Differential impact of sleep deprivation conditions on resting-state effective connectivity.

1563 

Abstract Submission 

Edoardo Bettazzi¹, Alexander Silchenko¹, David Elmenhorst^2,3, Simon Eickhoff^1,4, Masoud Tahmasian^1,3,4, Felix Hoffstaedter^1,4

¹Institute of Neuroscience and Medicine (INM-7), Research Centre Jülich, Jülich, NRW, Germany, ²Institute of Neuroscience and Medicine (INM-2), Research Centre Jülich, Jülich, NRW, Germany, ³Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne, Cologne, NRW, Germany, ⁴Institute for Systems Neuroscience, Medical Faculty, Heinrich-Heine University Düsseldorf, Düsseldorf, NRW, Germany

Edoardo Bettazzi  
Institute of Neuroscience and Medicine (INM-7), Research Centre Jülich
Jülich, NRW, Germany

Alexander Silchenko  
Institute of Neuroscience and Medicine (INM-7), Research Centre Jülich
Jülich, NRW, Germany

David Elmenhorst  
Institute of Neuroscience and Medicine (INM-2), Research Centre Jülich|Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne
Jülich, NRW, Germany|Cologne, NRW, Germany

Simon Eickhoff  
Institute of Neuroscience and Medicine (INM-7), Research Centre Jülich|Institute for Systems Neuroscience, Medical Faculty, Heinrich-Heine University Düsseldorf
Jülich, NRW, Germany|Düsseldorf, NRW, Germany

Masoud Tahmasian  
Institute of Neuroscience and Medicine (INM-7), Research Centre Jülich|Department of Nuclear Medicine, University Hospital and Medical Faculty, University of Cologne|Institute for Systems Neuroscience, Medical Faculty, Heinrich-Heine University Düsseldorf
Jülich, NRW, Germany|Cologne, NRW, Germany|Düsseldorf, NRW, Germany

Felix Hoffstaedter  
Institute of Neuroscience and Medicine (INM-7), Research Centre Jülich|Institute for Systems Neuroscience, Medical Faculty, Heinrich-Heine University Düsseldorf
Jülich, NRW, Germany|Düsseldorf, NRW, Germany

Sleep loss affects the human brain at multiple levels and leads to various cognitive dysfunctions, including attention lapses, impaired working memory and emotional hyperreactivity. Experimentally induced sleep deprivation (SD) is an interventional approach to investigate how the brain responds to sleep loss. SD affects the functional connectivity of the intrinsic neural networks of the brain, including the Salience Network (SN) and Central Executive Network (CEN). A previous neuroimaging meta-analysis identified reduced activity in the right intraparietal sulcus (rIPS). Thus, we investigated the influences of SD on the effective connectivity (EC) of SN and CEN, each with the rIPS.

We included resting-state fMRI across three conditions of SD: Stockholm SleepyBrain dataset (N=40, age 20-30 yrs; TR 2.5s, 193 volumes), acute partial SD (3-hours sleep); PETcoffee dataset (N=36, age 22-37 yrs; TR 2.29s, 262 volumes), chronic partial SD (5-hours sleep, 5 nights); Somnosafe dataset (N=35, age 20-39 yrs; TR 2s, 146 volumes), total SD (0-hours sleep). fMRI data was preprocessed using fMRIPrep in the FAIRly big workflow with Datalad, before time-series extraction based on Nilearn (version 0.9.2) and deploying DCM in SPM12 (version 7912) with Matlab, implemented in a custom-built pipeline (Fig.1). AROMA denoised first eigenvariate time series were extracted 12 SN and 13 CEN nodes from the Schaefer100 parcellation and the rIPS region, for SD and normal sleep (NS). The Spectral DCM framework was used to model EC of (1.) the SN with rIPS as well as (2.) the CEN with rIPS. Parametric Empirical Bayes (PEB) was then applied to estimate the effects of SD in contrast to NS. For each SD condition, we quantified EC patterns of each sleep state and their differences, which were filtered for significant state effects, excluding connectivity differences not present in either SD or NS. Resulting state-informed difference matrices were tested for pairwise correlations between the 3 SD conditions, yielding coefficients of <0.1 for all pairs. Finally, for heuristic comparison between SD conditions, two network-level metrics were computed: the absolute sum over connectivity values, to quantify the impact of treatment (i.e. the amount of change in EC); and the sum over connectivity values, to evaluate the network balance after treatment (i.e. overall increase/decrease in EC).

PEB reveals altered EC patterns in both neural networks after all SD conditions (Fig. 2). Regarding SN, the impact of SD on the network and the resulting balance were the following, acute partial SD: 0.84/+0.23; chronic partial SD: 0.24/+0.09; total SD: 0.89/+0.26. Effects observed in CEN were acute partial SD: 2.52/+0.11; chronic partial SD: 0.91/+0.04; total SD: 1.35/-0.22. The chronic partial SD treatment with 5h of sleep for 5 nights induced the smallest change in connectivity strength for both networks. The most impactful treatment was partial acute SD, with 3 hours of sleep, on the CEN, while the SN was altered generally less and similarly by acute partial and total SD. Finally, in total deprivation, EC of the rIPS to nodes of both CEN (-0.097) and SN (-0.043) is decreased, while EC toward the rIPS is increased with only 3h of sleep time. Overall, CEN to rIPS connections show the biggest change in EC for acute (0.362), chronic partial (0.155) and total SD (0.093).

SD differentially impacts resting-state networks depending on the amount of sleep, with acute (total and partial) SD showing more impact than chronic partial SD. As expected from previous reports of rIPS hypoactivity, PEB reveals a decreasing influence of rIPS on both networks in total SD. Since the rIPS is regarded as an attentional hub, the observed changes in EC from and to this region could disrupt communication between networks and impair the allocation of attentional resources.

Bayesian Modeling ²

Connectivity (eg. functional, effective, structural) ¹

Task-Independent and Resting-State Analysis

Workflows

Sleep and Wakefulness

Data analysis

FUNCTIONAL MRI

Modeling

Sleep

Statistical Methods

Systems

Other - Dynamic Causal Modeling (DCM)