Papers

A collaboration involving Nanchang University proposed a deep learning network to classify motor imagery tasks from EEG signals. Across three public BCI datasets, it achieved average accuracies of 80.21% for Dataset 1's 4-class tasks, 90.19% for Dataset 2's 2-class tasks, and 97.32% for Dataset 3's 4-class tasks, exceeding the strongest comparison network on each dataset. The approach combines multiscale processing with temporal and channel attention, while augmenting training data through segmentation and reconstruction. These results reflect classification under the tested dataset and training conditions, not demonstrated real-time BCI or clinical performance.

Authors affiliated with Tianjin University in China, UCSF and collaborating institutions review BCI-related work in cancer pain, brain tumor surgery and tumor mechanisms. They examine neurofeedback, closed-loop regulation and neural recording while distinguishing BCIs from recording and surgical decision-support techniques.

A systematic review and meta-analysis pooled 11 randomized controlled studies involving 559 participants to assess BCI training for cognition and activities of daily living after stroke. Compared with conventional rehabilitation, pooled results showed statistically significant improvements in both outcomes. However, evidence certainty was rated low for cognition and very low for daily living, so these findings do not establish definitive benefits. Most subgroup comparisons found no significant differences. The authors describe preliminary promise but conclude that current evidence is insufficient to recommend BCI for emotion or speech rehabilitation.

A systematic review and meta-analysis pooled seven randomized controlled trials to assess whether adding non-invasive brain stimulation to BCI training improves upper limb rehabilitation after stroke. The combined approach did not outperform BCI alone immediately after intervention and showed a smaller improvement on an upper limb motor scale. Subgroup analysis suggested that studies combining transcranial direct current stimulation with BCI mainly drove this result; repetitive transcranial magnetic stimulation plus BCI showed no significant difference. Evidence certainty ranged from moderate to very low, and sensitivity analyses indicated unstable findings. The results neither establish a clear additional benefit nor show that every combined protocol is ineffective.

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.

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.