Chronic microstimulation with StimNET

The published 2023 Cell Reports study uses StimNET hardware in awake mice. The earlier preprint is not substituted for this published report.

Separate imaging and behavioral groups

The study used nine mice for two-photon imaging and five for behavioral experiments. Three imaging mice were excluded because of early cranial-window occlusion or backend connector breakage. Imaging and behavioral follow-up should not be combined into one uninterrupted cohort.

Imaging found focal activation at 2 µA in superficial somatosensory cortex. Contacts separated by 60 µm produced greater-than-95% spatial selectivity at 2 µA under the reported analysis. Imaging reached only about 400 µm depth, not layers 4-6, and traded temporal resolution for volumetric coverage. It cannot establish the same activation geometry in deeper cortex or subsecond response timing.

Behavioral detection is not a human percept claim

Head-fixed mice learned to turn a wheel in response to stimulation for a water reward. Thresholds were estimated with an adaptive staircase. In two mice, a single site at 1.5 µA (0.25 nC/phase) supported stable low-threshold detection over long periods. This selected result is not the threshold of all contacts or all animals.

Across layers 4-6, the five animals’ mean thresholds were 1.12, 0.35, 0.88, 1.14 and 0.37 nC/phase. A lowest single measurement of 0.08 nC/phase (0.5 µA) was limited by stimulator resolution and should not replace those means.

Behavioral testing began weeks after implantation and lasted up to 226 days. Final tests occurred at implantation days 308, 294, 284, 264 and 242 for mice 1-5. Testing duration and time since surgery are different clocks.

Failures and tissue boundary

Mouse 1 still detected stimulation at day 308, when its backend connector failed. In mouse 3, accidental delivery of approximately 40 µA DC for a few seconds raised mean detection thresholds from 4.45 ± 0.33 to 13.73 µA. Over a subsequent 116 days, thresholds fell to 4.87 ± 2.14 µA. The proposed local-tissue-damage explanation is the authors’ interpretation, not an independently confirmed lesion measurement.

Histology compared stimulating sites, passive implanted sites and contralateral controls. Neuronal density was preserved in the sampled tissue, while mild glial-marker elevation remained near the implant; GFAP differed from control within the first 40 µm. No difference between stimulating and passive sites is not proof of no tissue response everywhere.

These mouse outcomes do not establish human sensory quality, clinical efficacy, safe dosing for arbitrary waveforms or lifelong implant service.

Primary sources

• 2023 published primary manuscript, Figures 3-5, limitations and methods.
• Publisher record/PDF.