Cognitive efficacy and neural mechanisms of music‐based neurological rehabilitation for traumatic brain injury

Abstract Traumatic brain injury (TBI) causes lifelong cognitive deficits, most often in executive function (EF). Both musical training and music‐based rehabilitation have been shown to enhance EF and neuroplasticity. Thus far, however, there is little evidence for the potential rehabilitative effects of music for TBI. Here, we review the core findings from our recent cross‐over randomized controlled trial in which a 10‐week music‐based neurological rehabilitation (MBNR) protocol was administered to 40 patients with moderate‐to‐severe TBI. Neuropsychological testing and structural/functional magnetic resonance imaging were collected at three time points (baseline, 3 months, and 6 months); one group received the MBNR between time points 1 and 2, while a second group received it between time points 2 and 3. We found that both general EF and set shifting improved after the intervention, and this effect was maintained long term. Morphometric analyses revealed therapy‐induced gray matter volume changes most consistently in the right inferior frontal gyrus, changes that correlated with better outcomes in set shifting. Finally, we found changes in the between‐ and within‐network functional connectivity of large‐scale resting‐state networks after MBNR, which also correlated with measures of EF. Taken together, the data provide evidence for concluding that MBNR improves EF in TBI; also, the data show that morphometric and resting‐state functional connectivity are sensitive markers with which to monitor the neuroplasticity induced by the MBNR intervention.


INTRODUCTION
According to the U.S. National Institute of Neurological Disorders and Stroke, traumatic brain injury (TBI) is "an alteration in brain function, or other evidence of brain pathology caused by an external force." 1 Most of TBI cases are closed head injury and vary in severity from mild (including concussion) to moderate and severe, though this classification is fairly crude and generally based on a unidimensional assessment of level of consciousness 2 or post-traumatic amnesia. 3 With estimate for annual TBI approaches 2.5 million new cases. [8][9][10] Owing in part to the demographic aging of the population in high-income countries, incidence of TBI in the elderly, mostly due to falls, is increasing. 11 Across all ages, TBI accounts for 30-40% of all injury-related deaths. 12 These figures not only highlight the magnitude of the problem but also the need to provide appropriate prevention and treatment strategies that are suitable for the subpopulations at higher risk of TBI.
People who suffer TBI show a wide spectrum of symptoms, ranging from physical, behavioral, emotional, to cognitive. [13][14][15] This wide variation in the clinical manifestations of TBI is likely due to the complexity of the brain's organization, as well as to the patterns and extent of damage caused by external forces leading to TBI. Cognitively, the majority of deficits affect high-level cognitive functions, such as attention, memory, communication, and executive function (EF). 16 EF is a broad umbrella term referring to high-level cognitive processes that enable individuals to regulate their thoughts and actions during goal-directed behavior. 17 Given their clinical importance and profound effect on the daily lives of the patients, executive dysfunctions are often viewed as a core symptom of TBI. While there is no consensus on the exact definition of EF, various cognitive processes, such as set shifting (switching from one task to another), inhibition (avoiding a prepotent response), and updating (continuously updating the contents of working memory), have been proposed as key components of EF. 18 A brain hallmark of TBI, compared with other neurological disorders (e.g., stroke), is that the underlying brain injury is difficult to measure clinically. Computed tomography and standard structural magnetic resonance imaging (MRI) acquired in the clinical setting are most sensitive for measuring focal injuries in the form of microbleeds 19 or hematomas, 20 though these are not always present in individuals with TBI and, if not found in given cases, may lead to the underestimation of the severity of the injury. In addition to focal brain damage, rapid acceleration and deceleration forces at the time of brain injury damage the axonal membrane, which can disrupt axonal transport; such socalled diffuse axonal injury (DAI) 21,22 is a key and common hallmark in the pathophysiology of all TBI severities. As mentioned, the detection of DAI poses challenges due to the lack of sensitive and low spatial resolution of current neuroimaging technologies-though some diffusion metrics, such as a neurite density index, could represent a more specific measure of the neurodegeneration and demyelinization caused by DAI. 22 Given the diffuse structural disconnection triggered by DAI, it is relatively straightforward to predict the coexistence of functional connectivity (FC) abnormalities in large-scale networks. 23 Despite the complexity of the effects of TBI on network function, there are numerous studies showing the consistent patterns of functional network disruption, 23 especially affecting the default mode, 24 the salience, and the executive control networks, 25 which suggests that the understanding and characterization of this neurological disorder may be greatly advanced from the study of static and dynamic brain interactions at the global level with the potential to be used as diagnostic or prognostic markers and readily implemented with current resting-state functional magnetic resonance imaging.
Although TBI is usually limited to a single event, it has been shown that it confers an increased long-term risk for cognitive impairment and dementia, 26,27 stroke, 28,29 parkinsonism, 30,31 and epilepsy, 32 as well as being associated with a higher long-term mortality rate compared with rates for the general population.
Taken together, these facts illustrate that TBI can have long-term consequences for the patients, and that despite some degree of recovery, patients might have to continuously adapt and cope with any remaining symptoms. In this regard, the optimization of rehabilitation methods that are cost-effective and can be tailored to the individual needs of patients may provide great benefits in terms of both reducing the economic impact to society and health care systems and improving clinical outcomes.

