administered | 128 (full WCST) | ||
WCST—errors | 40 | Mildly impaired | |
WCST—perseverative responses | 27 | Mildly impaired | |
WCST—perseverative errors | 23 | Mildly impaired | |
WCST—nonperseverative errors | 17 | Mildly impaired | |
WCST—conceptual level responses | 45 | Mildly impaired | |
WCST—categories | 4 | 11-16th %ile | Mildly impaired |
Trials to complete 1st category | 12 | >16th%ile | Intact |
MAE Token Test = Multilingual Aphasia Examination Token Test; WAIS = Wechsler Adult Intelligence Scale; RVDLT = Rey Visual Design Learning Test; BVMT = Brief Visuospatial Memory Test; WCST = Wisconsin Card Sorting Test; W = written.
The control TBI patient, GB, was a right-handed female, native speaker of English, with 13 years of education who was 54 years of age at the time of injury. She had sustained a moderate to severe TBI secondary to a motor vehicle accident. Her GCS at the scene was 5 and declined to 3 at the time of admission to the emergency department. A CT scan performed on the day of injury indicated a left temporal subarachnoid hemorrhage as well as an intraventricular hemorrhage. GB was in a coma for approximately 1 week following the accident. Neuropsychological assessment at 28 months postinjury did not indicate any persistent language deficits. Although the control patient is a native English speaker, while the treated patient speaks English as a second language, we do not expect this to affect the results because we are not comparing between them on language performance or on a language task-related functional imaging. Ethics approval was granted for the study, and both participants had signed an informed consent.
Thirty-six months after her second injury, as part of the current study, JV started receiving a melody-based treatment, which was a modified version of MIT [
Each target phrase, and its associated unique melody, was presented following the standard MIT hierarchy (e.g., [
Melody served as the primary cue to support word production within the target phrase, and the unique melody was played up to five times, as required, allowing the participant to gradually fill in the melody with the target words. Initially, the participant required maximal melodic cueing (i.e., 5 cues), but this need decreased (i.e., to 1-2 cues) over the course of treatment. Rhythm served as the secondary cue, whereby the rhythm of the target phrase was tapped on her left arm in order to help the participant overcome hesitations or sustained pauses during a phrase. Additional cues included modelling of oral-motor placement and providing the first word in the target phrase. The need for these cues, however, significantly reduced as treatment progressed. In addition to the hierarchical progression of cues, the treatment was designed to enhance generalization, by gradually decreasing the structure of the target phrases, thereby increasing the linguistic difficulty. Namely, the treatment involved progressing through 5 steps: steps 1 to 3 primarily involved repetition; step 4, sentence completion; and step 5, phrase production (in response to target questions). Novel responses were encouraged in steps 4 and 5.
The primary goal of steps 1 to 3 was to use melodic and rhythmic cues to encourage word retrieval and fluency in sentences of increasing length and complexity. The participant was required to produce target phrases with and without melodic intonation. The primary goal of step 4 (sentence completion task) was to encourage generalization beyond trained phrases, by asking the participant to generate novel words at the end of each rehearsed target phrase. In step 5, the primary goal of treatment was for JV to produce fully self-generated phrases, using two tones to produce nonrehearsed responses to target questions. Untreated items were also created and matched to all treatment stimuli in terms of number of syllables, syntactic complexity, and relevance to the participant’s life and interests. These items were not treated and served for testing before, immediately after, and eight weeks following treatment (see outcome measures).
