Effects of music composition on structural and functional connectivity in the orbitofrontal cortex

Abstract

Aims

Composer is a classic case for investigating the plasticity effect of musical composition on the brain. As an essential brain region associated with prediction and decision, the orbitofrontal cortex exhibits the significant difference in many studies between composers and controls. Meanwhile, these studies also show that musical composition induces changes in cognitive networks, such as auditory and attentional functions. However, these structural and functional connectivity changes are often measured independently, making it crucial to exam the link between these two changes in order to further understand the neuroplasticity of musical composition.

Methods

In this work, we recruited 18 composers and 20 controls under resting-state functional Magnetic Resonance Imaging (fMRI) scanning. First, based on the Tract-Based Spatial Statistics method, we found the differences in white matter skeleton between composers and the controls. Subsequently, we compared the differences in structural connectivity by probabilistic tracing from the orbitofrontal cortex. Finally, we examined the functional difference between groups.

Results

We found that composers had higher anisotropy scores and mean diffusion rates in white matter regions such as the corpus callosum, anterior radiating corona, anterior and posterior branches of the internal capsule than the control group. Functional connectivity also provided evidence for the more robust relationship between the orbitofrontal cortex and other regions, such as attentional networks.

Conclusion

This experiment suggests that the structural and functional connectivity between the orbitofrontal cortex and other higher cognitive areas of composers is stronger than nonmusicians, which improves the understanding of the effect of composition training on structural and functional neuroplasticity.

Keywords:

• Composer
• diffusion tensor imaging
• functional connectivity
• the based spatial statistics method

1. Introduction

Music training, including composition training, has a positive effect on brain plasticity, causing structural and functional changes in the primary sensory and higher multifunctional brain areas [1]. For example, music-trained children have different rates of cortical thickness maturation between the right and left posterior superior temporal gyrus and higher anisotropy in the fraction of callosities connecting the frontal, sensory, and superior motor segments [2]. A functional connectivity study also reported that the superior temporal gyrus of musicians showed preferential activity in response to musical stimulation compared to controls who did not receive musical training [3]. Arkin et al. performed an MRI investigation on jazz musicians, revealing that extended musical training influenced the grey matter volume in both the right inferior temporal gyrus and the bilateral hippocampus [4]. Besides, there is ample evidence, supported by dual processing streams [5], that orbitofrontal functions, in particular, may be critically essential across the lifespan from childhood to adulthood, especially in response to pleasurable or aesthetic aspects of musical stimulation [6,7], which is an important ingredient in composition. Among these, the orbitofrontal cortex plays an essential role in music perception, with its neurons responding to music and allowing musicians to activate primary auditory cortex neurons relatively quickly [8].

Compared to regular musicians, improvising musicians have distinctive brain network connections. According to Belden et al. improvising musicians exhibited significant differences connectivity between the ventral brain regions of the default mode network (DMN) and the bilateral executive control network (ECN). They also demonstrated superior connectivity within the primary visual network with the DMN and ECN, as well as between the ECN and DMN [9]. Another study found that improvization training was positively correlated with functional connectivity in the bilateral dorsolateral prefrontal cortex, dorsolateral premotor cortex, and pre-supplementary areas [10]. This corroborates that from the perspective of prefrontal theory, which is responsible for creative thinking [11,12], the functional connectivity of improvizing musicians in this region is superior to that of controls. Limb and Braun conducted a study involving professional pianists in which they observed that when pianists engaged in improvisational composition, there was a notable increase in activity in the medial prefrontal region (MPFC) [13], suggesting that improvization leads to enhanced functional connectivity in this region.

To some extent, the ability to improvize indicates greater creative power. Regarding social experience, compositional capacity refers to creating something innovative and impactful, often used as a yardstick by which composers can be distinguished from non-musicians. The abilities of composers, including the ability to focus and distract, to search memory stores intentionally or spontaneously, and to search using cognitive or emotional search processes [14], are enhanced after a long period of training in music composition. These abilities involve various integrated cognitive activities and brain area connections [15]. Previous research also has shown that engaging in music composition leads to a significant increase in cortical surface area or subcortical volume in specific regions associated with creative cognition (dorsomedial prefrontal cortex, middle temporal gyrus, and temporal pole), higher cognitive-motor functions and auditory processing (dorsal premotor cortex, supplementary motor areas, anterior supplementary motor areas, and temporal plane), as well as emotional areas (orbitofrontal cortex, temporal pole, and amygdala) [16]. Beaty et al. used fMRI imaging data to find significant brain structural differences between composers and controls in the prefrontal network commonly linked to improvisatory behavior [17]. These suggest that compositional training has a specific plasticity effect on the brain structure.