MUSIC THERAPY FOR TBI REHABILITATION
The diversity and complexity of the consequences of TBI are best addressed with a comprehensive, holistic approach to rehabilitation delivered by a specialized multidisciplinary team, in close liaison with the patient and family or caregivers (the patient-centered care approach). 33 As recently noted, 34 music therapy responds well to this framework and can take the shape of an active or passive intervention, a decision that must be made based on the primary clinical outcomes targeted by the intervention, as well as the severity of the symptoms presented by the patients. This becomes markedly clear when the stage of the recovery is considered. For instance, when a music therapy is provided to TBI patients with a minimally conscious or comatose state, a passive music listening intervention may be the most suitable strategy and help to regulate arousal and moderate physiological parameters. 35 Music listening has been proven to ameliorate cognitive symptoms-including attention and verbal memory-in stroke patients at the acute stage. 36 One possibility is that the neurobiological mechanisms underlying these effects may result from the widespread engagement of brain regions during music listening, which, in turn, may induce experience-dependent plasticity to support lost or impaired brain function. Another important benefit derived from music listening is self-regulation of emotions, as it recruits the reward system and an extensive number of limbic and paralimbic regions. 37,38 It then follows that music can be used to regulate mood alterations after a brain injury as has been shown in stroke patients. 36 In addition, since music listening can act as a reinforcer on its own, 37 it may contribute to the adherence to the therapy as opposed to traditional rehabilitation methods.
On the other hand, active music making by instrument playing or singing leads to further activation of the brain, requiring the precise timing, sequencing, and spatial organization of actions coordinated by the cerebellum, basal ganglia, premotor and supplementary motor areas, and prefrontal cortical regions. 39 Importantly, behavioral and neuroimaging studies in healthy individuals have revealed that musical training enhances EF and the recruitment of the cognitive control network, [40][41][42][43][44][45][46][47][48] raising the question of whether music-based interventions could have similar positive effects on the executive dysfunction experienced by TBI patients. In spite of the evidence suggestive of music therapy being suitable for TBI rehabilitation, little clinical research has been conducted thus far.
A review of the literature reveals that only a few studies have investigated the clinical efficacy and neural correlates of music therapy after brain injury. In one of the earliest, Thaut and colleagues 49 carried out a pre-and post-test exploratory study with a treatment (n = 31) and control group (n = 23) of patients, with the latter receiving no intervention.
The music therapy consisted of four 30-min sessions each one targeting a different domain: attention, memory, EF, and emotional adjustment.
The pre-post comparison suggested that treated patients improved in the mental flexibility aspect of EF, in addition to better self-efficacy and reduced anxiety and depression.
The next effort in the literature is the small-scale feasibility study conducted by Lynch and LaGlasse. 50 The patients (n = 14) were randomly assigned to a music therapy group, singing group, and control group. The primary outcome of this musical executive function training More recently, Vik et al. 51,52 carried out a piano training intervention in a group of seven mild TBI (mTBI) patients who received two 30-min one-on-one piano lessons per week for 8 weeks and, in addition, had to practice at home for a minimum of 15 min a day. The piano-training protocol involved reading musical notation of 28 pieces for beginners. The study design also comprised two control groups of healthy participants, one with the music intervention (n = 11) and one without (n = 12). The results from the neuropsychological assessment indicated that verbal learning improved in the patient and control group with the music intervention. Furthermore, the authors found an increased activation of the right orbitofrontal (OFC) cortex in a tonicdominant-tonic task in the mTBI patients after the intervention. Using a liberal threshold, seven regions of interest derived from this analysis were used as nodes in a spectral dynamic causal model. 52 When compared to the control group that received the intervention, the mTBI patients showed increased intrahemispheric FC between the left OFC and reduced connectivity between the right anterior OFC and the left posterior OFC. The results from these analyses need to be replicated in a larger sample of mTBI patients but, as noted by the authors, might suggest that the music training induced plastic changes in the activity and connectivity patterns of the OFC, a region in close proximity to bony protrusions that is particularly vulnerable to trauma-induced rotational acceleration of the brain. However promising, these findings must be taken with caution, as most of them include important limitations, such as small sample sizes, lack of a high-quality randomized controlled trial (RCT), heterogeneity in the brain injury mechanism, and no brain imaging measures (except for Vik et al.'s work).
To illustrate this, Thaut et al. 49 included not only patients with TBI but also some with stroke, toxic exposure, seizures, or brain tumor. In addition, there were differences in the number and severity of brain injuries between the control and intervention groups. Likewise, Lynch and LaGasse 50 included one patient with stroke, the severity of which was not reported. Another important limitation concerns the outcome measures used in the aforementioned studies, which were brief and did not cover different domains of EF and focused attention systematically.
Thus, there is a serious need for large RCT with a music-based intervention that can provide robust translatable evidence for TBI rehabilitation.

