When recording electrical activity from the surface of the human body—whether tracking neural oscillations via scalp electroencephalography (EEG) or cardiac propagation via electrocardiography (ECG)—sensors capture a highly degraded version of the underlying physiological reality. As David Steyrl et al. note in On Similarities and Differences of Invasive and Non-Invasive Electrical Brain Signals in Brain-Computer Interfacing, invasive and non-invasive electrical brain signals share the exact same biological origin: action potential spiking rates in the cortical layers. The performance gap between invasive and non-invasive modalities does not stem from different neural sources, but rather from the physical medium the signals must traverse.
The intervening layers of tissue—including the brain, cerebrospinal fluid, skull, scalp, or the human torso—act as a volume conductor. This anatomical barrier applies severe spatial smoothing and frequency-dependent attenuation to electrical signals, acting as a biological low-pass filter.
The Physics of Tissue-Induced Attenuation
To understand why this filtering occurs, we must look at the physics of signal propagation through heterogeneous biological tissue. The electric field generated by cellular activity decays relative to the square of the distance ($1/r^2$) from the source, as explained by David Steyrl et al.
Furthermore, high-frequency neural oscillations (such as the gamma band, >60 Hz) typically reflect highly localized, asynchronous, or fine-grained laminar cortical synchronization (Gabriel N. Costa et al., Interhemispheric Binding of Ambiguous Visual Motion Is Associated with Changes in Beta Oscillatory Activity but Not with Gamma Range Synchrony). Because these high-frequency signals are spatially restricted, their electrical fields tend to cancel each other out over a distance. Conversely, lower-frequency oscillations (like alpha or theta rhythms) involve large-scale, synchronous populations of neurons. This widespread synchrony allows low-frequency signals to summate constructively and survive the journey through the skull.
This phenomenon is not unique to the brain. In cardiac imaging, Corentin Dallet et al. demonstrate in Combined Signal Averaging and Electrocardiographic Imaging Method to Non-Invasively Identify Atrial and Ventricular Tachycardia Mechanisms that the human torso similarly acts as a low-pass filter. It smooths and obscures high-frequency electrical information from the heart, presenting a fundamental barrier to high-resolution non-invasive mapping that cannot be fully resolved through basic signal processing alone.
The High-Frequency Blind Spot
This biological low-pass filter creates a significant "blind spot" for non-invasive sensors, stripping away the fine-grained details of internal neural and cardiac representations.
Loss of Cognitive and Pathological Biomarkers
High-frequency signals are critical for understanding cognitive processes and pathologies, yet they are routinely obscured by skull attenuation. For instance, Christine Grützner et al. highlight in Deficits in high- (>60 Hz) gamma-band oscillations during visual processing in schizophrenia that high-gamma reductions are powerful biomarkers for schizophrenia, but standard scalp EEG struggles to capture them.
Similarly, deep-brain high-frequency events like hippocampal sharp-wave ripples (Takamitsu Iwata, Hippocampal sharp-wave ripples correlate with periods of naturally occurring self-generated thoughts in humans), preictal phase-amplitude coupling (Hui Ming Khoo, Phase-amplitude coupling between infraslow and high-frequency activities well discriminates between the preictal and interictal states), and early biomarkers of epileptogenesis (Diyuan Lu et al., Staging Epileptogenesis with Deep Neural Networks) remain largely locked behind the physical barrier of the skull, preventing straightforward non-invasive clinical monitoring.
The "Black-Box" Effect in BCIs
In Brain-Computer Interfaces (BCIs), the loss of high-frequency nuances limits communication speed and control resolution. Non-invasive scalp recordings are restricted to a simplified, mid-level representation of cortical activity (Thomas Prätzlich et al., Brain Beats: Tempo Extraction from EEG Data). This degradation forces non-invasive BCIs to rely on multiple stimulus repetitions to extract clear signals, severely limiting bit rates (Peter Brunner et al., Rapid Communication with a “P300” Matrix Speller Using Electrocorticographic Signals (ECoG); Thomas Navin Lal et al., Methods Towards Invasive Human Brain Computer Interfaces).
[Cortical Source] ---> (High-Frequency & Localized Signals)
|
v
[SKULL & TISSUE FILTER] <--- Attenuates fast oscillations,
| blurs spatial patterns
v
[Scalp Sensors] ---> (Low-Frequency & Smoothed Signals Only)
Strategies for Overcoming the Filtering Barrier
To bridge the gap between invasive high-resolution data and non-invasive accessibility, researchers employ several distinct paradigms:
- Invasive Bypassing (ECoG): By placing electrodes directly on the brain surface (Electrocorticography), researchers completely circumvent the skull's low-pass filter, enabling rapid communication speeds (Peter Brunner et al.) and localized feature selection (Thomas Navin Lal et al.).
- Super-Nyquist Spatial Sampling: Traditional wisdom suggests wide electrode spacing (20–30mm) is sufficient because the skull blurs the signal anyway. However, Amanda K. Robinson et al. demonstrate in Very high density EEG elucidates spatiotemporal aspects of early visual processing that ultra-high-density arrays (14mm spacing) can capture low-amplitude, high-spatial-frequency patterns that are normally missed, proving that spatial pattern information can be recovered even when total signal energy is attenuated.
- Anatomical Beamforming: By pairing scalp EEG with individual structural MRI scans, Melissa M. Smith et al. in Non-invasive detection of high gamma band activity during motor imagery successfully mapped high-gamma activity non-invasively, mathematically reversing some of the spatial blurring caused by the volume conductor.
- Cross-Modal and Algorithmic Alignment: Other approaches use mathematical projections or multi-subject averaging to cancel out noise and isolate weak high-frequency signals. For example, Mingxiong Huang et al. developed an "EMG-projected MEG" technique to capture elusive gamma-band brain-muscle coupling (EMG-projected MEG High-Resolution Source Imaging of Human Motor Execution: Brain-Muscle Coupling above Movement Frequencies), while Di Zhou et al. leveraged subject-alignment algorithms to uncover gamma-band tracking of speech envelopes from noisy scalp data (Neural Entrainment to Natural Speech Envelope Based on Subject Aligned EEG Signals).
Key Foundational Work
Rapid Communication with a “P300” Matrix Speller Using Electrocorticographic Signals (ECoG) (2011) — contributed 1 concept elements to this direction.
Methods Towards Invasive Human Brain Computer Interfaces (2004) — contributed 1 concept elements to this direction.