Moreover, from the perspective of functional connectivity, one study showed that the effect of training on composing was positively correlated with functional connectivity in the dorsolateral prefrontal cortex and dorsal premotor cortex [10]. Another study found that composers showed stronger connectivity in the orbitofrontal, temporal brain regions, right angular gyrus, and bilateral superior frontal gyrus regions [18]. EEG studies also have demonstrated stronger connections between the orbital frontal cortex and other brain regions when composing through coherence analysis [19], suggesting that more information is exchanged between brain regions during the compositional phase, allowing for more significant differences in functional connectivity than controls. However, research on whether and how composition training can affect the brain’s structural connectivity still needs to be completed. The structural and functional connectivity of the orbitofrontal cortex has rarely been explored in depth. Studying the connectivity specificity of the orbitofrontal cortex in composers would help to better understand the functional and structural plasticity of compositional training on the brain as well as musical intervention. Meanwhile, it would complement the mechanism of brain–apparatus communication research based on music training and promote the development of this field [20]. Therefore, this paper takes the orbitofrontal region as a starting point and uses diffusion tensor imaging (DTI), tract-based spatial statistical analysis (TBSS), and functional connectivity calculations to compare the structural and functional connectivity with other regions. It will lead to a deeper understanding of the plasticity of the brain in composition training and provide theoretical support for studying brain mechanisms in music composition.

2. Materials and methods

2.1. Materials

Thirty-eight subjects were recruited in this experiment, including 18 composers and 20 controls. The composers were recruited through an advertisement in the composition department of the Sichuan Conservatory of Music. All were recruited by questionnaire to ensure that they had at least three years of composition training. In order to ensure that the recruited composers had a high level of training in music composition, they all participated and passed a composition-level test organized by the composition department of the Sichuan Conservatory of Music. In addition, the demographic data were collected (Table 1). The control group had 20 nonmusician subjects who had no professional musical training and were recruited at the University of Electronic Science and Technology of China.

Table 1. Demographic data on subjects.

demographics	Composer (N = 18)	Controls (N = 20)
Age	19.22$\pm 1.31$	22.25$\pm 2.24$
Gender (M/F)	8/10	14/6
Start learning music age	11.36$\pm 4.92$	N/A
Start studying Composition Age	17.11$\pm 0.90$	N/A
Compose music days per week	4.17$\pm 1.38$	N/A
Composing hours per day	2.36 $\pm$1.03	N/A
Number of days per week in composition	4.17$\pm 1.38$	N/A
Play an instrument days per week	4.78 $\pm 1.70$	N/A
Total training time (hours)	1063.11 $\pm 928.87$	N/A
Instrumental hours per day	1.69$\pm 0.93$	N/A

All subjects had no neurological or psychiatric disorders. The composers and control groups did not differ in gender and age. They were paid for participating in the study. The study was done with the approval of the Ethics Committee of the School of Life Science and Technology at the University of Electronic Science and Technology of China. All procedures were carried as following approved guidelines.

2.2. Methods

2.2.1 fMRI scans

Functional MRI scans were acquired at the Magnetic Resonance Imaging Research Centre of the University of Electronic Science and Technology of China. MRI data were collected from each subject using a 3 T MRI machine (GE Discovery MR750, USA). During scanning, foam padding and earplugs were used to reduce head motion and scanning noise, respectively. In the resting-state functional MRI scans, each time point is spaced 2s apart, and 255-time points are captured. The functional images were acquired using gradient-echo EPI sequences (slices = 30, slice scan order: interleave, averages/measurements = 1/244, echo time [TE] = 30 msec, repetition time [TR] = 2000 msec, flip angle [FA] = 90°, field of view [FOV] = 24 × 24 cm², matrix = 64 × 64, acquisition voxel size = 3.3 × 3.3 × 4.0 mm³, reconstructed voxel size = 3 × 3 × 3 mm³, multisided mode/series: interleaved/descending, bandwidth = 2232, slice thickness/gap = 4 mm/0.8 mm), with an eight-channel phased array head coil. To ensure steady-state longitudinal magnetization, the first five volumes were discarded. Subsequently, high-resolution T1-weighted images were acquired using a 3-dimensional fast spoiled gradient echo (T1-3D FSPGR) sequence (TR = 5.948 msec, TE = 1.964 msec, FA = 9°, matrix = 256 × 256, FOV = 20.4 × 16.3 cm², slice thickness/gap = 1 mm/0 mm, slices = 154).

