Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Phase-amplitude coupling analysis to identify receptive language cortex in epilepsy patients
Poster Session D, Saturday, September 13, 5:00 - 6:30 pm, Field House
This poster is part of the Sandbox Series.
Srijita Das1, Haley Mendez2, Stephen Gliske3; 1Graduate Student, University of Nebraska Medical Center, 2Research Intern, University of Nebraska Medical Center, 3Associate Professor, University of Nebraska Medical Center
Patient-specific mappings of the eloquent cortex are crucial for minimizing deficits following resective surgery, particularly in epilepsy patients, among whom one-third are resistant to anti-epileptic drugs. For these individuals, surgery offers the next best option, but successful outcomes depend on accurately localizing eloquent cortical regions, such as language, to minimize post-surgical deficits. Language is innately complex and critical to localize due to the atypical cortical language representation in epilepsy patients. Electrocortical stimulation mapping (ESM), the current gold standard for language localization, is invasive and depends on accurate pre-implantation mapping. Magnetoencephalography (MEG) offers a non-invasive alternative by enabling clinicians to identify surgical candidates and guide electrode implantation for ESM. Its utility for language mapping using simple dipole method remains limited to single point in space for a given point in time. In this work-in-progress study, we evaluate our previously validated localization method based on phase-amplitude coupling (PAC) for receptive language cortex localization in epilepsy patients undergoing MEG. Our approach employs a passive auditory language (PAL) task, which consists of audio clips from natural speech played either forward or backward, each lasting 500-800 ms, presented every 2.5 seconds. From a retrospective database of 178 epilepsy patients who completed MEG language tasks, 25 patients underwent the PAL paradigm. Preprocessing was conducted in MNE-Python, including notch filtering at 60 Hz and independent component analysis for rejection of cardiac and ocular artifacts. We will proceed with further processing like our published method. Specifically, in Brainstorm we will create epochs of 1000 ms with 100 ms as pre-stimulus baseline. We will average the epochs across trials to generate evoked files and perform source-reconstruction using LCMV beamforming. PAC will be quantified using a time-resolved sliding window approach (500 ms) with 1-12 Hz as the low-frequency (LF) band and 30-300 Hz as the high-frequency (HF) range. PAC will be calculated on the time-courses of neural activity for 148 brain regions defined by the Destrieux atlas. Next, we will apply tensor decomposition to reduce the dimensionality of our six-dimensional data (brain scouts, LF bin, HF bin, patient, time window, speech or non-speech event). Density-based spatial clustering (DBSCAN) clustering with silhouette evaluation will be employed to identify task-activated brain regions as outliers, most of the brain is not involved in the language task. A linear mixed-effects (LME) model will be constructed with PAC as a response variable, expected regions as fixed effects and patients as a random variable. A binary support-vector machine (SVM) classifier will be developed for the language task to predict brain regions depending on PAC values. We will validate PAC-derived maps against ESM results to assess their clinical accuracy. We expect to observe elevated phase-amplitude coupling in the posterior superior temporal gyrus of the dominant hemisphere, consistent with known receptive language function. This abstract presents our analytic framework and early observations, with particular emphasis on developing individual-level PAC profiles to assess functional language lateralization and localization. This study advances the application of PAC as a biomarker for functional language mapping, with implications for broader neuroimaging research.
Topic Areas: Computational Approaches, Disorders: Acquired