A LARGE-SCALE RCT USING MUSIC THERAPY FOR TBI REHABILITATION
Given the nature and limitations of the prior studies for the efficacy of music-based neurological rehabilitation (MBNR) in TBI, we sought to address some of them with the first-ever large-scale RCT in moderate-to-severe TBI. [53][54][55] In this single-blinded RCT, we used a cross-over design with n = 40 TBI patients randomly allocated to two groups, AB and BA. The randomization was stratified for lesion laterality (left/right/bilateral) using an online random number generator. During the 3 months between the first two time points for data collection, the AB group (n = 20) received the 10-week music intervention in addition to standard care, and vice versa for the second 3-month period spanning data collection. The standard care included mostly individual therapies, such as physiotherapy, occupational therapy, neuropsychological rehabilitation, and speech therapy, provided by the health care system. Both groups received a similar amount of other rehabilitation (e.g., physical therapy or neuropsychological rehabilitation) at any time point (for more details, see Table 2 in our previous publication 53 ).
For the purpose of the present study, we administered an MBNR intervention specifically designed to target the needs of TBI patients.
The intervention model was adapted from two existing music therapy methods: functionally oriented music therapy (https://www. fmtmetoden.se/fmtsiteng/index.html) and music-supported training, which have been both applied in stroke rehabilitation. 56,57 Our approach was centered on supporting neurological and cognitive recovery after TBI, as a distinction from other types of music therapy that might focus more, for instance, on emotional aspects. The primary outcome of this RCT was the rehabilitation of cognitive deficits, especially EF, attention, and working memory. Secondary goals were to enhance mood, emotional adjustment, and upper extremity motor function.
The intervention consisted of 20 individual therapy sessions (2 times per week, 60 min per session) held by a trained music therapist. The length, frequency, and total duration (10 weeks) of the intervention were selected to balance training intensity with maintaining participant motivation and endurance, as well as logistic practicalities (e.g., traveling to sessions). The focus was on active music production with different instruments, and each session was structured in three 20-min modules: (1) rhythmical training ( Figure 1A and questionnaires, in addition to structural and resting-state functional magnetic resonance imaging. Self-report and caregiverreport questionnaires, as well as subjective feedback, were also collected 18 months after the intervention. In the following, we will summarize the main findings from the analyses of this data set.