Treatment stimuli were created based on an “interest” inventory completed by JV and her family, whereby the participant listed her primary hobbies and interests, important members of her family and friend groups, details from her past, and her speech-related goals. Treatment stimuli were phrases on a continuum of difficulty in terms of syllable length and syntactic complexity. Shorter, simpler (e.g., imperative and wh-question) phrases were used initially, followed by longer, declarative present- and past-tense phrases (based on HELPSS hierarchy, [
JV started receiving treatment 36 months after her second injury and received treatment three days a week for 16 weeks, for a total of 48 sessions. Each session lasted approximately 30 minutes. At the beginning of each session, three previously learned phrases were rehearsed, to encourage maintenance and continued practice of all treated items. Progression through steps 1 to 5 occurred if either (a) the participant produced 80% of the target phrases with 80% accuracy on two consecutive sessions or (b) the participant completed 9 sessions of therapy within a single step. Target accuracy was measured as the proportion of syllables correctly produced within the target phrase (for steps 1 to 3) and the appropriateness of the responses given to sentence completion and question probes (for steps 4 and 5). Each melody and phrase was notated and emailed to the participant, who was instructed to practice the target phrases between sessions for 30 minutes, three times per week. The participant was also required to sing personally significant songs for ten minutes a day, five days a week, in an attempt to reinforce word retrieval through familiar melody and word associations. These activities were monitored by the patient using a “homework log.” These logs indicated very good compliance of the patient with the homework.
Both treated and untreated items were administered pretherapy, immediately posttherapy, and at eight weeks posttherapy by a registered speech-language pathologist who was not involved in the treatment (coauthor Tijana Simic). The primary outcome measure for steps 1 to 3 was the ability to repeat target phrases accurately (measured as the
Whole head MR scans were acquired on a General Electric (GE) Signa-Echospeed 1.5 Tesla high-definition scanner, located at Toronto General Hospital—University Health Network, using an eight-channel head coil. The high-resolution 1 mm isotropic T1-weighted, three-dimensional radio-frequency spoiled-gradient recalled-echo (SPGR) images were acquired in the axial plane utilizing a 25 cm field of view (FOV) (TR/TE/TI = 12/5/300 ms), flip angle = 20°, slice thickness = 1 mm no gap, 160 slices, and matrix 256 × 256. Resting-state BOLD fMRI data were acquired with the following imaging parameters: TR/TE = 2000/40 ms, flip angle = 85°, FOV = 22, slice thickness 5 mm with no gap, 32 slices with 4800 images, and matrix 64 × 64. Resting-state data were collected at three time points. For JV, this was at 25 months, 35 months, and 39 months after the second injury, with treatment occurring between 36 and 39 months postinjury. Resting-state data for the control patient GB was collected at 28 months, 32 months, and 36 months postinjury, with no speech and language treatment. Additional sequences include DTI and axial fast spin-echo PD/T2-weighted images, which are not presented here. All sequences were obtained with a 22 cm FOV. The entire scanning session lasted approximately 55 min.
The preprocessing of resting-state data was performed using SPM12 (Wellcome Department of Imaging Neuroscience, London, UK;
Functional connectivity analysis was performed using a region of interest (ROI) to ROI approach within CONN’s functional connectivity toolbox (Whitfield-Gabrieli and Nieto-Castanon 2012;
Sources of physiological noise (based on white matter and cerebrospinal fluid segmentation) and movement covariates (motion correction and scrubbing using Artifact Detection Tools; ART
Similar to the approach taken by Sandberg et al. [
The proportions of correctly produced syllables before and immediately posttreatment, for both treated and untreated phrases, are presented in Table
Treatment outcome measures for JV.
Mean % correctly produced syllables (SD) | ||||
Treated phrases ( |
Untreated phrases ( | |||
Pre- | 71.34 (28.79) | 64.60 (32.59) | ||
Post- | 96.78 (8.05) | 61.55 (33.15) | ||
8-week | 93.33 (13.30) | 74.11 (30.88) | ||
Treated post | Treated 8 weeks | Untreated post | ||
Treated pre | ||||
Treated post |
Mean number of syllables produced (SD) | ||||
Treated phrases ( |
Untreated phrases ( | |||
Pre- | 1.44 (1.88) | 1.78 (1.20) | ||
Post- | 4.22 (2.44) | 4.78 (2.28) | ||
8-week | 2.78 (2.04) | 3.78 (1.72) | ||
Treated post | Treated 8 weeks | Untreated post | Untreated 8 weeks | |
Treated pre | ||||
Treated post | ||||
Untreated pre |
∧Approaching significance of Alpha = 0.01 (Bonferroni correction).