2.2.2. Preprocessing

In this paper, all subjects’ fMRI data were pre-processed using the MATLAB toolbox NIT (http://www.neuro.uestc.edu.cn/NIT.html) [21]. Preprocessing steps include realignment of head movements, slicing timing (time correction), spatial normalization through a standard spatial template developed at the Montreal Neurological Institute (MNI), spatial smoothing, and finally regression and filtering. Pre-processing of Diffusion tensor imaging data is mainly undertaken using the FMRIB Software Library (FSL) software. The preprocessing steps included scalp removal, head movement, and vortex correction.

2.2.3. Tract-based spatial statistics

Tract-Based Spatial Statistics is a method to quantify spatial information by aligning the fractional anisotropy (FA) and mean diffusivity (MD) values and projecting them onto an aligned invariant bundle representation (mean FA skeleton) [22]. The calculation process was as follows. First, the FA images of composers and non-musician subjects were aligned with the same FA template using a non-linear alignment method. The 'bundle vertical’ voxel with the highest FA was then defined as the center of the bundle, forming the bundle skeleton, and the estimated bundle verticality was normalized to improve the stability of the estimate. The final step is to perform voxel statistics by taking each subject’s FA data in skeleton space. A nonparametric test based on permutations [23] was chosen for this paper, and the results were corrected based on familywise errors (FWE).

2.2.4. Analysis of probabilistic tracer

Probabilistic fiber tract tracing of DTI data was calculated primarily using the FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki) toolbox. In the calculations, we first selected a specific ROI as the seed point. We constructed a voxel-based whole brain atlas by randomly constructing 5000 fiber tracts starting with each voxel point in that ROI region. We estimated FA and MD values using the DTI data after pre-processing. The number and direction of fiber bundles within each voxel were then estimated, setting the number of fibers modeled in each voxel to 2, the primary multiplicativity factor for additional modeled fibers to 1, and the number of iterations before sampling to 1000. In this work, the FSL software was used to align each subject’s structural image T1 data to the individual DTI space, and the aligned T1 image was transformed to the standard MNI T1 template to generate a transfer matrix. Finally, the ROI of each subject in the original space was acquired. Also, We have chosen a threshold of 0.2–0.8 for showing the difference in skeleton between the two groups so that voxels primarily in the grey matter or cerebrospinal fluid could be successfully excluded in most subjects, meaning that the skeleton would not run directly to the extreme outer edge of the cortex [22].

2.2.5. Functional connectivity

Functional connectivity was calculated between the regions of the interest (the orbitofrontal cortex) and other 399 ROIs based on Schaefer’s report [24]. After computing the functional connectivity matrix of 18 composer and 20 non-musician subjects, we did a Fisher-z transformation to convert Pearson’s coefficients into Gaussian distribution to perform a t-test. Afterward, the t-test was used on the functional connectivity to obtain the significant differences between the composers and controls (p < 0.05).

3. Results

TBSS was used for voxel-wise analysis of whole brain DTI measures, t-statistical maps of FA and MD group differences (composers vs. controls) of TBSS results views overlaid on the MNI template was shown in Figure 1. We applied a threshold range of 0.2 to 0.8 for the FA and MD values, with the aim of excluding peripheral tracts exhibiting notable inter-subject variability and potential partial volume effects with grey matter. Figure 1(a) shows that the composer had significantly greater FA values than controls for structures such as the corpus callosum knee, corpus callosum trunk, corpus callosum compression, cerebral peduncle, anterior branch of the internal capsule, the posterior branch of the internal capsule, anterior radioulnar corona, epithalamus radiation and so on. Figure 1(b) shows that composers had significantly greater MD values than controls for structures such as the anterior radial crown, superior radial crown, posterior radial crown, anterior branch of the internal capsule, the posterior branch of the internal capsule, splenium of the corpus callosum, superior longitudinal fasciculus, and superior frontal occipital fasciculus, etc.

Figure 1. Difference of FA and MD values in TBSS between composers and controls. The background is the T1 image template of MNI152 at 1${\mathit{\text{mm}}}^3,$ the green part is the calculated mean FA skeleton (threshold chosen from 0.2 to 0.8), the red-yellow part in Figure 1(a) is the region where the FA values of composers are significantly larger than those of controls (p < 0.001, FWE corrected). The red-yellow part in Figure 1(b) is the region where MD values are significantly greater for composers than for controls (p < 0.001, FWE-corrected).