EFFECTS OF MBNR ON COGNITIVE, BEHAVIORAL, AND EMOTIONAL RECOVERY AFTER TBI
The primary outcome was change in performance in the frontal assessment battery (FAB). 58 FAB is a measure of global executive functioning and consists of six subtests covering different aspects of frontal lobe functions: conceptualization, mental flexibility, motor programming, sensitivity to interference, inhibitory control, and environmental autonomy. Furthermore, computerized tests were administered to capture more narrowed aspects of EF (set shifting, updating, and inhibition) as defined by Friedman and Miyake. 17 Set shifting was measured with the number-letter task, where the subject is instructed to make a decision based on the number or the letter that appears on a screen by pressing one of two buttons. 59 Updating was measured with an auditory N-back, where the subject has to determine if the heard chord is the same or different compared with the previous (1-back) or the chord before that (2-back) by pressing one of two buttons. 60 Inhibition was measured with the Simon task, for which the subject has to press the right button each time a red square appears, or the left button each time a blue square appears, irrespective of which side the square is presented on. 61 As a measure of attention, we also used the Sustained Attention to Response Task, 62 a computerized task with digits ranging from 1 to 9, where the subject has to respond to every other digit, except digit 3, by pressing a response button.
To determine behavioral recovery, the patients were assessed with the Behaviour Rating Inventory of Executive Function-Adult (BRIEF-A) version. 63 In the current study, we used three indices: behavioral regulation index (comprising the inhibit and self-monitor scales), emotional regulation index (comprising the shift and emotional control scales), and metacognition index (comprising the remaining scales).
These indices together formed the global executive composite index.
In addition, we determined the effects of MBNR on everyday cognitive and emotional functioning as indicated by self-report and caregiverreport questionnaires on executive dysfunction, depression, and quality of life (QoL). Furthermore, we obtained subjective quantitative and qualitative feedback from the TBI patients and their caregivers regarding their experience of the intervention. To reduce potential bias introduced by the drop-outs, an intention-to-treat (ITT) analysis with multiple imputation was used for the cognitive tests and questionnaire

BRAIN MORPHOMETRY CHANGES INDUCED BY MBNR
In addition to cognitive improvement in EF outcomes, in this RCT, we investigated whether MBNR could have a neuroprotective effect on the progressive brain atrophy that is commonly reported after TBI.
Volumetric MRI measures, including voxel-based 64-67 and deformedbased morphometry, 67,68 have been used to identify this atrophy with sufficient statistical reliability and sensitivity to be applied in relatively small sample sizes. 69 Crucially, this progressive loss of brain volume has been linked to cognitive impairments, including verbal reasoning and memory. 66 67 Importantly, gray matter volume changes have been reported after a music listening intervention with stroke patients, 70,71 pointing to the sensitivity of volumetric analysis to monitor structural plasticity changes induced by music-based interventions in brain injury patients.
In our recent study, 53 we found that gray matter volume in the right inferior frontal gyrus was significantly greater in both the AB and BA groups after the music intervention period, both when the groups were compared with each other across time ( Figure 3) and when pooled together and compared with the control period ( Figure 4A,B). Critically, this change was correlated with better set shifting abilities during the intervention period ( Figure 4C). Other regions that most consistently showed an increase in gray matter volume after MBNR were the right middle frontal gyrus and the left superior frontal gyrus, cingulate cortex, insula, and cerebellum. One possible interpretation that fits well with the aforementioned spatial patterns of brain atrophy is that MBNR countered this neurodegeneration by providing environmental enrichment, which has been shown to protect against cognitive decline and atrophy after TBI in the chronic stage. 69,72

