Comparing Measures of Active and Passive Field Spread between non-, partially, and fully invasive human brain recordings

Abstract

Event Abstract Back to Event Comparing Measures of Active and Passive Field Spread between non-, partially, and fully invasive human brain recordings Mansoureh Fahimi Hnazaee1*, Benjamin Wittevrongel1, Evelien Carrette2, Ine Dauwe2, Alfred Meurs2, Paul Boon2, Dirk Van Roost2 and Marc Van Hulle1 1 Faculty of Medicine, KU Leuven, Belgium 2 Ghent University Hospital, Belgium BACKGROUND Precise details of the propagation of neural activity on remote cortical areas are still poorly understood; this propagation refers to both electric diffusion as a result of physiological propagation of action potentials along axons (active spread) and the passive spread of the electric field over the cortical surface, also referred to as field spread, or as more commonly known, electrical volume conduction. Principles of volume conduction is of significant interest for researchers interested in solving the M/EEG inverse problem and increasing the accuracy of source localization. Therefore we need measures that can separate instantaneous signal propagation from genuine causal and functional interactions, such as methods that can reveal the temporal delay. However, no study has yet attempted to compare the influence of volume conduction for different levels of invasiveness. In this study, by using measures of phase-based and amplitude based correlations both on actual data and reconstructed sources obtained from the data for each recording modality, we try to quantify the level of volume conduction, comparing between subcortical, cortical, and scalp recordings. METHODS We recorded 3 minutes of resting state (with eyes fixating) and a rich stimuli task in 5 patients (2 female and 3 male patients) from simultaneously implanted subdural grids, depth electrodes, and scalp EEG. Areas of subdural recording were spread over left and right temporal and frontal cortex, and depth electrodes were implanted in left and/or right hippocampus in 4 patients and left insula in 1 patient. The data is decomposed into traditional frequency bands using Morlet wavelet function. Phase-based connectivity measures were used to quantify the interactions include Phase Locking Value (PLV), and imaginary part of Phase Locking Value (iPLV) whereas for Amplitude-based we used Spearman correlation coefficient. The connectivity analysis is done on both the data and on the estimated sources obtained using Independent Components Analysis (ICA). The independent sources were estimated with statistical and algorithmic reliability with the help of bootstrapping and random initialization using the Icasso software package. The calculated measures from both the original data and on the estimated independent sources are then analyzed statistically using ANOVA. CONCLUSION Our study provides a comprehensive account of the spatial spread between SEEG, ECoG and EEG. As can be expected, IPLV consistently decreases for higher frequency bands. Additionally, the lower PLV and Spearman correlation for ECoG compared to SEEG can probably be attributed to the larger average distances (larger spatial coverage) spanned by ECoG. Using ICA analysis on partially invasive and fully invasive brain recording we consistently found between 4 and 7 cortical generators, which is in accordance with previous studies on functional networks during resting state. Similar overall trends in connectivity can be seen for both raw data and reconstructed sources, however PLV and Spearman are now higher for SEEG compared to ECoG, which is likely attributed to a higher number of cortical generators discovered from ECoG. FUTURE WORK - Current results show the analysis of a single subject, further analysis would include normalizing and pooling results from all subjects in order to investigate connectivity as a function of anatomical location. When pooling results we can also account for inter-electrode distances and control these for modalities. - It has been shown that correlations are more variable and on average higher during rest than when the patient is engaged in a listening task. In this study data has been recorded from the same patients engaged during a rich stimuli task (an audio-visual movie to ensure engagement of multi-sensory and higher-level cognitive functions). A future step would be to compare task and non-task connectivity measures.