The results of probabilistic tracing in diffusion tensor imaging of composers and controls are shown in Figure 2, we chose the ROI of the orbital frontal cortex as the seed point (starting point). It exhibits that fiber bundles originating from ROI have a higher probability of passing through more areas compared to the control group. The colored voxel dots (red and yellow) correspond to the voxels that the fiber bundle has a probability of passing through, with red to yellow indicating weak to strong probability of passing through. The fiber tracts from the seed point cover the entire left frontal region and even extend to the parietal and temporal regions. In contrast, in the control group, the fiber tracts from the seed point (frontal, middle gyrus, orbitalis) only reach the frontal middle and inferior gyrus. At the level of the fiber tracts, in composers, the fiber tracts from the seed point to the regions of the anterior radial crown, superior radial crown, corpus callosum, splenium of the corpus callosum, corpus callosum compression, anterior branch of the internal capsule, posterior branch of the internal capsule, and external capsule. Moreover, the probabilistic tracing results showed that many regions in the control group were not significantly connected. In contrast, the composer’s tracing results went through a large area of the brain and are therefore represented in this figure as results.

Figure 2. Probabilistic tracings in diffusion tensor imaging of composers and controls, with the FA standard template for HCP in FSL in the background. The color scale of 0.0095 (yellow) is the normalized maximum value for the number of fiber bundles, and voxels with no fiber bundles passing through have a value of zero, giving a transparent color.

Functional connectivity results showed that the differences between composer and controls. As shown in Figure 3, compared with controls, the stronger connections of composers are concentrated in bilateral dorsal as well as ventral attention network, such as insula, inferior temporal gyrus, postcentral gyrus, inferior parietal marginal angular gyrus, etc. The lower connections of composers are concentrated in the left DMN, such as angular gyrus, middle temporal gyrus, medial superior frontal gyrus, middle occipital gyrus and so on. All function connection results are statistically checked (p < 0.05).

Figure 3. Functional connectivity has a significant difference between composers and controls. Functional connectivity of composers between the left OFC and the right and bilateral dorsal and ventral attention network are more significant than those of controls. The strength of functional connections between the left OFC and the left DMN of composers is lower than controls (RH: Right hemisphere, red line: higher functional connection than controls, blue line: lower functional connection than controls) (p < 0.05).

4. Discussion

Metrics such as FA and MD in DTI can quantify local fasciculations’ directional strength and the rate of directional displacement. From a spatial perspective, the study of FA and MD in the brain can help locate changes associated with development, degeneration, and disease. A study showed that musicians who had received musical training since early childhood had greater FA in the corpus callosum knee than controls. This change may be due to the cognitive and motor effects of musical training [25]. Manzano et al. also found the structure difference that musicians had greater cortical thickness in the auditory-motor network in the left hemisphere and more developed white matter microstructures in both hemispheres with the associated tracts of the corpus callosum [26]. In addition, from the DTI perspective, instrumental music training also has an impact on brain structure. A study showed that after experiencing the long-time music exercise, the anterior, middle part of the corpus callosum was more significant in children compared with low-intensity practice and without any instrumental training [27]. Some studies focusing on FA indexes also showed a strong positive correlation between pianists’ fiber system integrity and piano practice time [28]. Han et al. showed that pianists had higher FA in the posterior branch of the internal capsule compared with controls, representing greater white matter integrity. This higher FA may be associated with somatic movements in playing the instrument [29]. These findings indicate that musical training induces unique structural plasticity in the brain. Our work focuses on composers with creative power, a rare subject in previous research, and verifies the differences in brain structure between composers and controls. In our study, we observed that the FA values in composers were notably elevated compared to those in the control group. This effect is mainly reflected in the structure such as the corpus callosum and internal capsule. This suggests that the process of compositional training, which involves extensive music appreciation and experience in composition training, enhances the cognitive and somatic motor functions of the composer.

MD is a highly sensitive biomarker of microstructural changes associated with speech-language learning [30]. Acer et al. showed that musicians have higher MD in the corpus callosum than controls [31]. The superior longitudinal tract is considered the most prominent pathway for processing and producing language and music [32]. Our study also found that composers had significantly higher MDs in structures of superior longitudinal fasciculus and corpus callosum than controls, which suggests that compositional training has significant changes in areas of the brain associated with language learning.

As a result, there is an increase in the number of white matter fiber bundles associated with function and a greater overall integrity related to structure. In addition, we found that composers showed more structures differences from the perspective of FA and MD indexes. This suggests that the disparities observed between composers and the control group cannot be solely attributed to musical or instrumental training. Instead, the structural alterations in the brain are likely influenced by the linguistic abilities, cognitive functions, and somatic motor skills honed through prolonged compositional training.

