The domain-separation low-dimensional language network dynamics in the resting-state support the flexible functional segregation and integration during language and speech processing

Abstract

Modern linguistic theories and network science propose that the language and speech processing is organized into hierarchical, segregated large-scale subnetworks, with a core of dorsal (phonological) stream and ventral (semantic) stream. The two streams are asymmetrically recruited in receptive and expressive language or speech tasks, which showed flexible functional segregation and integration. We hypothesized that the functional segregation of the two streams was supported by the underlying network segregation. A dynamic conditional correlation approach was employed to construct frame-wise time-varying language networks and investigate the temporal reoccurring patterns. We found that the time-varying language networks in the resting-state robustly clustered into four low-dimensional states, which dynamically reconfigured following a domain-separation manner. Spatially, the hub distributions of the first three states highly resembled the neurobiology of primary auditory processing and lexical-phonological processing, motor and speech production processing, and semantic processing, respectively. The fourth state was characterized by the weakest functional connectivity and subserved as a baseline state. Temporally, the first three states appeared exclusively in limited time bins (~15%), and most of the time (> 55%), the language network kept inactive in state 4. Machine learning-based dFC-linguistics prediction analyses showed that dFCs of the four states significantly predicted individual linguistic performance. These findings suggest a domain-separation manner of language network dynamics in the resting-state, which forms a dynamic meta-networking (network of networks) framework.

How do we imagine a speech? A triple network model for situationally simulated inner speech

“…The third CAP mainly included the seed points and the language network (LAN) (Figures 3 and 4). LAN predominantly consists of the left superior temporal gyri, superior temporal gyri, inferior frontal gyri, and ventral precentral gyri, as documented by various studies (Lipkin et al, 2022;Malik-Moraleda et al, 2022;Yuan, Xie, & Wang et al, 2023c). A high spatial similarity was observed between the time-recurring states across the 10 story scans (Supplementary Figure 4).…”

“…Secondly, an exponentially weighted moving average (EWMA) window is applied to the standardized residuals to compute a non-normalized version of the time-varying correlation matrix between Regions of Interest (ROIs). Mathematical expressions of the GARCH (1,1) model, DCC model, EWMA, and estimations of the model parameters can be found in (Lindquist et al, 2014) and related literature (Yuan, Xie, & Gong et al, 2023;Yuan, Xie, & Wang et al, 2023b).…”

“…This task necessitated not only retrieving word meanings but also employing them to weave into a complex ongoing stream of thought. To unveil the transient brain activations and network reconfigurations during these processes, we employed co-activation pattern (CAP) analysis (Bolton et al, 2020;Liu & Duyn, 2013;Liu et al, 2018) and dynamic conditional correlation (Aielli, 2013;Engle, 2002;Yuan et al, 2023aYuan et al, , 2023b. The core brain regions utilized for CAP analysis were derived from a term-based meta-analyses of 14 inner-speech-related terms, and further corroborated through.an independent fMRI dataset involving covert speech of words and numbers.…”

Compensatory Mechanisms for Preserving Speech-in-Noise Comprehension Involve Prefrontal Cortex in Older Adults

“…At the listener’s side, the clusters covered an extensive range of brain regions, including the wide range of the prefrontal cortex (CH2, CH5-16) and the classical language regions. Based on the anatomical localization and functional organization within the language network ( 35, 36, 39, 40 ), these language regions were divided into ventral language regions, including right STG, MTG (CH29/31/35), bilateral inferior parietal lobe (IPL, CH23/24/26/33/34/36), and dorsal language regions, including bilateral IFG (CH17/27), bilateral pre- and post- central gyrus (pre-/postCG, CH18/20/22/28/30/31). The channel and localization of these channel combinations are shown in Figure 3A and S1A.…”

A systematic evaluation of dynamic functional connectivity methods using simulation data

“…Network reorganization or dynamic functional connectivity (dFC) can be estimated frame by frame or within a short window (~ 20-30 time points). Numerous dFC methods have been proposed, such as the slidingwindow functional connectivity with L1-regularization (SWFC) (Allen et al, 2014;Shakil et al, 2016), dynamic conditional correlation (DCC) (Choe et al, 2017;Lindquist et al, 2014;Yuan et al, 2023a;Yuan et al, 2023b), Multiplication of Temporal Derivatives (MTD) (Shine et al, 2015), Flexible Least Squares (FLS) (Kalaba and Tesfatsion, 1989;Liao et al, 2014), general linear Kalman filter (Kang et al, 2011;Milde et al, 2010), Hidden Markov models (HMM) (Chen et al, 2022;Eavani et al, 2013;Vidaurre et al, 2017), and Hidden semi-Markov models (HSMM) (Shappell et al, 2019). Although these methods have demonstrated their accuracy, reliability, and potential caveats, a comprehensive comparison regarding the efficacy of these methods is lacking.…”