Applications

The original Neuropixels study records more than 700 isolated neurons with two probes and reports chronic event-rate stability alongside electronic damage, local activity loss and late implant detachment.

Application — Intracortical

Neuropixels: 2017 rodent recordings and reported failures

Neuropixels · rodent · chronic · failures · historical · preclinical

Original Neuropixels rodent recordings

Jun and colleagues’ 2017 paper describes a multi-phase probe development and recording program, with affiliations including Janelia, UCL, imec and the Allen Institute. The linked 1.0 hardware entry keeps the original 20 µm shank thickness distinct from the manufacturer’s later 24 µm specification.

Large-population recording

Two probes recorded more than 700 well-isolated neurons simultaneously across five structures in an awake mouse. This is a two-probe result, not a yield from one 384-channel probe. The paper also demonstrates recordings in freely moving rodents.

Chronic metrics and limits

Extended Data Figure 5 reports chronic rat medial-prefrontal-cortex recordings across development phases. Across 14 probes in phases 2 and 3, spiking activity generally did not decline over the eight-week observation, with an explicit exception: the upper half of one distal recording array lost nearly all activity during the first 30 days while the lower approximately 1.9 mm remained stable.

The activity metric counts time-coincident spikes on contiguous sites exceeding a threshold. Stability of this metric is not proof that every neuron remains identifiable across every session, or that all hardware survives indefinitely.

Failures the paper reports

  • Amplified, switchable probes were electronically damaged before enough recording data could be collected in the phase-3 chronic comparison.
  • One probe’s upper recording region lost nearly all activity over 30 days, despite stability in its lower region.
  • Three of 16 implants detached from the skull after 223-482 days.
  • Surgical-wound irritation required euthanasia in some animals, always after more than 20 weeks; the caption does not supply a separate count for this outcome.

These counts refer to different cohorts and events. They are not pooled into a single failure rate. The study does not report a universal chronic lifetime, clinical qualification or a completed human BCI trial.

Primary sources