Moreover, with the DTI probabilistic tracing results in this paper, composers have many fiber tracts connected to various parts of the corpus callosum from the seed point. Previous studies have shown that compared to controls, the anterior half of the corpus callosum, which is responsible for inter-hemispheric communication, including premotor, supplementary motor, and motor cortices of professional musicians, are significantly larger than in controls [33]. During musical training, the plasticity of the corpus callosum may reflect the bilateral motor coordination and auditory skills musicians need while playing their instruments [34]. This is highly related to the composer’s years of creative music training, with many bundles of fiber tract connections to the seed-points and corpus callosum, which is significantly associated with composers’ long-term instrumental training. In addition, structural white matter connections contribute to integrating neural information across functionally separated systems [35]. Composers in this paper have more fiber tract strips from the seed point to other regions than controls and are involved in a broader range of areas. This suggests that composers have stronger connectivity and more efficient information transferring from the seed point on the white matter structure to regions such as the radial crown, corpus callosum, internal capsule and external capsule, evidence that the composer’s primary perceptual areas are more connected to higher, multifunctional cognitive areas.

Also, we explored whether structural differences in composers’ brains affect functional brain connectivity from a seed point with the perspective of brain networks. Based on a priori knowledge, it is already known that composers only show more significant connection differences with other brain regions in the left orbitofrontal cortex compared to controls [36]. During our study, we found that composers had significantly enhanced connectivity between the orbitofrontal cortex and dorsal attention network as well as ventral attention network, which indicates the functional and structural connections in the cerebral cortex reflect each other. An article verifying the relationship between structure and function in the resting brain showed that the strength of functional connectivity at rest was positively correlated with the strength of structural connectivity [37]. Koch et al. also found that grey matter areas connected by white matter fiber tracts exhibited high levels of functional connectivity [38]. The structural and functional plasticity of the brain result of musical training is also synchronous and interrelated. One study, starting from resting-state EEG, found that patients had less intense activity in frontal areas than individuals under musical stimulation, concluding that the frontal lobe can act as an essential brain region for music recognition, giving more positive emotional feedback [39]. To explore the relationship between the OFC and other regions from a structural and functional point of view, we have intervened from a macro perspective. Composers exhibit stronger connections within the attention network compared to controls, particularly the ventral attention network for bottom-up information shifts and the dorsal attention network for maintaining attentional stability [34]. It may be explained that the composers, after long-term music composition, have increased functional-structural connectivity in brain areas related to prediction and choice with information shifts and brain areas for attentional maintenance, resulting in composers performing predictive choice activities generally more likely to focus their attention and be more amenable to bottom-up processing of information. This function is reflected in higher perceptual regions associated with attentional networks, and functionally associated brain regions can also exhibit some coincidence in the number of fibers in white matter structures. Moreover, we also found that the functional connectivity between the left OFC and the DMN of composers is lower than controls, possibly as compensation for increased connectivity between the left OFC and Attention Network.

However, in the daily training process, composers receive not only compositional training but also musical and instrumental training. Therefore, although the article focuses on composers and non-musicians, and the synthesis of prior research along with the current study’s results establishing distinct differences in structural and functional connectivity within the orbitofrontal cortex of composers compared to non-musicians, it still a limitation that the effects of musical (instrumental) training are not well differentiated from the effects of compositional training. So, in the future study, we plan to add a group of musicians (with no experience of composition training) as a control group to further studies the effects of composition and music training on brain plasticity respectively. In addition, the inter-group subject functional connectivity results did not pass the calibration, which may be due to the small sample size of the composers. Future studies should consider a larger sample size that reflects the significant difference Inter-group.

5. Conclusion

This work focuses on the brain structural differences between composers and controls. The white matter skeleton showed distinct differences between the two groups, particularly in the corpus callosum knee, trunk, and capsule. Composers exhibited higher FA values in structures like the corpus callosum and internal capsule, while MD values were elevated in specific regions including the radial crown and branches of the internal capsule, as well as the corpus callosum and longitudinal fasciculus. This indicates that training in composition enhances the integrity of the composer’s diverse white matter structures, reducing the level of separation among the different layers of white matter. Subsequently, we found that the fiber bundles originated from left OFC have a higher probability passing through more areas compared to the control group, suggesting that composers had better connectivity, greater integration, and more efficient information transfer. Moreover, we also found that composers’ functional connectivity in the left OFC and the dorsal as well as ventral attention network was stronger than controls. Starting from the seed point, the overall structural and functional results are consistent, collectively revealing specificity in the connectivity to the OFC between composers and non-musicians. This suggests that compositional training has the potential to induce plasticity in the brain.

Acknowledgments

We acknowledge all the subjects in this experiment. This work was supported by Chinese MOST Project (No. 2O22ZD0208500) and the Sichuan Science and Technology Program (2021YFS0135).

Disclosure statement

No potential conflict of interest was reported by the author(s).