Feeling drowsy? Brain structure-function coupling dynamics track alertness during resting-state

1567 

Abstract Submission 

Maria Giulia Preti¹, Dimitri Van De Ville²

¹EPFL / UNIGE, Geneva, Switzerland, ²École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

Maria Giulia Preti  
EPFL / UNIGE
Geneva, Switzerland

Dimitri Van De Ville  
École Polytechnique Fédérale de Lausanne
Lausanne, Switzerland

Resting-state functional magnetic resonance imaging (fMRI) is a well-established methodology to probe brain organization. However, previous evidence showed that different levels of drowsiness, or even sleep, might occur especially in later phases of the resting-state session (i.e., after 8-10 minutes) [1]. While it has been shown that such variations of conscious state are reflected by changes of specific functional network activity [2], it remains an open question whether and how the coupling between brain activity with the underlying structure is affected. By leveraging graph signal processing (GSP), we previously introduced a quantitative method to measure structure-function coupling through the structural decoupling index (SDI), whose temporal fluctuations have not yet been fully investigated. Here, we introduce a new time-resolved measure of SDI, and investigate its changes across the first and second half of the acquisition.

The 100 unrelated healthy participants from the Human Connectome Project (HCP) were considered [3]. Diffusion weighted images as well as two sets of fMRI datasets were considered during resting-state and motor task, with test and retest sessions. Conventional preprocessing pipelines [4] were adopted to obtain regional fMRI timecourses and structural connectomes, using a parcellation with 379 brain regions (360 from Glasser parcellation [5] and 19 subcortical [3]). The GSP pipeline outlined in [4] was used to filter individual fMRI signals into a liberal and a structurally aligned portion. A new metric of temporally resolved SDI was calculated as the instantaneous difference dt in amplitude between the structurally liberal and aligned portion. K-means clustering (k=[5,..,15] were tested, k=5 retained) was then applied to the patterns of dynamic SDI, temporally demeaned. For each subject, the percentage of occurrence of each cluster during the first and second half of the acquisition was computed and compared with a paired two-sample t-test. The intraclass correlation coefficient (ICC) was used to measure the reliability of cluster occurrences across test-retest sessions.

Fig. 1 shows the obtained cluster centroids Ck, ordered by their occurrence percentages along the resting-state acquisition. Given the temporal demeaning operation performed on dt before clustering, these can be interpreted as brain patterns of change in structure-function decoupling with respect to the average pattern (across time). A significant difference between occurrences in the first and second half of the session was found for all the clusters (paired t-tests corr. for multiple comparisons, pcorr < 0.5). We found that a pattern of decoupling (globally in C1 and for somatomotor/visual networks in C4) is most characteristic for the first part, while a significant trend towards increased coupling both globally (C2), and in sensory networks (C3, and C5) occurs in the second half (Fig. 2). This finding indicates higher decoupling for higher level of alertness, presumable at the beginning of the session, versus increased coupling for higher drowsiness. The same coupled networks, in fact, were found in previous work to emerge in resting-state later stages of the acquisition, more prone to falling asleep [1]. The results appear reliable across test-retest sessions (ICC> 0.8 for C1, C2 and C5, C> 0.6 for C3 and C4). As further support of this interpretation, the same analysis didn't lead to any significant differences when applied to motor task session, which are reasonably less subject to changes in alertness by the participants.

We showed for the first time that structure-function coupling can track changes during resting-state, opening the avenue to characterizing different states of consciousness through the structure-function coupling dynamic. Additional analyses on datasets annotated for drowsiness
level and/or sleep stages will be conducted to further confirm the association between coupling/decoupling and drowsiness/alertness.

Connectivity (eg. functional, effective, structural) ¹

Methods Development

Consciousness and Awareness ²

Consciousness

FUNCTIONAL MRI

Tractography

Other - resting-state