RESTING-STATE NETWORK PLASTICITY INDUCED BY MBNR
As alluded to above, TBI is often characterized by the presence of DAI and constitutes a pre-eminent structural disconnection syndrome. As Further, compared to other brain injury disorders in which gray matter is also commonly affected (e.g., stroke), TBI can occur with no visible focal brain injury and be most likely restricted to the strained damage of axons caused by DAI, which makes it a paradigmatic case for analyzing the effects of damage to the structural connectome on brain FC. Since music production engages widespread brain regions, MBNR stands as a suitable rehabilitation strategy that may contribute to restore function by driving neural plasticity that balances the damaged system and helps it to regain its optimal functioning.
For this reason, we investigated the effects of MBNR on the resting-state FC in large-scale networks and their relationship with EF outcomes. 55 More specifically, we analyzed FC patterns of four selected brain networks: the fronto-parietal network (FPN), the dorsal attention network (DAN), the salience (SAL) network, and the default mode network (DMN), which were used as seeds to assess FC within them and between them and every other node in the resting-state networks included in the CONN toolbox (for more details, see the Methods section in Ref. 55). The selection was based on brain networks whose constituent nodes have been previously found to play a functional role in the high-level cognitive functions that are typically damaged after TBI, most frequently EF, attention, and working memory. 23 This choice was further supported by the existing evidence showing that deficits in these cognitive domains are related to alterations in structural or FC after TBI. As mentioned above, it is possible to distill some commonalities in the abnormality patterns of large-scale networks following TBI. For example, disruption of the structural connectivity in the SAL network has been associated with deficits in sustained attention and increased DMN activation, suggesting that SAL network integrity is needed to switch from an internally oriented focus of attention, as represented by the DMN, to a salient external stimulus, as supported by the SAL. 73 The FC coupling between the SAL network and the DMN that allows for cognitive control has also been shown to be disrupted after TBI. 74 76 We also found an increased coupling between the FPN and the DAN that might be related to improved regulation of perceptual attention, as suggested by recent work using meta-analytic tools. 77 On the other hand, the MBNR reduced the connectivity between the DMN and the SM network, which could reflect a reduction in the interference between the DMN and taskdriven activities mediated by the SM network ( Figure 5D) Considering what has been previously outlined so far in our review, it may seem counterintuitive that the MBNR induces a reduction in the FPN and that this is associated with improvement in EF. However, this finding is in alignment with the hyperconnectivity framework. This hyperconnectivity is defined as an increase in FC strength after TBI. 78 According to the hyperconnectivity hypothesis, a major goal of the increase in FC following injury is to re-establish network communication through network hubs in order to maximize information transfer and minimize cognitive impairments. However, as has been shown in a recent cost-efficiency study, 79 this is made at a higher metabolic cost for the network that can ultimately lead to the long-term neurodegeneration patterns associated with TBI. The cost metric used by the authors includes the physical length as the Euclidean distance between nodes and, though it does not reflect a structural white matter connection directly, it is a good approximation for the cost that involves signaling through long distance connections in terms of metabolic energy and latency of neural transmission. This cost-efficiency study demonstrated that there is a concomitant increase in FC and network cost at early stages of TBI, mostly due to increased medium-range connections, whereas this cost is reduced at later stages during recovery.
Although we did not compute measures of network cost, it is worthwhile noting the effects of reduced within-network connectivity after the MBNR were located in frontal nodes, those showing the largest effect size for the increased network cost in Ref. 79.  However, there are still many questions that need to be addressed with regard to the mechanistic explanation for the efficacy of MBNR.

CONCLUSIONS AND FUTURE DIRECTIONS
For succinctness, we will focus on two of these questions in relation to some limitations in our resting-state analysis. One of these limitations concerns the low number of nodes included in the network analysis, which was restricted to cortical regions. Therefore, our results are preliminary, and it would be relevant to replicate these findings with a fine-grained brain parcellation scheme that includes subcortical regions. In particular, corticostriatal interactions could play a prominent role in the music-induced benefits, as they are important for cognitive and motivational aspects that are both recruited during music making and targeted in TBI rehabilitation. In this sense, it is important to highlight that a disruption in the connectivity between the caudate and cortical regions in TBI has been linked to executive dysfunction together with increased levels of fatigue and apathy. 81 Moreover, frontostriatal connections have been shown to be dependent on the levels of music reward sensitivity in the healthy population, suggesting that one mechanism for the rehabilitative effects of music might be experience-dependent plasticity in these connections. The music-induced dopamine release via the mesocortical and mesolimbic pathways could further contribute to the cognitive and emotional outcomes after TBI. From this perspective, music could provide a nonpharmacological dopaminergic stimulant that improves cognitive function as has been reported for methylphenidate in TBI patients. 82 The second limitation is inherent to the static FC approach adopted for the resting-state analysis, which overlooks the dynamic aspects of brain network function. Indeed, resting-state brain networks are constantly reconfiguring over time to allow optimal information processing capabilities. 83 In this vein, it is likely that the reduced executive functioning after TBI may be a consequence of reduced whole-brain synchronization at global, network, and/or node level. Recently, a computational model framework has proposed that brain dynamics exhibit turbulent-like behavior 84 and that the optimal fitting of the brain model to the empirical data occurs when the turbulence amplitude levels as well as the integration/segregation capabilities are maximal. Therefore, future work should examine how turbulent-like dynamics is disrupted after TBI, its relationship with behavioral measures, and its malleability with music-based interventions. It is our hope that this computational modeling approach will provide important insights into the causal mechanisms underlying the neural plasticity induced by MBNR in individuals with TBI.

ACKNOWLEDGMENTS
Financial support was provided by the Academy of Finland  Mänttäri. We would also like to thank the TBI patients and the family members who participated in our study.