中 / EN

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

Latest

October 2026
10-19 Policy & Regulation Industrial Planning

Shanghai Science Commission Sets November 4 Start for National Brain Research Applications

The Shanghai Municipal Science and Technology Commission in eastern China is coordinating applications for the first batch of 2027 projects under the national brain science and brain-inspired research initiative. Online submission runs from November 4 to December 4, 2026, through the national science and technology management platform, with attachments uploaded as scans.
Why it matters The notice supplies Shanghai applicants with a submission window and filing route; eligibility and project requirements remain in the national guidelines.
View details →
10-19 Papers Foundations & Methods

Huazhong University of Science and Technology Aligns Cross-User Visual EEG With One Labeled Trial per Class

Researchers at Huazhong University of Science and Technology proposed KDHNet, a hybrid network combining analytical signal processing with neural networks to decode steady-state visual evoked potentials, stable brain responses to visual stimulation, with limited individual calibration data. Across three public datasets, it achieved the highest mean accuracy and information transfer rate among the compared methods in 14 of 15 analysis-window settings and all ten calibration-ratio settings in within-subject evaluation. In cross-subject evaluation, performance ranked first or second, with generally larger gains when less calibration data were available. These dataset results do not establish real-world plug-and-play operation.
Why it matters Cross-subject decoding still requires one labeled trial per class from the target participant to align data. Limited calibration therefore does not mean zero calibration, a distinction that separates the authors' plug-and-play goal from the experimental setup.
View details →
10-18 Papers Foundations & Methods

Tongji and Collaborators Cut Token Use 64.02% in EEG Agent Experience-Reuse Test

Authors affiliated with Tongji University, Ant Group, Shanghai Jiao Tong University and collaborating institutions introduced the NeuroAct EEG research agent. In an experience-reuse experiment, BCIC-IV-2a balanced accuracy rose from 34.66% to 50.81% while LLM token use fell 64.02%. The preprint also evaluates general analysis, cross-task optimization, single-task research and collaborative prediction.
Why it matters Testing conditional recommendations on later tasks links experience reuse to both predictive performance and research cost.
View details →
10-17 Papers Foundations & Methods

Motor Imagery EEG Framework Classifies Signals Across Two Datasets

Researchers developed a motor imagery EEG classification framework, tested it on two competition datasets, and deployed it on an NVIDIA Jetson Nano. Subject-dependent accuracy reached 93.67% on BCIC III IVa and 79.14% on BCIC IV 2a. Under leave-one-subject-out evaluation, which tests generalization to a person excluded from training, average accuracy was 76.49% and 55.50%, respectively. Deployment tests also measured complete pipeline latency and throughput. These findings describe dataset classification performance and device processing speed, not demonstrated improvements in motor impairments.
Why it matters The framework performed differently under subject-dependent and cross-subject evaluation. Keeping these conditions distinct helps readers separate performance with person-specific training from generalization to someone excluded from training.
View details →
10-16 Papers Rehabilitation & Assistance

Review Finds Two BCI Training Approaches Improve Post-Stroke Arm Scores, With Low Certainty

A systematic review and network meta-analysis of 22 randomized controlled trials involving 956 post-stroke patients compared BCI training with different feedback modalities. Robotic or exoskeleton-assisted feedback and multimodal feedback produced statistically significant improvements in upper-limb motor scores versus usual rehabilitation, but evidence certainty was low. Non-BCI robotic training also showed benefit, so the robotic BCI effect cannot be attributed to BCI alone. No BCI feedback modality was statistically superior to another. Visual feedback ranked highest on a ranking metric, yet its comparison with usual rehabilitation was non-significant and supported by very low-certainty evidence.
Why it matters Treatment rankings do not establish superiority. Visual feedback had the highest ranking metric in this analysis but no significant advantage over usual rehabilitation. Higher-quality head-to-head trials are still needed to compare feedback options.
View details →
10-16 · Neuracle China Policy & Regulation Regulatory Approval

