Ideal current dipoles are appropriate source representations for simulating neurons for intracranial recordings

Combining independent component analysis and source localization for improving spatial sampling of stereoelectroencephalography in epilepsy

Stereoelectroencephalography is a powerful intracerebral EEG recording method for the presurgical evaluation of epilepsy. It consists in implanting depth electrodes in the patient's brain to record electrical activity and map the epileptogenic zone, which should be resected to render the patient seizure-free. Stereoelectroencephalography has high spatial accuracy and signal-to-noise ratio but remains limited in the coverage of the explored brain regions. Thus, the implantation might provide a suboptimal sampling of epileptogenic regions. We investigate the potential of improving a suboptimal stereoelectroencephalography recording by performing source localization on stereoelectroencephalography signals. We propose combining independent component analysis, connectivity measures to identify components of interest, and distributed source modelling. This approach was tested on two patients with two implantations each, the first failing to characterize the epileptogenic zone and the second giving a better diagnosis. We demonstrate that ictal and interictal source localization performed on the first stereoelectroencephalography recordings matches the findings of the second stereo-EEG exploration. Our findings suggest that independent component analysis followed by source localization on the topographies of interest is a promising method for retrieving the epileptogenic zone in case of suboptimal implantation.

Optimization of patient-specific stereo-EEG recording sensitivity

Stereo-EEG is a minimally invasive technique used to localize the origin of epileptic activity (the epileptogenic zone) in patients with drug-resistant epilepsy. However, current stereo-EEG trajectory planning methods are agnostic to the spatial recording sensitivity of implanted electrodes. In this study, we used image-based patient-specific computational models to design optimized stereo-EEG electrode configurations. Patient-specific optimized electrode configurations exhibited substantially higher recording sensitivity than clinically implanted configurations, and this may lead to a more accurate delineation of the epileptogenic zone. The optimized configurations also achieved equally good or better recording sensitivity with fewer electrodes compared with clinically implanted configurations, and this may reduce the risk for complications, including intracranial haemorrhage. This approach improves localization of the epileptogenic zone by transforming the clinical use of stereo-EEG from a discrete ad hoc sampling to an intelligent mapping of the regions of interest.

Multi-scale model of axonal and dendritic polarization by transcranial direct current stimulation in realistic head geometry

BACKGROUND

Transcranial direct current stimulation (tDCS) is a non-invasive brain stimulation modality that can alter cortical excitability. However, it remains unclear how the subcellular elements of different neuron types are polarized by specific electric field (E-field) distributions.

OBJECTIVE

To quantify neuronal polarization generated by tDCS in a multi-scale computational model.

METHODS

We embedded layer-specific, morphologically-realistic cortical neuron models in a finite element model of the E-field in a human head and simulated steady-state polarization generated by conventional primary-motor-cortex-supraorbital (M1-SO) and 4 × 1 high-definition (HD) tDCS. We quantified somatic, axonal, and dendritic polarization of excitatory pyramidal cells in layers 2/3, 5, and 6, as well as inhibitory interneurons in layers 1 and 4 of the hand knob.

RESULTS

Axonal and dendritic terminals were polarized more than the soma in all neurons, with peak axonal and dendritic polarization of 0.92 mV and 0.21 mV, respectively, compared to peak somatic polarization of 0.07 mV for 1.8 mA M1-SO stimulation. Both montages generated regions of depolarization and hyperpolarization beneath the M1 anode; M1-SO produced slightly stronger, more diffuse polarization peaking in the central sulcus, while 4 × 1 HD produced higher peak polarization in the gyral crown. The E-field component normal to the cortical surface correlated strongly with pyramidal neuron somatic polarization (R2>0.9), but exhibited weaker correlations with peak pyramidal axonal and dendritic polarization (R2:0.5-0.9) and peak polarization in all subcellular regions of interneurons (R2:0.3-0.6). Simulating polarization by uniform local E-field extracted at the soma approximated the spatial distribution of tDCS polarization but produced large errors in some regions (median absolute percent error: 7.9 %).

CONCLUSIONS

Polarization of pre- and postsynaptic compartments of excitatory and inhibitory cortical neurons may play a significant role in tDCS neuromodulation. These effects cannot be predicted from the E-field distribution alone but rather require calculation of the neuronal response.

