BCI Weekly - April 12, 2026
2026 week 15 (April 6–12). Methods-heavy week. No major industry news.
This was a quiet week for launches, rounds, and trials. No Neuralink update, no FDA clearance, no megaround. If you’re scanning for that kind of signal, you won’t find it here.
What you will find is a methods week that actually matters. Two optical voltage indicators. A new tACS hardware system with real impedance validation. A PNAS paper that addresses a question the field has been circling for decades. And a $24M Canadian neuroprosthetics initiative that deserves more attention than it’s getting.
The paper worth reading first
BCI-based neurofeedback training enables transferable control of cortical state switching in humans
PNAS, April 2026
Closed-loop regulation of sensorimotor activity has been possible since the 1960s. The question that hadn’t been settled: can people learn to switch cortical states volitionally, and does that learned control transfer → meaning it generalizes beyond whatever training setup produced it?
This paper says yes.
The transfer result is what makes it interesting. Most neurofeedback work shows that people can learn to modulate a signal in a specific context. Fewer papers ask whether that control extends when the context changes. If it does, the implications for decoder longevity and session-to-session calibration are direct: users who build genuine state-switching skill may need less recalibration over time.
The clinical story of the week
Dr. Christian Iorio-Morin — From Gamma Knife to Neuroprosthetics: The Future of Functional Neurosurgery
Neural Implant Podcast, April 2026
Dr. Iorio-Morin is a functional neurosurgeon and professor at the Université de Sherbrooke. The episode covers movement disorders, chronic pain, gamma knife, and neuroprosthetics. But the real reason to listen is RE-MOVE.
RE-MOVE is a 14-investigator interdisciplinary project, awarded a $24M grant from Canada’s New Frontiers in Research Fund (Transformation stream). The core problem it’s solving: spinal circuits responsible for muscle control often survive injury intact. The failure is signal interruption, not the loss of the muscles or neurons themselves. RE-MOVE builds a bridge → cortical or spinal sensors detect movement intention, and neuromodulation activates the paralyzed limbs.
The system uses cortical sensors, haptic gloves, and human-machine interfaces. It was designed from the beginning with “ethics by design”; law and ethics are integrated into the research structure, not retrofitted after the engineering is done.
This is Canada’s most-resourced closed-loop neuroprosthetics initiative currently active. It hasn’t gotten the coverage it deserves outside of the Québec academic press. Worth tracking.
The tooling news: two GEVI preprints in the same week
Two bioRxiv preprints dropped this week targeting the same bottleneck: two-photon voltage imaging with genetically encoded voltage indicators (GEVIs) hasn’t been practical enough for most labs.
FORCE1s — accessible two-photon recordings in vivo
FORCE1s is a green, positive-going GEVI that brightens from a dark baseline during depolarization. In awake mice, it reports spikes at approximately 100% ΔF/F. The “accessible” framing is substantive: it works on standard resonant-scanning microscopes, not just specialist hardware. It also supports MEMS-based two-photon miniscopes for freely moving mice and multiplexed voltage-neurotransmitter imaging. The equipment bottleneck is the main reason GEVI adoption has been slow. FORCE1s directly targets that.
Designer indicators for subthreshold voltage dynamics
The companion preprint pushes further: subthreshold voltage (the integration layer below spike threshold) has been nearly inaccessible with two-photon GEVIs because sensitivity is too low for millivolt-scale signals. This work reports a multiparametric refinement approach to close that gap.
Together, these two preprints represent a coordinated tooling push: spike-resolved readout on accessible hardware (FORCE1s), plus subthreshold dynamics for richer ground-truth data. Both are preprints, so replication and journal review pending. But for labs choosing optical recording stacks, these are the ones to track.
tACS hardware gets a real validation standard
PD-stim — phase-difference tACS system for dual-site neuromodulation
Frontiers in Neuroscience, April 2026
Dual-site transcranial alternating current stimulation (tACS) requires precise phase control between stimulation sites. Conventional setups often lack it. Prior validation work is seldom done under realistic biological impedance conditions.
