Search Abstracts | Symposia | Slide Sessions | Poster Sessions
ECoG Reveals Dynamics of Cortical Connectivity and Semantic Representations During Speech Preparation
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
Itsuki Hamada1, Atsushi Miura1, Manabu Tanifuji1, Yuichi Kubota2, Hiromu Sakai1; 1Waseda University, 2Tokyo Women's Medical University
Frontal–parietal connectivity has long been considered important for the neural architecture of speech and language and has classically been linked to the dorsal language pathway, including the arcuate fasciculus and superior longitudinal fasciculus (Catani et al., 2005). Resting-state fMRI has also characterized a distributed language network encompassing frontal, temporal, and parietal regions (Tomasi & Volkow, 2012). However, these approaches provide limited information about how such connectivity evolves during the brief period from stimulus presentation to overt articulation. It also remains unclear whether its temporal dynamics are related to information about the upcoming spoken response or instead reflect more general coordination. Because the stimuli in the present tasks were organized into semantic categories, category information provided an experimentally tractable index of task-relevant speech content. We therefore used electrocorticography (ECoG), which directly measures cortical population activity with high temporal resolution, to examine cortical connectivity dynamics and semantic category information during picture naming and word reading aloud. The participant was a 68-year-old right-handed native Japanese-speaking man with a glioma in the left frontal lobe. A 36-channel intracranial electrode array covered the inferior parietal lobule (IPL), middle frontal gyrus (MFG), and inferior frontal gyrus (IFG). Electrical stimulation of channels 25–26 induced speech arrest, confirming the functional relevance of the implanted region to speech production. The participant performed picture-naming and word-reading-aloud tasks using the same 32 items from four semantic categories. After common average referencing, we extracted the high-gamma envelope from 70–160 Hz activity using a Gaussian filter bank and the Hilbert transform. Time-resolved functional connectivity (FC) was estimated as sliding-window Pearson correlations between trial-wise high-gamma envelopes, separately for stimulus-locked and speech-onset-locked epochs, with significance assessed by permutation testing corrected for multiple comparisons. Four-class category decoding was performed within four frontal electrode groups—anterior MFG, posterior MFG, anterior IFG, and posterior IFG—using raw ECoG signals with a sliding-window support vector machine and five-fold cross-validation. FC analysis revealed characteristic frontal–parietal connectivity involving the anterior MFG and anterior IFG groups. The anterior MFG showed persistent FC with the IPL across tasks, but category information could not be decoded from this group. In contrast, the anterior IFG showed stimulus-locked FC with the IPL, increasing after trial onset and differing in temporal profile between picture naming and word reading aloud; category information was also decodable from this group in the picture condition. The posterior MFG was not detected as an independent long-range FC cluster, yet category information was decodable in both tasks. The posterior IFG was also not detected as a robust long-range FC cluster, but showed the highest category decoding accuracy in the picture condition. These findings indicate that long-range frontal–parietal FC during speech preparation was not uniform. The anterior MFG–IPL connection showed persistent FC without decodable category information, whereas the anterior IFG–IPL connection showed stimulus-locked FC accompanied by category information in the picture condition. Thus, ECoG revealed temporally distinct frontal–parietal connectivity dynamics during speech preparation and their different relationships to category information indexing task-relevant speech content.
Topic Areas: Language Production,