Single voxel autocorrelation uncovers gradients of temporal dynamics in the hippocampus and entorhinal cortex during rest and navigation

Abstract

During navigation, information at multiple scales needs to be integrated. Single-unit recordings in rodents suggest that gradients of temporal dynamics in the hippocampus and entorhinal cortex support this integration. In humans, gradients of representation are observed, such that granularity of information represented increases along the long axis of the hippocampus. The neural underpinnings of this gradient in humans, however, are still unknown. Current research is limited by coarse fMRI analysis techniques that obscure the activity of individual voxels, preventing investigation of how moment-to-moment changes in brain signal are organized and how they are related to behavior. Here, we measured the signal stability of single voxels over time to uncover previously unappreciated gradients of temporal dynamics in the hippocampus and entorhinal cortex. Using our novel, single voxel autocorrelation technique, we show a medial-lateral hippocampal gradient, as well as a continuous autocorrelation gradient along the anterolateral-posteromedial entorhinal extent. Importantly, we show that autocorrelation in the anterior-medial hippocampus was modulated by navigational difficulty, providing the first evidence that changes in signal stability in single voxels are relevant for behavior. This work opens the door for future research on how temporal gradients within these structures support the integration of information for goal-directed behavior.

Single voxel autocorrelation reflects hippocampal function in temporal lobe epilepsy

“…Consistent with findings from our previous work (Bouffard, Golestani et al, 2023;Coughlan et al, 2023), we found that data-driven clustering of the signal voxel autocorrelation in TLE patients and controls resulted in three distinct clusters. Our clustering approach does not use a predefined number of clusters, and visual examinations of the cluster maps revealed that there were no differences in number of clusters between patients and controls, nor was there a difference between healthy and epileptogenic hemispheres.…”

“…In this study we examined single voxel autocorrelation and clusters of single voxel autocorrelation in patients with unilateral left and right TLE to investigate how hippocampal damage present in TLE affects the temporal signal dynamics in the hippocampus. Here, we replicated our past work (Bouffard, Golestani et al, 2023;Coughlan et al, 2023) and found that all groups demonstrated a pattern of high autocorrelation values clustered in the anterior-medial hippocampus and low autocorrelation values clustered in the posterior-lateral. There were two major findings in the TLE groups, both selective to the anterior-medial cluster.…”

“…In past work, the intermediate cluster has been shown to have low reliability within participants and includes significantly fewer voxels compared to the anterior-medial and posterior-lateral clusters (Bouffard, Golestani et al, 2023;Coughlan et al, 2023). Here we again replicate this finding and found that the intermediate cluster had low reliability (See Supplemental Materials Figure S1) and was the smallest cluster (See Supplemental Materials Table S1 for list of cluster sizes).…”

“…In previous work correlating the temporal dynamics of the BOLD signal with behavior, task-dependent changes in signal dynamics have been associated with better cognitive performance (Armbruster-Genç et al, 2016;Bouffard, Golestani et al, 2023;Protzner et al, 2013). We had predicted that our measure, single voxel autocorrelation, could be an indicator of this aspect of functional integrity and thus would be related to patient's memory performance.…”

“…In the healthy hippocampus, single voxel autocorrelation follows a gradient from high-to-low along the anterior-posterior and medial-lateral axes Coughlan et al, 2023;Raut et al, 2020). These continuous gradients can be segmented into distinct clusters based on the autocorrelation of individual voxels and result in three segments: an anterior-medial cluster of high autocorrelation voxels, an intermediate cluster, and a posterior-lateral cluster of low autocorrelation voxels (Bouffard, Golestani et al, 2023;Coughlan et al, 2023). Our past work used a datadriven clustering approach and found that changes in single voxel autocorrelation, specifically in the anterior-medial cluster, were related to changes in navigation difficulty.…”