Brain–computer interfacing (BCI) offers novel methods to facilitate participation in music production, providing access for individuals who might otherwise be unable to take part (either due to lack of training or physical disability). This article describes the development of a BCI system for conscious or ‘active’, control of parameters on an audio mixer by generation of synchronous MIDI Machine Control messages. The mapping between neurophysiological cues and audio parameters must be intuitive for a neophyte audience (i.e. one without prior training or the physical skills developed by professional audio engineers when working with tactile interfaces). The prototype pilot system, dubbed MINDMIX (a portmanteau of ‘mind’ and ‘mixer’), was subsequently evaluated by neophyte and experienced music producers across utility and mapping congruency. Neophyte participants rated the system higher in utility than experienced participants, whilst both groups rated mapping congruency similarly. Assuming a degree of synonymy between utility and usefulness, and between congruency and intuitiveness, this suggests that whilst the system might be useful for the neophyte audience, experienced users are likely to exhibit a preference for existing technology over the MINDMIX system. In the future, specific evaluation of discrete mappings would be useful for iterative system design.

Developing effective Brain-Computer Interfaces (BCIs) based on Imagined Speech (IS) is a significant challenge, largely due to high inter-subject variability in neural patterns. This study introduces a novel analytical framework to address this issue by integrating functional, effective, and complex network analyses with a more naturalistic sentence-level experimental protocol. Our findings confirm that while IS connectivity networks are characterized by considerable variability across individuals, our methodology successfully identifies a core set of stable pathways that persist across subjects. Specifically, we identified three principal pathways: a motor-language network in the left hemisphere driven by delta-band activity (CL→FR,CR consistent in 60% of subjects), a right-hemisphere network relayed to motor planning areas via gamma-band activity (TR→CL in 40% of subjects), and a top-down visual-spatial network involving parietal regions (POL→CR in 60% of subjects). In parallel, complex network analysis reveals the gamma frequency band to be the most functionally integrated and robust spectral signature, exhibiting significantly higher mean connectivity strength compared to all other bands (e.g., p=0.0015 vs. beta) and appearing consistently in 6/10 subjects. By distinguishing these stable neural markers from subject-specific activity, this work provides more reliable EEG-based signatures for the future development of advanced speech BCIs.

Multimodal MRI–EEG fusion for brain–computer interface applications using a lightweight CNN and attention in offline Parkinson’s disease diagnosis

Brain–Computer Interface driven BOPPPS: Empirical evidence for enhanced educational practices

A Brain-Computer Interface (BCI) is designed for human-computer interactions without body movement. To improve the expression of input features closely related to a given BCI task, we propose a convolution network called GANet to analyze Event-Related Potential (ERP) in the BCI task. This model introduces the parallel convolution to extract multi-scale features in electroencephalogram (EEG) data. In addition, a Graph-based Attention (GA) mechanism is used to model interdependencies among different EEG channels. Experiments are conducted on a public dataset of 15 subjects in the specific-subject and cross-subject scenarios. The results indicate that the GANet achieves state-of-the-art performance with an accuracy of 99.75% in the specific-subject scenario and an accuracy of 81.37% in the cross-subject scenario. Different structures are discussed to analyze the contributions of the parallel convolution in feature extraction and the GA module in feature expression. GANet shows satisfied performance and good generalization in the BCI task. Our codes are publicly available at https://github.com/Debbie-85/GANet.

Bioactive carbon dots as antibacterial neural interface for neuron recording

Deployable medical devices are designed to be compact during insertion and expand after surgical placement. Devices such as neural interfaces can leverage deployment to minimize the size of the foreign body cascade near the electrode, potentially improving chronic recording and stimulation performance. To trigger deployment, a stimuli responsive material can be used. However, external stimuli are difficult to supply within tissues. Intrinsic changes upon implantation, such as water uptake, are difficult to control and may lead to device failure. Here, we describe a strategy to deploy small-scale structures into soft tissues after insertion without the need for any stimulus. Photoresponsive liquid crystal networks (LCNs) are crosslinked after self-assembly of monomers and adopt a programmed 3D form at room temperature (RT). The trans-cis isomerization of azobenzene enables the 3D LCN films to be flattened by UV light before insertion and revert to 3D forms over 5h at body temperature. A film programmed to adopt a cone shape with a diameter of 531µm can actuate to 53µm in height. Rigidity of the films enables penetration into and deployment within soft tissues. The described materials could potentially enable self-deployable biomedical devices, including neural interfaces with sub-mm features.

The unique gyral patterns of the human brain demand patient-specific neural interfaces to achieve precise neuromodulation, mitigate adverse tissue responses, and optimize therapeutic efficacy and safety. One-size-fits-all, conventional rigid electrocorticography (ECoG) electrodes, standardized for mass production through lithographic techniques, exhibit limited conformability to the brain's heterogeneous cortical topography. This mechanical mismatch results in poor electrode-tissue contact, signal loss, and foreign body responses. To address these limitations, we present an integrated novel platform, synergizing MRI-based anatomical mapping, finite element analysis (FEA)-optimized mechanical design, and direct ink writing (DIW) 3D printing to fabricate electrodes customized to individual gyral patterns. The resulting honeycomb-inspired printable gel electrode (HiPGE) employs a bioinspired honeycomb architecture with ultra-soft hydrogels, engineered to match the bending stiffness of brain tissue (0.1-10kPa) while maintaining cost-efficiency and long-term durability. This mechanical congruence ensures exceptional cortical conformability and adaptive interfacing, circumventing the geometric and material limitations of traditional rigid electrodes. By combining patient-specific design with scalable fabrication, our platform establishes a transformative framework for neural interface engineering, enhancing precision, biocompatibility, and functional performance in neuromodulation therapies and neuroprosthetic applications.