Comparison of EEG Source Localization Using Simplified and Anatomically Accurate Head Models in Younger and Older Adults

Accuracy of electroencephalography (EEG) source localization relies on the volume conduction head model. A previous analysis of young adults has shown that simplified head models have larger source localization errors when compared with head models based on magnetic resonance images (MRIs). As obtaining individual MRIs may not always be feasible, researchers often use generic head models based on template MRIs. It is unclear how much error would be introduced using template MRI head models in older adults that likely have differences in brain structure compared to young adults. The primary goal of this study was to determine the error caused by using simplified head models without individual-specific MRIs in both younger and older adults. We collected high-density EEG during uneven terrain walking and motor imagery for 15 younger (22±3 years) and 21 older adults (74±5 years) and obtained [Formula: see text]-weighted MRI for each individual. We performed equivalent dipole fitting after independent component analysis to obtain brain source locations using four forward modeling pipelines with increasing complexity. These pipelines included: 1) a generic head model with template electrode positions or 2) digitized electrode positions, 3) individual-specific head models with digitized electrode positions using simplified tissue segmentation, or 4) anatomically accurate segmentation. We found that when compared to the anatomically accurate individual-specific head models, performing dipole fitting with generic head models led to similar source localization discrepancies (up to 2 cm) for younger and older adults. Co-registering digitized electrode locations to the generic head models reduced source localization discrepancies by  ∼  6 mm. Additionally, we found that source depths generally increased with skull conductivity for the representative young adult but not as much for the older adult. Our results can help inform a more accurate interpretation of brain areas in EEG studies when individual MRIs are unavailable.

Relative Contributions of Different Neural Sources to the EEG

Dogma dictates that the EEG signal is generated by postsynaptic currents (PSCs) because there are an enormous number of synapses in the brain, and PSCs have relatively long durations. However, PSCs are not the only potential source of electric fields in the brain. Action potentials, afterpolarizations, and presynaptic activity can also generate electric fields. Experimentally it is exceedingly difficult to delineate the contributions of different sources because they are casually linked. However, using computational modeling, we can interrogate the relative contributions of different neural elements to the EEG. We used a library of neuron models with morphologically realistic axonal arbors to quantify the relative contributions of PSCs, action potentials, and presynaptic activity to the EEG signal. Consistent with prior assertions, PSCs were the largest contributor to the EEG, but action potentials and afterpolarizations can also make appreciable contributions. For a population of neurons generating simultaneous PSCs and action potentials, we found that the action potentials accounted for up to 20% of the source strength while PSCs accounted for the other 80% and presynaptic activity negligibly contributed. Additionally, L5 PCs generated the largest PSC and action potential signals indicating that they the dominant EEG signal generator. Further, action potentials and afterpolarizations were sufficient to generate physiological oscillations, indicating that they are valid source contributors to the EEG. The EEG emerges from a combination of multiple different source, and, while PSCs are the largest contributor, other sources are non-negligible and should be included in modeling, analysis and interpretation of the EEG.

iCanClean Improves Independent Component Analysis of Mobile Brain Imaging with EEG

Motion artifacts hinder source-level analysis of mobile electroencephalography (EEG) data using independent component analysis (ICA). iCanClean is a novel cleaning algorithm that uses reference noise recordings to remove noisy EEG subspaces, but it has not been formally tested in a parameter sweep. The goal of this study was to test iCanClean’s ability to improve the ICA decomposition of EEG data corrupted by walking motion artifacts. Our primary objective was to determine optimal settings and performance in a parameter sweep (varying the window length and r2 cleaning aggressiveness). High-density EEG was recorded with 120 + 120 (dual-layer) EEG electrodes in young adults, high-functioning older adults, and low-functioning older adults. EEG data were decomposed by ICA after basic preprocessing and iCanClean. Components well-localized as dipoles (residual variance < 15%) and with high brain probability (ICLabel > 50%) were marked as ‘good’. We determined iCanClean’s optimal window length and cleaning aggressiveness to be 4-s and r2 = 0.65 for our data. At these settings, iCanClean improved the average number of good components from 8.4 to 13.2 (+57%). Good performance could be maintained with reduced sets of noise channels (12.7, 12.2, and 12.0 good components for 64, 32, and 16 noise channels, respectively). Overall, iCanClean shows promise as an effective method to clean mobile EEG data.