中 / EN

BCIwiki — Brain-computer interface news, research and industry database

Latest

September 2026
09-15 United States Papers Foundations & Methods

AI Should Read Intent in BCI Exoskeletons While Controllers Enforce Limits, Review Says

A brain-computer interface can read neural activity tied to movement, while a powered exoskeleton supplies the force needed to carry it out. But EEG is noisy, muscle signals shift with fatigue and recovery, and the right level of assistance depends heavily on the user and the task, so AI's main value lies in handling several changing signals at once rather than relying on a single fixed input. The review surveys how EEG, EMG and mechanical sensing are combined for exoskeleton control and contrasts two applications with very different goals: stroke rehabilitation and healthy users. The author argues that higher classification accuracy alone is not enough, since latency, calibration, fatigue, uncertainty and physical safety also determine whether a system is truly usable.
Why it matters The review splits exoskeleton control into two layers: AI interprets the user's intent and chooses a high-level assistance strategy, while conventional controllers still enforce mechanical limits on torque, joint angle and speed. It also points out that the goals diverge, with stroke rehabilitation requiring assistance to taper as voluntary control improves, and healthy users mainly wanting to reduce physical effort. The review offers no new trial data and reports no clinical outcomes for any system; its discussion centers on latency, calibration, fatigue and safety, the issues that decide whether such systems can actually be used.
View details →
09-15 Papers Foundations & Methods

Review of 129 EEG-BCI Papers Finds 61.2% State Data Availability, Only 20.9% Share Code

Public datasets for EEG-based brain-computer interfaces (BCIs) keep multiplying, but electrode layouts, task definitions, preprocessing and participant records differ from study to study, making the data hard to reuse across datasets and the results hard to replicate. This review screened 16,920 records down to 129 publications, mapped the fragmentation across structural, semantic, procedural, human/contextual and computational layers, and scored reporting transparency against 10 criteria, with a median score of 9. Data availability was stated in 61.2% of the papers; code or pipeline availability in just 20.9%. Existing standards, ontologies, software platforms and transfer-learning methods each solve part of the problem, the authors conclude, but none yet delivers full semantic interoperability.
Why it matters The bottleneck this review exposes is not a shortage of EEG-BCI datasets but a mismatch between them. Only about a fifth of the 129 papers released code or pipelines, so most published results cannot be rerun as-is. For algorithm developers, that helps explain why models trained across datasets often lose accuracy; for teams building new datasets, it lists the fields to standardize before collection starts.
View details →
09-15 Papers Foundations & Methods

Low-Cost 8-Channel EEG and VR Headset Reach Up to 90% Accuracy in Single-Subject Test

Researchers paired an 8-channel OpenBCI Cyton board and an EEG cap laid out on the international 10-20 system with a Meta Quest 2 headset to capture motor imagery and attention signals, reaching up to 90% subject-specific online classification accuracy with the EEGNet model when the signal was stable. The work targets cognitive control training for attention-deficit/hyperactivity disorder (ADHD), building EEG sensing into a virtual reality (VR) serious game so users can regulate their mental state in real time; earlier approaches mostly relied on hardware-heavy multichannel systems and did not combine motor imagery with attention levels. The team also documented artifacts from high impedance, channel saturation and mechanical tension from the headset strap, showing that a low-cost setup is usable only once skin-electrode impedance and mechanical interference are under control. This is a single-subject feasibility study with no clinical evaluation.
Why it matters The study assembles low-cost 8-channel EEG and a consumer VR headset into a rig that reads mental state in real time, and it is candid about what got in the way: high impedance, channel saturation and strain from the headset strap. For anyone building cognitive-training hardware for homes or clinics, the takeaway is a hardware threshold rather than an efficacy result: 90% is reachable with a stable signal, but stability has to be bought with conductive gel and mechanical fixation. With a single subject, therapeutic use remains a long way off.
View details →
09-15 · Icahn School of Medicine at Mount Sinai +1 more United States Papers Foundations & Methods

4,096-Channel μECoG Array Maps Brain Function During Surgery With 91.3% Channel Yield

During removal of a right parafalcine meningioma, surgeons placed four Layer 7 μECoG arrays, 4,096 electrodes in total, on either side of the central sulcus and recorded somatosensory evoked potentials under 5 contralateral stimulation conditions. Using a 2 MΩ impedance cutoff, 3,739 channels were usable, a 91.3% yield; phase-reversal latencies were 19, 21, 25, 26 and 25 ms, consistent with the standard intraoperative mapping performed in the same operation. The researchers caution that the setup provides dense spatial sampling rather than submillimeter physiological resolution: measured responses were correlated across roughly 3–4 mm of cortex.
Why it matters Surgeons usually locate the central sulcus with sparse subdural strip electrodes that yield only a handful of points. By covering both sides of the sulcus with 4,096 electrodes, this study renders the polarity reversal as a continuous two-dimensional boundary instead of a jump between two neighboring contacts. Notably, the authors draw their own line: electrode pitch is 400 μm, but measured signals correlated across 3–4 mm, so the gain is in sampling density, not resolution. In the same operation, the automated classification agreed with standard mapping.
View details →
09-15 · Central Michigan University United States Papers Foundations & Methods

Adding fNIRS to EEG Fails to Improve Brain-Controlled Stimulation in 16-Person Trial