Neuracle UDI Listings Cover Two BCI System Models and Two Electrode Kit Models

China’s NMPA UDI database lists Neuracle’s NEO-M and NEO-M Pro hand motor function compensation system models under registration certificate 国械注准20263120537, and K1014-25 and K1014-35 implantable EEG electrode kits under 国械注准20263120536. The descriptions identify the system’s implantable components, EEG transceiver, pneumatic glove and software.
Why it matters Reading the records together distinguishes system and component registrations and avoids counting model-specific UDI entries as four separate product approvals.
View details →
10-16 Papers Foundations & Methods

New Framework Improves Motor Imagery Decoding with Missing EEG Channels

Researchers proposed TriRouter to classify motor imagery when EEG channels are sparse or corrupted, allocating contributions from different processing paths according to channel availability. On BNCI2014-001, accuracy exceeded the study's strongest corrupted-EEG baseline by 3.68, 4.02, and 6.46 percentage points at nominal missingness settings of 25%, 50%, and 75%, respectively. The authors also report effectiveness under structured regional missingness and in an additional low-channel dataset evaluation. These dataset results do not directly establish BCI performance in practical use.
Why it matters Beyond varying the amount of missing data, the study tests channels missing in regional groups. This adds a spatial dimension to evaluating how the model handles different patterns of electrode availability.
View details →
10-16 Papers Foundations & Methods

Compact Model Converts Brain Signals to Text in Two Benchmark Tests

Researchers proposed BEST, a two-stage method that converts neural activity into text, to examine whether speech brain-computer interfaces need large models. On the Brain-to-Text '24 and '25 benchmarks, BEST achieved word error rates of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Word error rate measures errors relative to reference text; lower values indicate fewer errors. The authors argue that larger decoders do not consistently perform better and that alignment between neural signals, speech, and text matters more than model size. These benchmark results do not directly establish everyday communication performance.
Why it matters Beyond parameter count, the study examines how different forms of information connect. The authors identify neural-speech-text alignment as a key factor, offering a specific angle for understanding how a compact model can decode neural activity.
View details →
10-16 Papers Rehabilitation & Assistance

Study Involving Air University Tests Ankle-Movement Brain Signal Decoding Offline and in Simulated Online Use

A pilot study involving authors from Air University and other institutions recorded functional near-infrared spectroscopy signals from 20 healthy participants and 7 patients with ankle motor impairment. It sought to distinguish rest, dorsiflexion and plantar flexion and generate exo-ankle control commands. In cross-participant offline validation among healthy participants, a support vector machine achieved a mean accuracy of 72.50% ± 13.8%. Simulated-online testing with participant-specific validation reached its highest mean accuracy of 62.2% ± 14.1% at a 1.5-s step size. The patient group achieved 57.23% ± 5% under leave-one-trial-out validation. These are proof-of-concept results, not demonstrations of prospective real-time exo-ankle control or rehabilitation efficacy.
Why it matters The offline evaluation tests performance on a participant excluded from training, whereas simulated-online testing uses participant-specific leave-one-trial-out validation. Their accuracy difference cannot simply be interpreted as the performance cost of moving from offline to real-time operation.
View details →
10-15 Papers Foundations & Methods

Study Links Income and Product Features to Interest in Mind-Writing Pens

A study of individuals with disabilities examined purchase intention and price acceptance for mind-writing pens. Income was significantly positively associated with both outcomes, while age showed no significant overall association. Battery life and head-mounted design were associated with greater purchase intention; weight and recognition accuracy were positively associated with price acceptance, although this does not establish a preference for lighter or heavier devices. Price acceptance declined as prices rose and was highest below 1,000 yuan. Subgroup analyses identified differences in priorities across age and income groups. These findings concern purchasing preferences, not verified product performance or clinical benefits.
Why it matters Different features were associated with willingness to buy and price acceptance: battery life and wearing design with the former, weight and recognition accuracy with the latter. Keeping these outcomes separate helps distinguish interest in a product from acceptance of its price.
View details →

© 2026 BCIwiki.com Digest Topics Tips Subscribe Revisions About