PD-stim addresses both problems: programmable inter-regional phase-difference stimulation with impedance-aware hardware validation.
Why phase precision matters: recent beta-band ds-tACS work shows in-phase stimulation and anti-phase stimulation produce opposite effects. In-phase enhances fronto-motor connectivity and predicts better inhibitory control; anti-phase reduces connectivity and speeds go responses. The direction of effect flips based on the phase relationship. A system without stable phase control can’t reliably reproduce those results. PD-stim is the platform that makes this class of experiments reproducible at scale.
A second tACS paper this week (Scientific Reports) reported that network-nodal tACS induces right-lateralization of thalamocortical connectivity. Connectivity readout is still a proxy, and site/frequency specificity is unresolved… but it’s another data point that non-invasive network-level stimulation produces measurable, directional effects.
Also on the radar
Epidural and DRG stimulation in SCI: therapy-specific motoneuron activation (JNP) — Recruitment is selective, not uniform, and shifts with stimulation parameters. Not a decoding BCI study, but directly relevant to the stimulation side of RE-MOVE-class systems.
Motor unit firing rates persist during neurofeedback (JNP) — Motor-unit firing-rate structure in tibialis anterior is preserved, not reorganized, during neurofeedback. Practical constraint for motor-BCI signal interpretation: feedback may not reshape the firing distributions that some decoder designs assume it will.
Combined electrophysiology + functional ultrasound: roadmap review (JNP) — fUSI’s whole-brain spatial coverage combined with electrophysiology’s temporal precision is a compelling pairing for chronic closed-loop interfaces. This review maps the practical tradeoffs. Still mostly a pre-human-trial modality, but the roadmap is clear.
EEG data augmentation survey (Frontiers in Neuroscience) — Systematic review on augmentation strategies for EEG emotion recognition and cognitive workload decoding. Prior work shows augmentation improves EEG decoding by ~29% on average; this survey organizes the taxonomy and flags task-specific pitfalls. Now the canonical starting reference for this problem.
sEMG bulbar assessment in ALS (Frontiers in Neuroscience) — 60 quantitative features from six craniofacial muscle groups, fully automated. Directly relevant to speech-BCI candidacy: objective bulbar stratification determines who is a viable candidate for speech-motor neuroprosthetics and when in disease progression to intervene.
Error equation predicts the brain’s ability to generalize (The Transmitter) — Four statistical measurements of neural network geometry predict generalization in both biological and artificial networks. Trade-press summary, primary paper still needed, but the bridge between ML-style generalization metrics and neuroscience is a useful conceptual frame for decoder design.
What I’m following up on
PNAS neurofeedback: Need full text to extract training duration, signal type, and transfer paradigm specifics. Would someone send me the paper?
FORCE1s: Tracking toward journal submission; benchmark comparison against ASAP4 GEVIs and rhodopsin-based 2P indicators.
RE-MOVE: Watching for ClinicalTrials registrations and first human-implant protocol publications from the Sherbrooke team.
PD-stim: Does the supplement include E-field characterization data comparable to the prior HD ds-tACS validation literature?
My take
Thin week for news. Strong week for boundaries.
The best work here didn’t promise new capabilities, it mapped where systems hold up. PNAS suggests learned cortical control can transfer beyond its training context. The motor-unit paper shows a limit: firing-rate structure stays rigid during neurofeedback. PD-stim pushes dual-site tACS toward real validation. FORCE1s makes two-photon voltage imaging work on standard microscopes. RE-MOVE is building a $24M closed-loop neuroprosthetics pipeline with ethics baked in from day one.
No demos, only infrastructure. The bottleneck isn’t decoding accuracy at the moment - it’s transfer, stability, and knowing where the system breaks before you put it in a person.
If you’re building, investing in, or researching this space: how do you weigh non-invasive neuromodulation at scale against the slower, harder path of implant infrastructure?