Combining EEG with functional near-infrared spectroscopy (fNIRS) did not make brain-controlled electrical stimulation more accurate in a blinded randomized trial of 16 healthy volunteers. The hybrid and EEG-only groups showed no statistically significant differences in real-time three-class recall, sense of agency, attention or physical comfort; median recall was 53.5% in the hybrid group and 57.3% with EEG alone. The researchers also released the full EEG-fNIRS dataset.
Why it matters BCI-driven functional electrical stimulation (FES) is seen as one of the more promising routes in post-stroke rehabilitation, but adding fNIRS means more sensors, longer setup and higher cost. This trial tested the add-on head to head and found no difference, a check on the intuition that more signals mean better control; the public dataset lets other groups rerun the analysis. With only 16 healthy volunteers, the result cannot be extended directly to patients.
View details →
09-15 China Clinical Implantation & Procedure

Peking University Third Hospital Joins Chinese Invasive BCI Trial for Spinal Cord Injury

Peking University Third Hospital, a top-tier (Grade 3A) general hospital in Beijing whose neurosurgery department is one of its key specialties, is taking part in a clinical trial of a Chinese-developed invasive brain-computer interface (BCI), neurosurgeons Yang Jun and Yu Tao wrote in the 16th issue of China Hospital CEO magazine. The trial enrolls patients with spinal cord injury and relies on a multidisciplinary team (MDT) to restore hand function after implantation. Invasive BCIs read neural signals through electrodes placed inside the skull or on the cortical surface, so surgery and rehabilitation depend on close coordination among neurosurgery, rehabilitation and imaging departments. The disclosure marks one step in Chinese hospitals' clinical validation of domestic systems. Public information has not named the system or disclosed enrollment, follow-up milestones or primary endpoints, and no efficacy or safety data have been released.
Why it matters For invasive BCIs, the clinical bottleneck is often not the electrode itself but how surgery, rehabilitation and imaging teams work together. By building an MDT into the trial workflow, Peking University Third Hospital signals that Chinese systems are moving from one-off technical demonstrations toward organized clinical validation. Disclosure is thin, with no system name, enrollment figure or follow-up schedule; what can be confirmed is that the hospital is now on the clinical front line of China's domestic invasive BCI effort.
View details →
09-15 · Université de Montréal +1 more CanadaUnited States Papers Foundations & Methods

Adaptive BCI Decoders Concentrate Task Information in Fewer Monkey Motor Neurons

After days of practicing a brain-computer interface with adaptive decoders, rhesus macaques needed significantly fewer motor cortex neurons to reach 80% task classification accuracy than in early sessions (p = 9.8 x 10^-4). Fixed decoders produced no such compaction. A neural network model suggests the adaptive decoders directly contribute to these more compact representations; the authors propose this may be because the decoders alter the error information used for credit assignment. The finding implies that assistive algorithms do not merely boost immediate BCI performance but actively reshape how the brain learns over time -- a consideration for real-world decoder design. The study used 128-channel microwire arrays in motor and premotor cortices of two monkeys and has not been tested in humans.
Why it matters This study shows that BCI decoders do not just passively translate neural signals: adaptive decoders reshape motor cortex representations, concentrating task information into fewer neurons and neural patterns. Decoder designers will need to balance immediate performance against long-term neural plasticity. The results come from two macaques and await human validation.
View details →
09-14 · Beijing Municipal Science and Technology Commission +2 more China Policy & Regulation Industrial Planning

Beijing's BCI Plan Sets No Output Targets, Bets on Companies Instead

Beijing issued China's first provincial-level brain-computer interface (BCI) action plan on January 7, 2025, setting every 2027 and 2030 target in terms of companies and platforms rather than output or revenue. The plan calls for three to five unicorns or would-be unicorns by 2027 and around 100 innovative small and medium-sized firms by 2030. On the payment side, Beijing prices invasive BCI implantation at RMB 7,980 per procedure and covers all three BCI service fees under Category A of basic medical insurance, limited to four hospitals.
Why it matters Where other provinces count products, procedures or revenue, Beijing counts companies, leaving delivery to how investors value local start-ups; its payment policy is further ahead, with the highest verified implant price in China and full Category A coverage, but only at four hospitals.
View details →
09-14 · Shanghai Municipal Science and Technology Commission China Policy & Regulation Industrial Planning

Shanghai Aims to Partially Restore Sight With BCIs by 2030

Shanghai's science and technology commission issued a BCI action plan on January 10, 2025, pledging to help blind patients partially regain sight by 2030 and to have more than five invasive or semi-invasive products complete medical device type testing by 2027. The city's healthcare security bureau created three BCI service fee items in August 2025 but left prices to hospitals and explicitly excluded them from basic medical insurance payment.
Why it matters Shanghai wrote the most clinically specific goals of any province, down to restoring vision, yet its own pricing notice sets no government price and no reimbursement, a combination found nowhere else in China.
View details →
09-14 · Department of Economy and Information Technology of Sichuan Province China Policy & Regulation Industrial Planning

Sichuan Targets 3,000 Invasive BCI Operations a Year by 2030

Sichuan's industry department and seven other agencies issued a BCI and human-machine interaction plan on May 12, 2025, targeting 3,000 invasive BCI operations a year by 2030 and the registration of three invasive and five non-invasive products by 2027. The province prices BCI services in five tiers by hospital grade, with implantation ranging from RMB 6,583 to RMB 4,389, all paid out of pocket.
Why it matters Sichuan is the only province to make procedure volume a headline target, but with no interim milestones the 3,000 figure works more as a mobilizing signal; its lowest price tier is the lowest verified in China, and none of it is reimbursed.
View details →

© 2026 BCIwiki.com Digest Topics Tips Subscribe Revisions About