Paul Manuel Müller, Christian Meisel
Abstract
The ability of neural circuits to integrate information over time and across different cortical areas is believed an essential ingredient for information processing in the brain. Temporal and spatial correlations in cortex dynamics have independently been shown to capture these integration properties in task-dependent ways. A fundamental question remains if temporal and spatial integration properties are linked and what internal and external factors shape these correlations. Previous research on spatio-temporal correlations has been limited in duration and coverage, thus providing only an incomplete picture of their interdependence and variability. Here, we use long-term invasive EEG data to comprehensively map temporal and spatial correlations according to cortical topography, vigilance state and drug dependence over extended periods of time. We show that temporal and spatial correlations in cortical networks are intimately linked, decline under antiepileptic drug action, and break down during slow-wave sleep. Further, we report temporal correlations in human electrophysiology signals to increase with the functional hierarchy in cortex. Systematic investigation of a neural network model suggests that these dynamical features may arise when dynamics are poised near a critical point. Our results provide mechanistic and functional links between specific measurable changes in the network dynamics relevant for characterizing the brain’s changing information processing capabilities.
Introduction
An essential ingredient for information processing is thought to be the ability of neural circuits to integrate information over appropriate periods of time and across different cortical areas. Characterization of neural systems according to their temporal and spatial correlations (TC and SC, respectively) has consequently provided important insight into their information processing capabilities. For example, in decision-making and working memory tasks [1–4], the ability to integrate over extended periods of time may increase the signal-to-noise ratio and afford to maintain some memory of past activity. In non-human primates, temporal correlations have been found to increase along the functional hierarchy providing a unifying principle for information integration across different timescales [1,5,6]. Similarly, the ability to integrate information in space, across functionally specialized regions, is considered essential for normal brain functioning [7–9]. Consequently, effective propagation and integration of information in space has been shown to depend on vigilance state where it is maximized during wake and breaks down during slow-wave sleep [10].
Methods
Ethics statement
Multi-day invasive electroencephalographic (iEEG) recordings of 23 patients with epilepsy undergoing presurgical evaluation at the Epilepsy Center of the University Hospital of Freiburg, Germany, were analysed (12 female, age 28±13 years, mean ± standard deviation). The data set was made available in the Epilepsiae database, their use for research was approved by the ethics committee of the University of Freiburg and written informed consent that the clinical data might be used and published for research purposes was given by all patients [61]. The study was approved by the local institutional review board (EK 92022019). Detailed data on the included patients can be found in the S2 Table.
Results
We analysed invasive electroencephalography (iEEG) recordings from 23 patients to characterize spatial and temporal correlations (STC) as functions of vigilance state, antiepileptic drug (AED) action, and of functional cortical hierarchy. Broadband γ-power (56–96 Hz) was used as an index of population firing rate near an electrode [14–18] to derive spatial and temporal correlations and based on prior comparative work across broad frequency ranges. In line with these prior studies, we observed this high-frequency domain activity to be best suited to resolve the correlations within cortical dynamics compared to other, lower frequency bands (see S2 and S3 Figs for TC and SC respectively). Specifically, we measured the decay speed of the autocorrelation function of this signal as an index for temporal correlations (TC; [5,19]) and the decay of the pairwise cross-correlation function over distance as an index of spatial correlations (SC; [20–22]; Fig 1).
Discussion
Accumulating evidence suggests temporal and spatial correlations to reflect integration properties of neural circuits in time and space. Using high-density mapping across full days of recording and different cortical regions, we here showed that these indices fluctuate and exhibit systematic co-variation. We observed that temporal correlations align with functional hierarchy in cortex, and that spatio-temporal correlation decline under AED action and break down during SWS. The results may thus have implications for characterizing the brain’s changing information processing capabilities.
Conclusion
While our research highlights the importance to study temporal and spatial correlations over extended periods of time to capture their variability, the sampling across cortical areas is naturally limited by electrode positioning determined by the clinical need. This leads to a relative under-sampling for example of the ACC in comparison to other regions. This sparse spatial sampling also limited us in calculating spatial correlations for individual regions along the hierarchy. Empirical evidence indicative of a strong co-variation between spatial and temporal correlations along with conceptual insights from our model, however, strongly suggest that also spatial correlations should follow a gradient along the functional hierarchy, similar to temporal correlations. Future research should thus address further validation of the spatial structure of temporal and spatial correlations, especially in such parts of the cortex.
Citation: Müller PM, Meisel C (2023) Spatial and temporal correlations in human cortex are inherently linked and predicted by functional hierarchy, vigilance state as well as antiepileptic drug load. PLoS Comput Biol 19(3): e1010919. https://doi.org/10.1371/journal.pcbi.1010919
Editor: Peter Neal Taylor, Newcastle University, UNITED KINGDOM
Received: September 29, 2022; Accepted: February 3, 2023; Published: March 3, 2023
Copyright: © 2023 Müller, Meisel. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data is available within the Epilepsiae data repository (http://epilepsy-database.eu). The code to analysis is available under https://gitlab.com/computational-neurologie/iEEG_STC.git.
Funding: PM and CM are funded by Charité — Universitätsmedizin Berlin. CM and PM are supported by NeuroCure Cluster of Excellence, funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany´s Excellence Strategy EXC 2049-89829218. CM acknowledges support from a NARSAD Young Investigator Grant. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