Mean number of syllables produced (SD) | ||||
Treated phrases ( |
Untreated phrases ( | |||
Pre- | 0.50 (1.08) | 0.40 (0.97) | ||
Post- | 6.20 (2.90) | 2.10 (3.03) | ||
8-week | 4.60 (3.06) | 2.40 (3.13) | ||
Treated post | Treated 8 weeks | Untreated post | Untreated 8 weeks | |
Treated pre | ||||
Treated post | ||||
Untreated pre |
Performance on treated and untreated phrases at pre-, post- and 8-week follow-up tests. Sentence repetition (steps 1–3; (a)), sentence completion (step 4, (b)), and probe questions (step 5, (c)) are presented. Statistical comparisons with Wilcoxon signed-rank test are indicated at the bottom of each panel.
The Wilcoxon signed-rank test for related samples was used to assess the treatment effect on syllable production when repeating sentences of increasing length and complexity (steps 1 to 3) and compared to performance on untreated stimuli sets. Three comparisons were made with these data; using the Bonferroni correction, alpha was set at 0.017. JV’s ability to repeat syllables within treated phrases significantly improved pre- to posttreatment (
In the sentence completion task (step 4), five comparisons were made using the Wilcoxon signed-rank test for related samples; thus, alpha was set at 0.01 (Bonferroni correction). The difference between pre- and posttreatment tests in the number of syllables produced by JV when given a sentence completion cue approached, but did not reach, significance (
Finally, JV’s ability to answer questions (step 5) was also assessed using the related-sample Wilcoxon signed-rank test; five comparisons were again made here; thus, alpha was set at 0.01. JV’s ability to answer treated questions significantly improved pre- to posttreatment (
JV was reassessed on various language measures following treatment (Table
Changes in resting-state connectivity during the treatment interval in the treated patient and the equivalent interval in the control patient are presented in Figure
Changes in resting-state connectivity during the treatment interval in the treated patient JV (a) and control patient GB (b). Increase: green; decrease: red. Values represent differences (T3 − T2) in the semipartial correlation. Arrows point to the target region in the calculation of semipartial correlations. Only changes in T3 − T2 which were significantly greater than changes during the baseline period (T2 − T1) are shown. Significance is determined with FDR correction for 66 correlations with
This study examined the effect of melody-based treatment on a chronic patient with moderate to severe aphasia due to extensive left frontotemporal lesions following two temporally proximal brain injuries three years earlier. The patient’s performance on treated and untreated phrases was examined before, immediately after, and eight weeks following treatment. Resting-state connectivity was examined at three time points: T1—baseline (25 months postinjury), T2—pretreatment (36 months postinjury), and T3—posttreatment (39 months postinjury). This was compared to a control patient, who did not receive treatment during the same time period. We expected that improvement in language performance following treatment would be associated with an increase in functional connectivity between right frontal homologues of language areas and regions involved in motor speech control.
Behavioral measures of performance on treated phrases showed improvement in the repetition of sentences of varying lengths and complexity. Likewise, JV’s ability to answer question probes improved following treatment, with more appropriate and longer answers; these improvements were maintained at the eight-week follow-up, and similar improvements were not seen in the untreated phrases and/or question probes. Performance on the sentence completion task improved numerically during treatment, but improvement was only marginally significant. Moreover, visual inspection of the data indicated similar levels of improvement in both treated and untreated phrases in the sentence completion task, with no difference between treated and untreated phrases posttreatment. These results, together with improved performance on the responsive naming subtest of the BDAE (a task similar to sentence completion), suggest that treatment effects generalized to untreated stimuli in the sentence completion task (step 4). However, the small number of phrases in step 4 may have underpowered the analysis and masked this effect. Alternatively, the marginal treatment effect for treated phrases in the sentence completion task may suggest that relative to the more open-ended nature of the probe questions, the constrained sentence completion task may have proven especially difficult.
