Comparing methods to analyze functional dynamics in at-rest musicians
Poster No:
2032
Submission Type:
Abstract Submission
Authors:
Abia Fazili1, Harrison Watters1, Shella Keilholz2
Institutions:
1Emory University, Atlanta, GA, 2Emory University and Georgia Institute of Technology, Atlanta, GA
First Author:
Co-Author(s):
Introduction:
A resting state pattern of anti-correlated brain activity between default mode (DMN) and task positive networks (TPN) has been implicated in attentional processing (1). This pattern is quasi-periodic, completing a cycle about once in 20 seconds in humans (2).
We detected quasi-periodic patterns (QPPs) using an algorithm based approach and complex principal components analysis (CPCA), to compare dynamic functional connectivity in a dataset of classical, improvisational, and non-experienced musicians. Based on previous literature, we hypothesized that we would detect differences in dynamic functional connectivity between the DMN and visual network and that these groupwise differences would be robust across two methods. In improv musicians, both methods detected an increase in visual-DMN correlation during the QPP and an unexpected increase in QPP correlation between the amygdala and dorsal attention network (DAN).
We detected quasi-periodic patterns (QPPs) using an algorithm based approach and complex principal components analysis (CPCA), to compare dynamic functional connectivity in a dataset of classical, improvisational, and non-experienced musicians. Based on previous literature, we hypothesized that we would detect differences in dynamic functional connectivity between the DMN and visual network and that these groupwise differences would be robust across two methods. In improv musicians, both methods detected an increase in visual-DMN correlation during the QPP and an unexpected increase in QPP correlation between the amygdala and dorsal attention network (DAN).
Methods:
Functional scans were obtained at the Psyche Lab at Northeastern Universeity. 48 subjects were classified by musical training: classical, improvisational, or minimal (MMT). Each group had 16 subjects (n = 4 females, n = 12 males). Groups were matched in age and cognitive and musical ability. For more details see Belden et al., 2020 (3).
T1 weighted structural scans and resting state functional scans were obtained on 3T Siemens scanners. T1-weighted sequences were 3D magnetization prepared rapid-acquisition gradient-echo (MPRAGE) with a voxel size of 0.8 x 0.8 x 0.8 mm3 (TR = 2.4 s, TE = 2.09 ms, flip angle = 8°, FOV = 256 mm). Resting state scans were 7.5 minutes with an echo-planar imaging sequence of 947 volumes (TR = 475 ms; TE = 30 ms; flip angle = 90°, 48 slices; FOV = 240 mm; acquisition voxel size = 3 × 3 × 3 mm3). Pre-processing and global signal regression was done with the CPAC pipeline (https://fcp-indi.github.io/) and Brainnetome atlas (5).
QPPs were detected with an in-house algorithm (8) then plotted as Yeo's seven networks and one subcortical network to compare network correlations (7). We used CPCA, a dimensionality reduction method that captures the QPP (4), to visualize QPPs in brain space with FSLeyes (see Fig. 1). Algorithm-based and CPCA-based waveforms were plotted (Fig. 2).
T1 weighted structural scans and resting state functional scans were obtained on 3T Siemens scanners. T1-weighted sequences were 3D magnetization prepared rapid-acquisition gradient-echo (MPRAGE) with a voxel size of 0.8 x 0.8 x 0.8 mm3 (TR = 2.4 s, TE = 2.09 ms, flip angle = 8°, FOV = 256 mm). Resting state scans were 7.5 minutes with an echo-planar imaging sequence of 947 volumes (TR = 475 ms; TE = 30 ms; flip angle = 90°, 48 slices; FOV = 240 mm; acquisition voxel size = 3 × 3 × 3 mm3). Pre-processing and global signal regression was done with the CPAC pipeline (https://fcp-indi.github.io/) and Brainnetome atlas (5).
QPPs were detected with an in-house algorithm (8) then plotted as Yeo's seven networks and one subcortical network to compare network correlations (7). We used CPCA, a dimensionality reduction method that captures the QPP (4), to visualize QPPs in brain space with FSLeyes (see Fig. 1). Algorithm-based and CPCA-based waveforms were plotted (Fig. 2).
Results:
In the algorithm-based QPP approach, the improv musicians showed positive correlation for visual-DMN (r = 0.830) while the classical (r = -0.214) and MMT (r = -0.879) groups had negative correlations. Using CPCA, the improv group's visual-DMN was also positively correlated (r = 0.371), and the classical (r = -0.549) and MMT (r = -0.960) groups' visual-DMN were negatively correlated.
For both methods, the improv group had greater amygdala-DAN correlation (r = 0.878, CPCA: r = 0.369) than the classical (r = 0.212, CPCA r = -0.275 ) and MMT groups (r = 0.014, CPCA: r = -0.502). While the sign of the correlations was consistent across methods, they differed in their strength. This was expected given that networks were defined differently in each method. For example, for CPCA, the entire DMN was defined as the Posterior Cingulate Cortex while the algorithm-based method used a mask of DMN-associated ROIs.
For both methods, the improv group had greater amygdala-DAN correlation (r = 0.878, CPCA: r = 0.369) than the classical (r = 0.212, CPCA r = -0.275 ) and MMT groups (r = 0.014, CPCA: r = -0.502). While the sign of the correlations was consistent across methods, they differed in their strength. This was expected given that networks were defined differently in each method. For example, for CPCA, the entire DMN was defined as the Posterior Cingulate Cortex while the algorithm-based method used a mask of DMN-associated ROIs.
Conclusions:
We used an algorithm and a CPCA-based approach to detect QPPs, and found convergent results with both methods in a resting dataset of musicians. The increased visual-DMN correlation in the improv musicians' QPP aligned with the static functional connectivity analysis of Belden et al., 2020. Also, both methods detected a positive correlation between amygdala-DAN in improv musicians. This shows the sensitivity of QPP analysis to groupwise connectivity differences and supports a potential relationship between visual-DMN activity and creative cognition. Given the link between amygdala functional connectivity and anxiety processing (6), the novel finding of increased amygdala-DAN connectivity in improv musicians points to future studies that relate anxiety to creative training.
Higher Cognitive Functions:
Music 2
Modeling and Analysis Methods:
Task-Independent and Resting-State Analysis 1
Perception, Attention and Motor Behavior:
Attention: Visual
Keywords:
Data analysis
FUNCTIONAL MRI
1|2Indicates the priority used for review
Provide references using author date format
2) Abbas, A. (2019), 'Quasi-periodic patterns contribute to functional connectivity in the brain', Neuroimage, vol. 191, pp. 193-204.
3) Belden, A. (2020), 'Improvising at rest: Differentiating jazz and classical music training with resting state functional connectivity', Neuroimage, vol. 207, 116384.
4) Bolt, T. (2022), 'A Parsiminious description of global functional brain organization in three spatiotemporal patterns', Nature, vol. 25, pp. 1093-1103.
5) Fan, L. (2016), 'The Human Brainnetome Atlas: A New Brain Atlas Based on Connectional Architecture', Cerebral Cortex, vol. 26, no. 8, pp. 3508-3526.
6) Ye, H. (2016), 'Lifespan anxiety is reflected in human amygdala cortical connectivity', Human Brain Mapping, vol. 37, no.3, pp. 1178-1193.
7) Yeo, B. T. (2011), 'The organization of the human cerebral cortex estimated by intrinsic functional connectivity', Journal of Neurophysiology, vol. 106, no. 3, pp. 1125-1165.
8) Yousefi, B. (2021), 'Propagating patterns of intrinsic activity along macroscale gradients coordinate functional connections across the whole brain', Neuroimage, vol. 231, 117827.