Simulating Scalp EEG from Ultrahigh-Density ECoG Data Illustrates Cortex to Scalp Projection Patterns

Abstract

Ultrahigh-density electrocorticography (μECoG) provides unprecedented spatial resolution for recording cortical electrical activity. This study uses simulated scalp projections from a μECoG recording to challenge the assumption that channel-level electroencephalography (EEG) reflects only local field potentials near the recording electrode. Using a 1024-electrode μECoG array placed on the primary motor cortex during finger movements, we applied Adaptive Mixture Independent Component Analysis (AMICA) to decompose activity into maximally independent grid activity components and projected these to 207 simulated EEG scalp electrode channels using a high-definition MR image-based electrical forward-problem head model. Our findings demonstrate how cortical surface-recorded potentials propagate to scalp electrodes both far from and near to the generating location. This work has significant implications for interpreting both EEG and ECoG data in clinical and research applications.

Why This Matters

The widespread use of channel-level EEG analysis in both research and clinical settings is fundamentally flawed. Despite decades of research demonstrating volume conduction effects, many studies still attribute EEG channel activity to cortical areas “under” the electrode - an assumption that ignores basic biophysics. This study provides concrete evidence of how misleading this approach can be.

By using μECoG data as ground truth cortical activity, we can definitively show where brain signals are actually generated and where they appear on the scalp. The results are striking: activity from a small motor cortex patch appears most strongly not overhead, but in distant parietal and frontal regions. This isn’t just a technical curiosity - it has profound implications for how we interpret EEG in everything from cognitive neuroscience to clinical diagnosis.

Key Findings

Widespread Cortical-to-Scalp Projection: Activity from a 3×3 cm cortical patch projected to all 207 simulated scalp electrodes, with strongest signals appearing in left parietal and right frontal areas rather than directly overhead.

Multiple Independent Sources: AMICA decomposition revealed 582 independent cortical sources within the small μECoG patch, challenging the notion of simple, stationary local field potentials.

Non-monotonic Distance Relationships: EEG signal metrics did not decrease monotonically with distance from the source, demonstrating complex volume conduction patterns that defy simple proximity-based interpretations.

Dynamic Activity Patterns: The independent components showed non-stationary, dynamic activity patterns resembling traveling waves rather than static dipolar sources, suggesting conventional ICA assumptions may be inadequate for cortical data.

Projection patterns across the scalp

Top row shows topographic distributions of signal range, standard deviation, and entropy from the μECoG patch projected to scalp electrodes. Bottom row shows spectral power distributions across frequency bands. Note how the strongest activity appears in parietal and frontal regions, not directly over the motor cortex source.

Signal metrics versus electrode distance

The relationship between EEG signal metrics and electrode distance from the μECoG patch center reveals non-linear patterns. Only signal range shows a clear decreasing trend with distance, while standard deviation and entropy show complex, non-monotonic relationships that challenge simple distance-based interpretations.

Implications for EEG Research

These findings have immediate practical consequences. Channel-level EEG analysis - still the norm in many labs and clinics - systematically misattributes the cortical origins of recorded activity. This affects everything from cognitive ERP interpretations to clinical seizure localization.

The solution isn’t to abandon EEG, but to embrace source-level analysis methods. Techniques like ICA, when properly applied, can disentangle the mixed signals that volume conduction creates. The widespread projection patterns we observed actually represent an advantage: each EEG electrode contains information from multiple cortical areas, potentially allowing detection of activity from regions traditionally considered “too distant” to influence scalp recordings.

For the clinical community, these results suggest that EEG-based diagnoses relying on channel-level features may be fundamentally limited. For researchers, they underscore why source-level analysis should be the standard, not the exception.

Citation

@ARTICLE{Shirazi2025-uecog,
  title    = "Simulating Scalp {EEG} from Ultrahigh-Density {ECoG} Data 
              Illustrates Cortex to Scalp Projection Patterns",
  author   = "Shirazi, Seyed Yahya and Onton, Julie and Makeig, Scott",
  journal  = "bioRxiv",
  year     =  2025,
  doi      = "10.1101/2025.06.24.660870",
  language = "en"
}