S129. An open source platform for concurrent EEG/ECoG comparisons

Abstract

Concurrent EEG/ECoG recordings are crucial in providing a gold standard to develop EEG models, especially using primate models. These could be crucial for selecting methods for determining valid brain connectivity measures. Two essential points are to eliminate the unavoidable artifacts that contaminate such types of recordings as well as to provide sufficiently detailed biophysical models, such as lead fields, for both EEG and ECoG data that take into consideration intracranial electrode materials. We present a Matlab open source platform for EEG/ECoG comparisons, based upon the Multidimensional Recording in Nagasaka et al. (2011) data set shared at http://neurotycho.org. EEG/ECoG registration, artifact rejection, and preprocessing routines are illustrated with data obtained from awake and anesthetized monkeys. This platform has already been used to produce the results described in Papadopuoulus et al. (2014) and Todaro et al. (2017). The platform uses BrainSuite2 for T1 MRI pre-processing (intensity bias correction and segmentation). The white/gray matter interface surface was chosen as the source space model for both EEG/ECoG. The specific head compartments considered were (1) brain (enclosed by the pial surface), (2) brain plus surrounding cerebrospinal fluid, (3) skull and (4) skin. Final manually correction was carried out by an expert. Our volume conduction model comprises compartments (2), (3) and (4). Lead Field Calculation: Tetrahedral meshes were created from the surfaces of the head model using Tetgen 2.0. Both EEG and ECoG lead fields were calculated by NeuroFEM using Finite Element Method from SimBio. Care was taken to model the plastic strips supporting ECoG electrodes. Model-Based Artifact Removal: The raw EEG data suffered from mainly 2 types of artifacts: Low frequency drifting, and Sparse spikes. We introduce the Transient Artifact Reduction Algorithm. We formulate the artifact removal process as minimizing the residual, and penalizing sparsity of both the trend and spikes. More details can be found in Selesnick et al. (2014). Time Frequency Analysis: Spectrogram (squared magnitude of the Short-Time Fourier Transform (STFT)) is useful in analyzing neurological signals, but with drawback of leakage effect. To overcome this, we use the multi-tapper method and the implementation from Chronux (Bokil et al., 2006) and we also calculate the cross-spectrum. Spectrogram: EEG and ECoG show slower oscillation in anesthesia state than the awake state (<5 Hz), but ECoG can detect obvious beta rhythm (13–30 Hz) in the awake state while EEG cannot; Cross-Spectrum: Averagely(0.3–45 Hz), EEG is stronger in the frontal area while ECoG is stronger in the occipital and temporal areas in both awake and anesthesia states. The proposed open source platform and allows co-registration, artifact removal and biophysical modeling of use in both sensor and source level analyses. We believe that this platform will be helpful for EEG/ECoG multimodal comparisons and integration.