Multiscale structural architecture of human neural dynamics
Poster No:
2091
Submission Type:
Abstract Submission
Authors:
Jessica Royer1, Casey Paquola2, Raúl Rodriguez-Cruces3, Hans Auer3, Alexander Ngo3, Ella Sahlas4, Daniel Mansilla De Latorre5, Raluca Pana3, Jeffrey Hall3, Birgit Frauscher6, Boris Bernhardt3
Institutions:
1MNI, Montreal, QC, 2INM-7, Jülich, Jülich, 3Montreal Neurological Institute and Hospital, Montreal, Quebec, 4McGill University, Montreal, Quebec, 5Institute of Neurosurgery, Asenjo, Santiago, 6Duke University, Durham, NC
First Author:
Co-Author(s):
Introduction:
Neural dynamics are complex and heterogeneous across the cortex. Local spectral signatures are constrained by structural properties [1,2], but the relationship between neural dynamics and cortical structure remains incompletely understood. The present work explores this question by investigating the balance of extrinsic and intrinsic structural constraints on regional neural dynamics. We provide an integrated account of this interplay by combining the high temporal and spatial precision of intracranial electroencephalography (iEEG) with multiscale measures of cortical wiring, specifically inter-regional distance, structural connectivity estimated from diffusion-weighted MRI tractography, and microstructural profile similarity.
Methods:
The MNI open iEEG atlas[3] provides iEEG recordings acquired during conditions of resting wakefulness in 106 patients with intractable epilepsy (atlas dataset; Fig1A). By excluding channels involved in ictal and interictal activity, this dataset provides a putative reference space for normal human neurophysiology. Data pre-processing included band-pass filtering (0.5-80Hz), downsampling (200Hz), and demeaning. We computed each channel's power spectral density (PSD; Welch's method; 2-second blocks, 1-second overlap, Hamming window weighting). Channel PSDs were log-transformed, mapped to a single hemisphere, and parcellated[4]. Parcel-wise PSDs were cross-correlated while controlling for the average PSD across all channels and underwent Fisher R-to-Z transformation. We then applied diffusion map embedding to derive principal axes of variation in neural dynamics (Fig1B)[5]. In a second dataset of 20 patients (multimodal dataset; 13F; mean±S.D. age=33.90±9.02years), iEEG was recorded during resting wakefulness, and pre-operative, high-resolution structural (T1-weighted, quantitative T1, 0.8mm isovoxels) and multi-shell diffusion-weighted MRI (DWI; 1.6mm isovoxels) were acquired, enabling dataset-specific correlations between brain wiring and macroscale neural dynamics. We used micapipe[6] to derive subject-specific measures of geodesic distance, microstructural profile similarity, and structural connectivity across all node pairs. Electrophysiological data underwent identical processing as in the atlas dataset. Channel-level PSDs were averaged across patients within each parcel (Fig2A), and multimodal dataset embeddings were aligned to the atlas embedding space using Procrustes rotation. We assessed structure-function coupling with a multiple linear regression model using three structural features as predictors of inter-node distances in the multimodal dataset embedding space.
·Figure 1
Results:
The first gradient (G1) of neural dynamics differentiated primary motor and surrounding frontal cortices from occipito-temporal regions, segregating channels with dominant beta and gamma-range activity (>13hz) from those with high-delta, theta, and alpha-range peaks (<13Hz) (Fig1C). The second gradient (G2) differentiated unimodal sensory cortices, encompassing channels with peaks in the alpha frequency range (8-13Hz), from limbic and paralimbic regions with strong low frequency activity (<4Hz). This compact representation reflected distinct spectral signatures of unimodal sensory, motor, and association areas (Fig1D), and could be replicated in the multimodal dataset (median r=0.34; Fig2B). Structural features explained up to 60% of variance in distances within the embedding space: highest R2 values were observed in frontopolar and lateral temporal areas (Fig2C). This model outperformed models implementing different functional response variables (Fig 2D). These results show diverse contributions of cortical wiring to regional neural dynamics, with variable coupling strengths across the neocortex.
·Figure 2
Conclusions:
By mapping gradients of neural dynamics, our approach resolves macroscale trends in spectral similarity of local cortical regions and opens the way for assessments of structure-function coupling from direct measurements of neural activity.
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Anatomy and Functional Systems 1
Cortical Anatomy and Brain Mapping 2
Keywords:
Cortex
ELECTROPHYSIOLOGY
Myelin
STRUCTURAL MRI
White Matter
Other - Structure-function coupling, Gradients
1|2Indicates the priority used for review
Provide references using author date format
2. Wang X-J. Macroscopic gradients of synaptic excitation and inhibition in the neocortex. Nature Reviews Neuroscience. 2020/03/01 2020;21(3):169-178. doi:10.1038/s41583-020-0262-x
3. Frauscher B, von Ellenrieder N, Zelmann R, et al. Atlas of the normal intracranial electroencephalogram: neurophysiological awake activity in different cortical areas. Brain. 2018;141(4):1130-1144. doi:10.1093/brain/awy035
4. Schaefer A, Kong R, Gordon EM, et al. Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cerebral cortex. 2018;28(9):3095-3114.
5. Vos de Wael R, Benkarim O, Paquola C, et al. BrainSpace: a toolbox for the analysis of macroscale gradients in neuroimaging and connectomics datasets. Communications biology. 2020;3(1):1-10.
6. Cruces RR, Royer J, Herholz P, et al. Micapipe: A pipeline for multimodal neuroimaging and connectome analysis. NeuroImage. 2022;263:119612.