Temporal Characteristics of Speech in Cortical and Subcortical Recordings

Introduction: Speech requires complex motor planning for activating coordinated jaw, tongue, lips, and larynx muscles. Motor-associated areas of the cortex, particularly frontal regions, are known to drive speech motor planning and execution. However, little is known about subcortical regions' roles in speech production. The basal ganglia is one subcortical region engaged in motor planning. Disorders of the basal ganglia, such as Parkinson's disease (PD), can lead to significant speech production-related issues (e.g., dysarthria, monotone). Furthermore, electrophysiological recordings of the subthalamic nucleus — a subregion within the basal ganglia and a common stimulation target for the surgical management of PD — have shown a decrease in 15–30 Hz "beta" power (Hebb et al., 2012) and an increase in 70–150 Hz "high-gamma" power (Chrabaszcz et al., 2019) during speech production.
Nevertheless, the relationship between the cortical and subcortical regions in speech production remains poorly understood. In this study, we utilize simultaneous local field potential recordings from the cortical and subcortical regions of volunteers undergoing awake Deep Brain Stimulation (DBS) surgery for PD and Essential Tremor (ET) during a three-speech-syllable repetition task. We investigated possible communication mechanisms by calculating synchronization between the cortical and subcortical regions and identifying the direction of any information transfer during speech production.

Materials and

Methods: Neural recordings were collected from deep brain stimulator electrode contacts (n=3 Boston Scientific Vercise Directed; n=1 Abbot 6172) and a 64-channel electrocorticography array (2110-64-154 PMT Electrode Gride 21x3, PMT Corporation, Chanessen, MN, US) from 4 volunteers undergoing DBS surgery (n=3 PD; n=1 ET). Participants completed a three-syllable repetition of non-word consonant-vowel pairs selected from 9 possible syllables (e.g., \ghah\, \tee\, \voo\) for a total of 144 trials. Recordings were done during awake implantation of the deep brain leads and were sampled at 1kHz (Ripple Grapevine system, Salt Lake City, Utah, US). The data were evaluated by visual quality control, preprocessed, and bipolar re-referenced to account for volume conduction. We quantified the synchronization between cortical and subcortical regions for each frequency band of neural activity based on the degree of phase locking determined from the Morlet Wavelet Transform. The magnitude of the exponentiated phase difference between cortical and subcortical contacts at each point in time represents the phase locking value (PLV). The PLV ranges from 0 to 1, with 0 representing no synchronization and 1 representing perfect synchronization. We compared differences in PLV before and after the onset of speech production using a non-parametric t-test with cluster-linked permutation testing. Furthermore, we estimated information flow by calculating Granger Causality (Multivariate Granger Causality (MVGC) Toolbox in MATLAB).

Results, Conclusions, and Discussions: Comparing the PLV before and after the onset of speech production, one PD volunteer showed increased synchronization in the high gamma range (p-value < 0.001) and desynchronization in the beta range (p-value < 0.05) range. The other two PD volunteers showed similar trends but with non-significant findings. There were no significant differences for the ET volunteer, whose deep-brain electrode was implanted in the ventral intermediate nucleus (VIM). Using Granger Causality between electrodes from cortical and subcortical contacts, all combinations of contacts were found to predict the flow of information from cortical to subcortical (p-value < 0.001). This unidirectional causality is consistent across subjects.
In this study, we aimed to characterize the temporal complexity of local field potentials from cortical and subcortical brain regions during a three-syllable repetition task in patients undergoing awake DBS implantation. We found a high gamma synchronization and a beta desynchronization pattern across participants with PD. Given that these patterns were found across brain regions, these regions are likely involved in common cognitive processing, such as the onset of speech production. However, when analyzing the ET patient with the implanted electrode in the VIM, there were no patterns found. Using PLV, the temporal characteristic of signals can be analyzed to understand the timing of communication between brain regions, allowing for temporal characterization. Furthermore, we found the signal in cortical electrodes to predict the signal in subcortical contacts through Granger Causality measures.
In summary, PLV is an informative measure of communication between cortical and subcortical regions in the brain. Investigating the temporal characterization of neural signals can contribute to our understanding of mechanisms underlying speech disorders.

Acknowledgements (Optional):