The interplay between information flux and temporal dynamics in infraslow frequencies

Abstract

The brain exhibits both spatial and temporal hierarchy with their relationship remaining an open question. We address this issue by investigating the brain's spatial hierarchy with complexity, i.e., Lempel-Zev Complexity (LZC) and temporal dynamics, i.e., median frequency (MF) in rest/task fMRI (including replication data). Our results are: (I) topographical differences in rest between higher-order networks (lower LZC and MF) and lowerorder networks (higher LZC and MF); (II) task-specific increases and task-unspecific decreases in LZC and MF; (III) non-linear topographical relationship with low MF mediating higher LZC rest-task changes as confirmed in various simulations. Together, we demonstrate convergence of spatial (LZC) and temporal (MF) hierarchies in a non-linear topographical way along the lines of higher-order/slow frequency and lowerorder/fast frequency networks.The main aim of our study was to investigate the convergence of spatial and temporal topographies including how that shapes rest and task states. For that purpose, we used the Human Connectome Project (HCP) 7 Tesla fMRI datasets (and the HCP 3T for replication) applying novel measures like Lempel-Ziv complexity (LZC) and median frequency (MF) during different states, i.e., rest and two different task states with two completely different complexity and temporal structure (i.e., movie and retinotopy) (see below).LZC measures the number of distinct patterns in a binary sequence i.e., the regularity or repetitiveness of a signal 34,35 , which reflects the amount of information (number of bits) required to reconstruct a signal 34 . Recently, LZC has been applied in fMRI 36-39 as well as in other imaging modalities including MEG 40-43 and wn spatial 2,43,56 , we h rest and 1,2,12,32,33 , i.e., more toparietalThe second specific aim was to investigate regional temporal dynamics by using median frequency (MF) of ISF in rest and task states. Recent studies demonstrated that lower-and higher-order regions exhibit different intrinsic neural time scales with short and long durations in their temporal receptive windows during task states 7,8,[17][18][19]32,33,[9][10][11][12][13][14][15][16] . Based on these findings and the fact that the duration of intrinsic neural time scale may be related to the length of cycle durations 57 as reflecting the frequency range 7,8 , we hypothesized that lower-order networks would exhibit higher MF, i.e., stronger power in faster ISF frequency ranges with shorter cycle duration. While we assumed that higher-order ones would show lower MF with stronger power in slower ISF frequency ranges, i.e., longer cycle durations, in both rest and task states.The third specific aim consisted in directly linking spatial, i.e., LZC, and temporal, i.e., MF dimensions including their respective topographies. For that purpose, we conducted correlation and regression models as well as simulation. Given the scale-free driven discrepancy in power of slower (strong power) and faster (weaker power) ISF frequency ranges 7,13,58,59 , we assumed non-l...

Topographic gradients of intrinsic dynamics across neocortex

“…Our results demonstrate that regional haemodynamic activity, often overlooked in favour of electrophysiological measurements with greater temporal resolution, possesses a rich dynamic signature [17,44,69,71,90,109]. While multiple reports have suggested the existence of a timescale or temporal receptive window hierarchy [48,53,59,62,67,84,120], these investigations typically involved (a) incomplete spatial coverage, making it difficult to quantitatively assess correspondence with other microscale and macroscale maps, and (b) a priori measures of interest, such as spectral power or temporal autocorrelation, potentially obscuring other important dynamical features. Here we comprehensively benchmark the entire dynamic profile of the brain, by nearexhaustively estimating 6000+ features from the wider time-series literature.…”

Intracortical recordings reveal Vision-to-Action cortical gradients driving human exogenous attention

“…TRW length was assessed by computing the across-trial autocorrelation (Golesorkhi et al, 2020; Murray et al, 2014) of the non-filtered iEEG signal (down-sampled to 100Hz), for each of the contacts in the three target-locked clusters. An exponential decay function (e^(−t/τ)) was fit to the contacts autocorrelation coefficient across time-lags.…”

“…An exponential decay function (e^(−t/τ)) was fit to the contacts autocorrelation coefficient across time-lags. TRW length for each contact was defined as the time constant (τ) of the contact’s fitted exponential decay function, i.e., the time it takes for the autocorrelation to decrease by a factor of e (Golesorkhi et al, 2020; Murray et al, 2014).…”

Default mode network mediates low‐frequency fluctuations in brain activity and behavior during sustained attention

“…The DMN does show strong low‐frequency fluctuations in exactly the frequency range of RT‐fluctuation (0.01–0.1 Hz) during attention tasks (Fox & Raichle, 2007 ; Golesorkhi et al, 2020 ; Raichle et al, 2001 ). We, therefore hypothesize that the low‐frequency fluctuations in DMN are related to corresponding low‐frequency fluctuations (0.01–0.1 Hz) in behavior, that is, RT, during sustained attention.…”

“…Recent investigations show that the magnitude of DMN's activity is correlated with changes in behavior over time (Esterman et al, 2013 ; Kucyi et al, 2016 ; Kucyi et al, 2017 ). Given that the DMN during rest displays strong fluctuations as observed in both low‐frequency range (Fox & Raichle, 2007 ; Golesorkhi et al, 2020 ; Raichle et al, 2001 ) and long timescales (Golesorkhi et al, 2021 ; Ito et al, 2020 ; Raut et al, 2020 ), it may be considered a suitable neural candidate for the low‐frequency fluctuations in sustained attention. Probing the role of DMN in the fluctuation of sustained attention by investigating its neural fluctuation (fMRI) is the main objective of our study.…”