Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Functional connectivity underlying speech monitoring processes as revealed by Graph Signal Processing
Poster A6 in Poster Session A, Tuesday, October 24, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
This poster is part of the Sandbox Series.
Yusheng Wang1,2, Elizabeth Anderson1,2, Sophie Kajfez2, Carrie R. McDonald1, David J. Lee1, Leena Kansal1, June Yoshii-Contreras1, Jerry J. Shih1, Sharona Ben-Haim1, Eric Halgren1, Ashkan Ashrafi2, Stephanie K. Riès2; 1University of California San Diego, 2San Diego State University
Speech monitoring has been the subject of extensive research due to its importance for accurate speech production. Several brain regions, including the left posterior temporal gyrus, medial frontal cortex, and inferior frontal gyrus, have been associated with speech monitoring processes. However, how these brain regions interact with one another during speech monitoring has been understudied. Our study examines how these different brain regions are functionally connected to support speech monitoring in individuals with intractable epilepsy undergoing a picture-word interference (PWI) experiment using an innovative Graph Signal Processing (GSP) technique. Whereas traditional functional connectivity analyses use multiple pairwise correlations measurement of the temporal synchrony between the signal recorded in different regions, the GSP approach considers the signal space as a graph which allows for the analysis of signals defined on the entire graph, thereby bypassing the multiple comparison issues. Moreover, GSP considers all of the signal information, including phase, and amplitude, thereby bypassing the need to focus on a restricted part of the signal. Therefore, this novel GSP approach can reveal connectivity patterns that may not be detectable with traditional functional connectivity analysis. Our study utilizes intracranial EEG (iEEG), a brain imaging tool that integrates excellent spatial and temporal resolution, to investigate the brain network supporting speech monitoring in eleven individuals (mean age = 29 years, SD = 7.68 years) with intractable epilepsy as they engaged in a PWI task. Individuals were asked to name pictures while ignoring the overlayed distractor words in three conditions (i.e., semantically related, semantically unrelated, or identical to the picture name). The result shows a main effect of condition on reaction time (χ2(1,11) = 105.8, p < .001) and error rate (χ2(1,11) = 11.9, p =.003). Post hoc analyses indicated significantly faster RTs in the identity than related (t = -9.94, p < .0001), or unrelated conditions (t = -7.02, p < .0001), and significantly faster RTs for the unrelated than related condition (t = 2.87, p = .01). There are significantly fewer errors in the identity than related (z = -3.3, p = .0031) or unrelated conditions (z = -3.3, p =.0031), but no difference between related and unrelated conditions (z = -0.006, p > .05). Our GSP predictions are as follows based on the above results, we hypothesize that the connection strengths and the number of connected nodes between the different brain regions supporting speech monitoring (i.e., showing larger activity response-locked in error than correct trials and in related than unrelated trials) will be higher in errors than in correct trials and in the related than unrelated condition. This is an ongoing study. We believe that the prospective outcomes of this study will not only shed light on understanding the neural underpinnings of speech monitoring but also could further support the use of GSP for the analysis of functional connectivity in language and cognitive neuroscience more generally.
Topic Areas: Language Production, Control, Selection, and Executive Processes