Devices

A CMOS silicon intracortical probe family (Neuropixels v1/v2): extremely high-density, low-noise spike recording for animal research; not a chronic human implant.

Device — Intracortical

Neuropixels Probe

Official site →

BCI · intracortical · Neuropixels · recording · CMOS · IMEC · UCL · Allen Institute

Neuropixels Probe (IMEC / UCL / Allen Institute)

One-line verdict: Neuropixels is the highest-density, lowest-noise intracortical recording platform widely used in neuroscience, defining an upper bound on spike-recording throughput — but it is a research tool, not a chronic human implant.

Quick tags: Recording · Stimulation: no · Channels: 384–960 (generation-dependent) · Species: rodent + NHP · First demonstrated: 2017

Overview

What it is: A silicon CMOS-based neural probe with hundreds of recording sites densely packed along a thin shank, multiplexed into on-chip amplifiers/ADCs and streamed out through a wired headstage.

Why it matters: It sets a practical ceiling for channel density and signal quality in vivo, and it has become a benchmark substrate for spike-sorting and large-scale systems neuroscience.

Most comparable devices: Utah microelectrode arrays (human intracortical baseline), flexible-thread intracortical systems (e.g., Neuralink-style architecture, different goals), other CMOS shank probes.

Spec Card Grid

Identity

  • Device name: Neuropixels (v1, v2.x family)
  • Canonical ID: BTSD-0004
  • Inventor / key authors: IMEC / UCL / Allen Institute teams (collaborative platform)
  • Org / manufacturer: IMEC (platform)
  • First demonstrated (year): 2017
  • First implanted (year): N/A (research tool; not a chronic human implant)
  • Species: rodent, NHP
  • Regulatory / trial status: research only
  • Primary use: recording
  • Primary target: cortex, hippocampus, thalamus (and other deep targets depending on placement)

Geometry & Architecture

  • Interface type: intracortical
  • Penetrating?: yes
  • Form factor: single thin silicon shank
  • Shank width (µm): ~70 (generation-dependent)
  • Shank thickness (µm): ~20 (generation-dependent)
  • Shank length (mm): ~10 (common)
  • Site spacing (µm): ~20 (common)
  • Array layout: linear column(s) along the shank
  • Insertion method: micromanipulator
  • Anchoring method: rigid shank + skull fixation (typical acute/semichronic prep)
  • Packaging location: external headstage

Electrode & Channel Physics

  • Total sites: ~960–1280 (generation-dependent)
  • Simultaneous channels: ~384–768 (generation-dependent)
  • Electrode material: titanium nitride (common)
  • Site area (µm²): small (often reported on the order of ~10 µm²; varies by generation)
  • Impedance @ 1 kHz: often reported around ~100–200 kΩ (varies)
  • Noise floor / SNR: low-noise, often reported around a few µV RMS (varies by setup)
  • Recording modality: single-unit spikes + LFP
  • Stimulation capability: no
  • Charge injection limit / safe stim range: N/A

Tissue Interface & Bioresponse

  • Target tissue: gray + white matter (depending on trajectory)
  • BBB disruption: high (penetrating)
  • Vascular disruption risk: moderate–high (placement dependent)
  • Micromotion sensitivity: high (rigid silicon in moving brain)
  • Gliosis / encapsulation: major constraint for chronic use
  • Neuron loss (if reported): can be significant near shank in chronic contexts
  • Foreign-body response mitigation: not the core design goal; platform optimized for research recording
  • Typical failure mode: tissue response + micromotion → drift/instability; not designed for long-term human implantation

System Architecture

  • Onboard electronics: on-shank amplification + multiplexing/ADC (platform feature)
  • Data path: high-speed wired headstage
  • Telemetry bandwidth: wired (lab system; varies)
  • Sampling rate: often ~30 kHz per channel (common in practice)
  • Power: external
  • Thermal management: low but nonzero dissipation; setup-dependent
  • Hermeticity: none (lab device)
  • MRI compatibility: no/unknown (assume no)
  • Surgical complexity: high-precision placement (research surgery)

Performance Envelope

  • Spike yield: extremely high (relative to most other single-probe technologies)
  • Neuron count per probe: often thousands (analysis-dependent)
  • Stability over time: hours to days/weeks (prep-dependent); not a chronic implant platform
  • Longevity (median / max): not intended for chronic implant lifetimes
  • Revision / explant: remove and replace
  • Notable demos / tasks: large-scale circuit mapping; decoding benchmarks; whole-brain-scale recordings in animals

Clinical / Preclinical Evidence

  • N implanted subjects / animals: very large adoption across labs (animal research)
  • Follow-up duration: acute / short-term / semichronic
  • Indications: none (research)
  • Trial registry links: N/A
  • Primary outcomes: research recordings
  • Key limitations of evidence: not a clinical device; chronic human translation would require different packaging/biocompatibility strategy

Engineering Verdict

Strengths:

  • unmatched channel density (in a single shank)
  • excellent noise performance
  • on-chip multiplexing enables scale
  • strong benchmarking ecosystem (software + datasets)

Limitations / failure modes:

  • rigid silicon + penetrating geometry
  • external tether / headstage
  • chronic tissue response and micromotion sensitivity

Scaling constraints:

  • mechanical mismatch with brain
  • wiring/bandwidth and connector complexity
  • thermal + packaging constraints if translated

What next-gen tries to fix:

  • softer mechanics (compliance matching)
  • implantable packaging and hermeticity
  • long-term stability

Simulation Hooks (for BuildTheSimulation)

  • Minimal model to reproduce: dense electrode column in cortical tissue with spike propagation + distance-dependent attenuation
  • Parameters to expose as sliders: site spacing, shank thickness, noise floor, tissue damage zone / encapsulation thickness
  • What outputs to visualize: spike yield proxy, neuron count proxy, SNR proxy, stability vs time proxy

References