Wiring the metanetwork of language: a connectome-driven neurosurgery approach
Poster No:
2103
Submission Type:
Abstract Submission
Authors:
Ludovico Coletta1, Paolo Avesani1, Luca Zigiotto2, Martina Venturini3, Luciano Annicchiarico2, Laura Vavassori4, Sharna Jamadar5, Emma Liang5, Justine Hansen6, Emmanuel Mandonnet7, Sam Ng8, Hugues Duffau8, Silvio Sarubbo2
Institutions:
1Fondazione Bruno Kessler, Trento, Trento, 2Azienda Provinciale per i Servizi Sanitari, Trento, Trento, 3Università dell'Insubria, Varese, Lombardia, 4Università di Trento, Trento, Trento, 5Monash University, Melbourne, NA, 6McGill University, Montreal, QC, 7Lariboisière Hospital, Paris, Paris, 8Gui de Chauliac Hospital, Montpellier, Occitania
First Author:
Co-Author(s):
Introduction:
Recent advances in non-invasive mapping techniques have provided invaluable insights into the organizational principles of the anatomo-functional architecture of the human brain. Specifically, the ability to measure spontaneous brain fluctuations using resting-state functional MRI has highlighted how segregation into functionally specialized areas is paralleled by integration into highly connected brain networks, while diffusion weighted imaging and tractography have proved incredibly useful in charting the wiring diagram of the brain.
Even though both non-invasive mapping techniques represent powerful tools to investigate the functional and structural anatomy of the major functional systems of the brain, they have been mainly used to promote a segregated view of the brain, with parallel networks acting in isolation. Instead, adaptive and context specific behavior is thought to emerge from the continuous interaction of large-scale functional systems, suggesting that a 'meta-network' framework is better suited to encapsulate this view of brain functioning.
Leveraging a recently developed computational approach able to integrate direct electrical stimulation (DES, a causal brain mapping technique) with connectomics, here we seek to (de)compose the meta-network of language. We found that spontaneous brain fluctuations in white matter are critical in delineating both the function specific and shared portion of the structural scaffold underlying semantics, phonology and speech articulation, core aspects of language.
Even though both non-invasive mapping techniques represent powerful tools to investigate the functional and structural anatomy of the major functional systems of the brain, they have been mainly used to promote a segregated view of the brain, with parallel networks acting in isolation. Instead, adaptive and context specific behavior is thought to emerge from the continuous interaction of large-scale functional systems, suggesting that a 'meta-network' framework is better suited to encapsulate this view of brain functioning.
Leveraging a recently developed computational approach able to integrate direct electrical stimulation (DES, a causal brain mapping technique) with connectomics, here we seek to (de)compose the meta-network of language. We found that spontaneous brain fluctuations in white matter are critical in delineating both the function specific and shared portion of the structural scaffold underlying semantics, phonology and speech articulation, core aspects of language.
Methods:
We integrated white matter DES points causing transient speech arrest, semantic or phonological aphasia (N=297 patients, 485 stimulations) in combination with resting-state fMRI via lesion network mapping for deriving functional networks (N=1'000 control subjects). To corroborate the use of resting state fMRI in the white matter, we tested whether DES derived networks are predictive of future stimulation points via cross validation, and we investigated – accounting for spatial autocorrelation – the degree of correspondence in the spatial organization of spontaneous brain fluctuations and glucose metabolism (FDG-PET, N=25).
For each functional category (Fig. 1), we defined resting-state fMRI driven white and gray matter hubs that were subsequently used as waypoints for tractography filtering in an independent cohort of control subjects (N=753). Individually filtered tractographies were merged, clustered, and visually inspected to remove artefactual reconstructions. Before clustering, we generated joint structural-functional maps of the white matter via track-weighted functional connectivity, and used them to predict symptom severity in a cohort of aphasic stroke patients via quantile regression (N=105).
For each functional category (Fig. 1), we defined resting-state fMRI driven white and gray matter hubs that were subsequently used as waypoints for tractography filtering in an independent cohort of control subjects (N=753). Individually filtered tractographies were merged, clustered, and visually inspected to remove artefactual reconstructions. Before clustering, we generated joint structural-functional maps of the white matter via track-weighted functional connectivity, and used them to predict symptom severity in a cohort of aphasic stroke patients via quantile regression (N=105).
·Integrating gray and white matter functional connectivity
Results:
White matter DES derived functional networks accurately predict future stimulation points in the white/gray matter (94/64, 95/70, and 96/81% accuracy for phonological, semantics, and speech arrest, respectively). After FDR correction, 25/40 regions showed a statistically significant correlation between spontaneous brain oscillations in the white matter as measured by resting state fMRI and FDG-PET derived glucose metabolism.
Our functionally driven tractography filtering procedure revealed both subnetwork specific connectivity signatures and the existence of multiple integration points across the three functional categories.
Notably, we found that the overlap between lesion and joint structural functional white matter maps of semantics and phonological thresholded at the 95th percentile significantly predicted symptom severity in aphasic stroke patients (r2=0.34, r2=0.35, pval < 0.001 after permutation testing), well beyond total lesion size (r2=0.14, pval < 0.001).
Our functionally driven tractography filtering procedure revealed both subnetwork specific connectivity signatures and the existence of multiple integration points across the three functional categories.
Notably, we found that the overlap between lesion and joint structural functional white matter maps of semantics and phonological thresholded at the 95th percentile significantly predicted symptom severity in aphasic stroke patients (r2=0.34, r2=0.35, pval < 0.001 after permutation testing), well beyond total lesion size (r2=0.14, pval < 0.001).
Conclusions:
Our results suggest that language is a complex function subserved by specific subnetworks strategically wired to act in cooperation, supporting the adoption of a 'meta-network' framework to understand complex cognitive functions.
Brain Stimulation:
Direct Electrical/Optogenetic Stimulation
Language:
Language Other 2
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Anatomy and Functional Systems 1
White Matter Anatomy, Fiber Pathways and Connectivity
Keywords:
Computational Neuroscience
Data analysis
FUNCTIONAL MRI
Language
Neoplastic Disease
Neurological
Statistical Methods
Tractography
1|2Indicates the priority used for review
Provide references using author date format
Coletta, L., Avesani, P., …, Sarubbo, S. (2023), ‘Integrating direct electrical brain stimulation with the human connectome’. Accepted for publication in Brain.
Hayashi S., et al., (2023), ‘brainlife.io: A decentralized and open source cloud platform to support neuroimaging research’. Accepted for publication in Nature Methods.
Porro-Muñoz, D., …, Avesani, P. (2015), ‘ Tractome: a visual data mining tool for brain connectivity analysis’, Data Mining and Knowledge Discovery, vol. 29, pp. 1258–1279.
Calamante, F., …, Cannelly, A., (2013), ‘Track-weighted functional connectivity (TW-FC): A tool for characterizing the structural–functional connections in the brain’, NeuroImage, vol. 70, pp. 199-210.