Newsfeed
Real-time intelligence on brain-computer interface funding, clinical trials, research, and policy.
EEG Gait Decoder Hits 55% Closed-Loop Success at 70ms
A 2-block EEG BCI achieves 55.3% closed-loop gait initiation and 70.5ms prediction latency in pilot exoskeleton testing.
Read →SoniSpeech: 26.3% WER Open-Vocab Silent Speech Dataset
SoniSpeech debuts 34hrs of trimodal wearable silent speech data, benchmarking 26.3% WER on open-vocab recognition.
Read →32–128 Mbps Battery-Free BMI via RF Backscatter
New preprint proposes RF backscatter + near-field charging to hit 32–128 Mbps wirelessly without an implant battery.
Read →Octopus 16: 16-Electrode Wireless EEG BCI for $250
PiEEG's Octopus 16 packs 16 electrodes into a coin-sized wireless EEG device for $250, enabling hands-free game control via brain signals.
Read →Tether Evo: 3 Papers Target Cross-Subject BCI Decoding
Tether Evo's 3 peer-reviewed papers show one model can decode speech, vision, and music across multiple patients.
Read →Organoid Stimulation Shrinks Response from 93% to 10%
Repeated daily stimulation collapsed organoid network engagement from 93% to 10%, with a naive-control design isolating stimulation from maturation.
Read →EMG Gesture Decoding Hits 0.688 F1 Across Sessions
A montage-agnostic encoder retains 0.688 macro-F1 across EMG sessions vs. 0.540 for per-user LDA, without any recalibration.
Read →0.234 F1 Gain: EMG Encoder Skips Per-User Calibration
A montage-agnostic sEMG encoder beats per-user LDA by 0.234 macro-F1 on DB1, potentially ending calibration bottlenecks in myoelectric prosthetics.
Read →GRU Beats Mamba on Brain-to-Text '25 Benchmark
GRU decoders outperform Mamba hybrids on Brain-to-Text '25, hitting 12.62% PER and 21.19% WER.
Read →More Electrodes, Faster Minds? BCI Bandwidth Limits
A new arXiv perspective argues BCI bandwidth gains yield nonlinear returns on meaningful human I/O.
Read →XAI Requirements for Stroke BCI: 3 Patients, 4 Methods
A 6-participant formative study reveals how to elicit XAI preferences from stroke survivors with aphasia for rehabilitation BCIs.
Read →Bayesian HD-MEA Sampling Cuts Electrode Waste by 17pp
Bayesian Thompson sampling selects 100 of 529 electrodes dynamically, beating static selection by 17.2pp at 34 hours.
Read →84.1% Perception Decoding from Resting-State V1 Activity
MAE pretrained on 14.6h of spontaneous V1 activity hits 84.1% perception decoding accuracy with no labeled training data.
Read →SSCDL Tackles Neural Drift in Long-Term BMI Decoding
New SSCDL framework uses teacher-student consistency and disentangled motor decoding to maintain BMI accuracy across days.
Read →China's 15th Five-Year Plan Backs Non-Invasive BCI Rehab
China's CDPF commits to non-invasive BCI-powered smart rehab devices in its 2026–2030 disability plan.
Read →Nimbus Personalizer Runs One EEG API Across Five Encoders
Nimbus Personalizer proposes a single trunk-agnostic API for EEG personalization across five classical encoders and one foundation model.
Read →Single EMG Grid Hits 89.5% Accuracy for Hand Gesture Recognition
Single extensor-grid HDsEMG achieves 89.5% balanced accuracy across 16 gestures, nearly matching dual-grid systems at half the hardware cost.
Read →China's NEO Brain Implant Gets Reimbursement in 48 Hours
China's NEO motor BCI got a reimbursement billing code 48 hours after NMPA approval — and the US is still catching up.
Read →BFT Eliminates Backprop in EEG-BCI Test-Time Adaptation
BFT framework removes backpropagation from EEG-BCI test-time adaptation, cutting compute overhead while preserving decoding robustness.
Read →Zero-Shot EEG Decoding Reaches 0.9799 nRMSE Across Subjects
Zero-shot Transformer EEG decoding hits 0.9799 nRMSE on unseen subjects via progressive unfreezing on Healthy Brain Network data.
Read →