Resting-state connectivity measures for the treated patient showed increases in connectivity between right frontal language areas (R.Tri and R.Operc) and regions involved in speech motor control (bilateral SMA and L.Insula) during the treatment period. There were also significant increases in connectivity within the right frontal language areas (R.Orb–R.Operc), compared to the baseline period. Moreover, these changes were specific to the RH and to the treated patient. In contrast, the control patient, who did not receive treatment, showed increases in connectivity between the left frontal language (L.Operc) and speech motor control area (L.SMA) and within the left frontal language areas (L.Orb–L.Operc) during the same time period.
The behavioral improvements observed in the repetition of treated phrases in the current study are consistent with previous research showing the effectiveness of melody-based treatments for patients with nonfluent aphasia [
In addition to improvement on treated phrases, our findings also show some evidence for generalization of treatment effects to untreated phrases with the sentence completion task (step 4) and a nonsignificant trend for improvement on the untreated question (step 5). These tasks, which are not typically included in melody-based treatment protocols, such as the standard MIT [
Analyses of resting-state connectivity, which show an increase in connectivity within the right frontal language areas, and between these areas and regions involved in speech motor control during treatment, are consistent both with our hypothesis and with the underlying assumptions of melody-based treatment approaches, namely, that musical abilities typically associated with RH areas [
Our findings are consistent with a number of functional imaging studies, showing a greater increase in activation in the right compared to the left frontal cortex following MIT ([
By using resting-state data, the current results further demonstrate that these treatment effects are not solely task dependent. Treatment-related changes in resting-state connectivity were previously shown with other types of aphasia therapies [
In contrast to functional imaging measures, structural-imaging results indicate a more stable change, which is not task dependent. The evidence, for the involvement of the RH in melody-based treatment in the current study, is broadly consistent with findings from DTI studies showing an increase in the number of fibers in the right arcuate fasciculus in patients following MIT [
A significant limitation of this study is its single-case design, which does not allow for the correlation of changes seen in brain connectivity and language behaviors. Nevertheless, we believe that the comparison to a baseline period, as well as testing of a control patient enabled us to make conclusions about the specificity of the results to the treatment administered. Lastly, although the measurable focal lesions of this patient were in the left hemisphere and secondary to hemorrhages, the probable presence of diffuse axonal injury secondary to traumatic injury does not allow us to rule out the presence of RH damage; indeed, impairments on measures of visuospatial measures are likely attributable to such damage. Arguably, the RH changes and response to treatment may have been more robust in the absence of these putative changes.
Our results show the benefit of melody-based treatment for a patient with moderate-severe nonfluent aphasia, more than three years following traumatic brain injury with focal lesions in the left frontotemporal areas due to hemorrhages. The patient’s ability to repeat sentences and answer question probes improved on treated items, and this treatment effect was maintained eight weeks following treatment. Improvement in sentence completion was marginally significant and similar for treated and untreated phrases, suggesting generalization. The results of the resting-state imaging suggest that the effect of melody-based treatment is right lateralized. This effect was stable and found at the level of the functional network even during rest and not only in localized task-dependent activation. Our results further show that the effects are treatment specific and are not shown in a patient that did not receive therapy during the relevant period. Beyond the lateralization of treatment, our results further show that melody-based treatment affects the interface between language retrieval and motor speech control and articulation. A treatment affecting this level of processing may be a good basis for generalization to untreated stimuli. Further research is needed in a larger sample of patients with discrete LH focal lesions.
The authors declare that they have no conflicts of interest.
Tables 4 and 5 in the supplementary material show changes in resting-state connectivity in the treated and control patients separately for the treatment and baseline periods. These were used to compute the differences in Figure