Retina and awake-mouse recordings

The 2020 microwire-CMOS bundle paper tests different tissue preparations. Stanford affiliation grounds the institution link, not a claim that its human BCI laboratory conducted this work.

Isolated rat retina

A 138-wire bundle contacted isolated retina held on a dialysis membrane in a perfusion chamber. Spike sorting yielded 152 units, approximately 1.1 units per wire. Light-onset and light-offset responses supported a retinal ganglion-cell interpretation. This is an ex vivo experiment, not a retinal prosthesis trial or a living-animal chronic implant.

Acute head-constrained mice

Mouse bundles contained 135-251 PtIr wires and reached deep motor/somatosensory cortex or dorsal striatum. Recordings began within two hours of implantation. Mice could run voluntarily on an air-supported spherical treadmill while their heads were constrained; this is not unrestricted free movement in a home cage.

Typical five-minute sessions yielded 92-221 putative neurons, with 0.56 ± 0.11 units per wire (reported range 0.3-0.89). Those session results are not the 8,640-wire connectivity demonstration or the full CMOS pixel count.

Motor-cortex and striatal firing rates increased during running in the reported comparisons. These are activity/behavior correlations, not an assistive-device decoder or causal proof that the measured neurons produced the movement.

Evidence boundary

The 14-day chip-interface test was performed on a mechanically pressed bundle, with daily connectivity/noise measurements. It does not establish chronic implanted single-unit tracking, long-term tissue safety or human clinical readiness. The paper’s larger-animal scaling proposals remain distinct from its retina and acute mouse outcomes.

Primary sources

• 2020 paper, Figures 5 and 6 and methods.
• Full primary text.
• Stanford publisher PDF.