The Dynamics of Speech Motor Control Revealed with Time-Resolved fMRI

Abstract

Holding a conversation means that speech must be started, maintained, and stopped continuously. The brain networks that underlie these aspects of speech motor control remain poorly understood. Here we collected functional magnetic resonance imaging (fMRI) data while participants produced normal and fast rate speech in response to sequences of visually presented objects. We took a non-conventional approach to fMRI data analysis that allowed us to study speech motor behavior as it unfolded over time. To this end, whole-brain fMRI signals were extracted in stimulus-locked epochs using slice-based fMRI. These data were then subjected to group independent component analysis to discover spatially independent networks that were associated with different temporal activation profiles. The results revealed two basic brain networks with different temporal dynamics: a cortical network that was activated continuously during speech production, and a second cortico-subcortical network that increased in activity during the initiation and suppression of speech production. Additional analyses explored whether key areas involved in motor suppression such as the right inferior frontal gyrus, sub-thalamic nucleus and pre-supplementary motor area provide first-order signals to stop speech. The results reveal for the first time the brain networks associated with the initiation, maintenance, and suppression of speech motor behavior.

Exploring the temporal dynamics of speech production with EEG and group ICA

“…First, it is worth pointing out the diversity of temporal profiles that underlie the production of speech. For example, for all ICs there is strong synchronized activity across trials, but for some this synchronized activity is of a rather short duration (IC5, 8,9), for others this synchronized activity has a longer duration (IC6, 10,11). Similarly, for some regions activity has a sharp rising edge while for others there is a more gradual slope (cf., IC5 and IC6).…”

“…The anatomical sources and their temporal dynamics observed in the current study correspond closely to those reported in previous studies. Previous hemodynamic studies of picture naming have typically reported activity in medial and lateral regions of the occipital cortex, in medial inferior and middle gyri of the temporal cortex, in posterior medial and inferior lateral portions of the parietal cortex, and in medial anterior (cingulate) and in medial and lateral (motor) regions of the frontal cortex 8,[10][11][12]57,58 . The current study found activations in occipital (cuneus, lateral, fusiform), temporal (inferior, middle), frontal (medial cingulate, lateral), and parietal areas (inferior, superior) and thus align with those previously reported.…”

“…Previous Positron Emission Tomography (PET) and functional Magnetic Resonance Imaging (fMRI) studies have repeatedly shown that picture naming yields activation in occipital, temporal, frontal and parietal areas of the cortex, as well as in the striatum, thalamus and brain stem of the subcortex e.g. [8][9][10][11][12] . Although these studies are informative about the location of brain activity underlying picture naming, other techniques such as Magnetoencephalography (MEG) and Electroencephalography (EEG) are needed to examine the precise temporal dynamics of neural activity underlying the task.…”

Medial temporal lobe contributions to resting-state networks

“…Finally, we relied on the spatially restricted group ICA technique (srICA; Blessing et al 2016;Ezama et al 2021;Formisano et al 2004). Whole-brain ICA is frequently used to separate signal from noise in fMRI studies (e.g., Janssen and Mendieta 2020;Smith et al 2013). Instead, by applying ICA to a particular brain region, the noise profile that is specific to that brain region will be taken into account and result in a more sensitive separation of signal from noise in that region.…”

A method to mitigate spatio‐temporally varying task‐correlated motion artifacts from overt‐speech fMRI paradigms in aphasia

“…More recently, Xu et al (2014) developed a more sophisticated dual‐masked s‐ICA technique to denoise TCM in an overt‐language fMRI paradigm collected on healthy young participants. In the realm of routine bulkhead motion correction, Janssen and Mendieta (2020)) employed the standard ICA‐based denoising package (i.e., AROMA) to correct for motion in an overt picture naming task collected on healthy young subjects, while (Sebastian et al, 2016) utilized a rigid body alignment approach (i.e., MCFLIRT) to correct for motion in an overt picture naming task collected from acute stroke patients. However, given that speech related motion is not just spatially selective, but also temporally correlated with task‐induced hemodynamic response, and can vary from epoch to epoch (Gopinath et al, 2009), optimization of TCM denoising approaches to account for the noise jointly encoded in both spatial and temporal domains is equally important.…”