Neural Noise- Crosstalk in neural recordings can bias decoding and connectivity estimates; bridging circuit-level modeling and signal analysis quantifies contamination risk.1
- The approach is critical for high-density and invasive arrays and has immediate impact on BCI and iEEG pipeline design.1
- Nature-published methods support validity assessment of neural recordings.1 1
Weekly enrichment (2026-07-20)
- The underlying source is a Nature Communications paper (2025) showing that epicortical micro-ECoG recordings from anesthetized rat brains acquired with a state-of-the-art system are compromised by crosstalk, with signal-coherence maps depending on the hardware routing layout rather than on cortical inter-electrode distance.2
- For high-frequency activity (above 300 Hz), coherence was highest between channels wired adjacently even when their electrodes sat far apart on the cortex — a signature that distinguishes crosstalk from genuine volume conduction.2
- The authors fully characterized the electrode-to-amplifier chain via impedance spectroscopy, built a lumped-element transfer-function model, and developed a novel crosstalk back-correction algorithm that reconstructs how signals would look under a zero-crosstalk scenario.2
- After back-correction, coherence dropped in the multi-unit activity (spike) band while remaining unchanged in the LFP band; coupled signals modulated raw waveform amplitudes by as much as 4.3%, enough to sway spike waveforms past the detection threshold and cause spurious spike detection.2
- Spike cross-correlation was proposed as an indicator that separates crosstalk from volume conduction, and validating recorded data against the routing layout is recommended as a routine data-quality-control step.2
- A circuit-model study found cross-talk amplitude rises nearly linearly with recording-electrode impedance and grows at higher frequencies, spatially smoothing current-source-density estimates and inducing artifactual phase shifts, though effects are modest when coupled channels record similar-magnitude signals.3
- An IEEE NER 2019 study using a thin-film ECoG array (MuSA) in rats reported high in vivo coherence at 1 kHz between electrodes with neighboring routing tracks, corroborated by saline (PBS) controls, flagging crosstalk as a barrier to extending recordings into the spike band (≥1 kHz).4
- Spike-to-LFP contamination is realistic down to roughly 10 Hz in awake-monkey motor cortex, so simply high-pass filtering above 300 Hz does not guarantee spike-free LFPs; per-dataset assessment is advised before LFP–behavior or LFP–spiking correlations.5
- On Neuropixels probes (inter-electrode spacing as small as ~20 µm), crosstalk between distant channel pairs generates hyper-synchronous spike events that survive Kilosort sorting; a ZCA-whitening-based cleaning step removes nearly all above-chance synchronous events.6