Handwriting brain-computer interfaces (BCIs) have enabled high performance brain-to-text communication for paralyzed individuals. However, the detailed parameters of handwriting movement and their cortical representations remain incompletely understood. Here, we recorded intracortical neural activity from a paralyzed subject and found distinct neural representations for strokes and pen lifts with respect to two-dimensional (2D) velocity on the writing plane, indicating that 2D kinematics alone cannot fully account for the observed neural variance. To address this, we acquired multidimensional handwriting data from healthy subjects, including 3D velocity, grip force, writing pressure, and multi-channel electromyographic (EMG) signals. Incorporating these additional dimensions beyond 2D velocity significantly improved the interpretability of neural signals for both strokes and pen lifts. We further leveraged these additional dimensions to enhance handwriting decoding performance. Together, our findings indicate the motor cortex encodes handwriting as multidimensional movement and highlight the importance of multidimensional features for improving the performance of handwriting BCIs.

Brain–computer interfaces (BCIs) enable direct communication between neural activity and external devices, offering novel input channels that expand and redefine human–computer interaction (HCI). Originally developed from early electroencephalography (EEG), BCIs have evolved to include both invasive and implantable systems, with applications extending beyond clinical settings into rehabilitation, assistive communication, and interactive entertainment. This paper provides a concise survey of BCIs within HCI. The review begins by introducing neuroscience and signal-processing foundations, including signal acquisition—such as EEG, electrocorticography (ECoG) and intracortical recordings—as well as preprocessing, feature extraction, and decoding methods. Subsequently, representative application domains are explored, including: (1) assistive technologies, exemplified by brain-controlled wheelchairs, prosthetics, neurorehabilitation systems; (2) communication systems, such as P300 spellers, intracortical spelling interfaces, and emerging speech neuroprostheses; and (3) entertainment and gaming applications, including neurogaming and VR/AR integration. The discussion then turns to emerging trends—AI-enhanced decoding, wearable and wireless form factors, and multimodal integration—along with cross-cutting challenges such as neural data privacy, user autonomy and consent, usability, and regulation. The paper concludes with design implications for HCI and proposes a research agenda for safe, reliable, and equitable BCI deployment.

Auditory Brain–Computer Interfaces (BCIs) constitute the vital intervention for profound sensorineural hearing loss where the auditory nerve is compromised, yet their clinical efficacy remains restricted by substantial biological bottlenecks and limited spectral resolution. This review critically examines the evolutionary paradigm of auditory restoration, tracing the transition from static physical replacement to dynamic biological symbiosis. We systematically analyze physiological barriers across cochlear, brainstem, and cortical levels, elucidating how rigid interfaces provoke chronic tissue responses and why linear encoding protocols fail in distorted central tonotopy. The article synthesizes emerging methodologies in material science, demonstrating how soft, bio-integrated electronics and biomimetic topologies effectively address mechanical impedance mismatches. Furthermore, the trajectory of neural encoding is evaluated, highlighting the paradigm shift from traditional envelope extraction to deep learning-driven non-linear mapping and adaptive closed-loop neuromodulation. Finally, the potential of high-resolution modulation techniques, including optogenetics and sonogenetics, alongside AI-facilitated intent perception for active listening, is assessed. It is concluded that future neuroprostheses must evolve into symbiotic systems capable of seamlessly integrating with neural plasticity to enable high-fidelity cognitive reconstruction.

Small-world scale-free brain networks from EEG with application to motor imagery decoding and brain fingerprinting.

Abstract The similarity of neural activity in finger movements poses challenges for accurate decoding using many non-invasive imaging techniques. Magnetoencephalography (MEG), with its relatively high spatial resolution, offers the potential to capture the underlying dynamic neural differences. In this study, we recorded MEG signals during single extension movements of the right-hand fingers, examining the time-varying cortical activation patterns across different frequency bands and their contribution to decoding finger movements. Our results demonstrate that signals below 8 Hz not only enable effective movement classification but also reveal millisecond-scale neural activation patterns in the sensorimotor cortex. Furthermore, incorporating the spatiotemporal dynamics of neural activity may enhance decoding performance for fine motor control. These findings highlight the value of integrating temporal, frequency, and spatial dimensions in studying motor neural activity and underscore MEG’s potential for broader applications in movement-related neurophysiology and brain-computer interface research.

Semantic decoding is a crucial approach for investigating the neural mechanisms underlying language processing and representation. Informed by brain-computer interface (BCI) technology, this study investigated methods for decoding semantic information, with an emphasis on the neural representations of semantics in language perception. Due to the limited availability of electroencephalography (EEG) datasets containing Chinese linguistic stimuli, we have specifically designed a semantic task paradigm as a promising attempt to decode language comprehension and expression in patients with aphasia using scalp EEG. This paradigm fully incorporates the processes underlying both speech perception and speech imagery by adopting tasks such as overt speech perception and silent speech imagery. Firstly, Seventeen participants of aphasia patients and healthy subjects were recruited for EEG data collection. Secondly, we constructed a deep learning model termed Time-Frequency-Spatial Channel Attention Network (TFSANet), which processes both time-domain and frequency-domain features to extract key neural signatures associated with semantics. By optimizing the model and employing multidimensional feature extraction mechanisms, we significantly improved the model's ability to decode semantically relevant EEG features. Finally, the experimental results demonstrate the proposed TFSANet could decode semantic information from EEG for ten categories of four-word phrases under an "auditory-guided" paradigm with an accuracy of 60.73% and 75.09% for aphasia patients and healthy subjects respectively.

Reconstructing speech from neural recordings is crucial for understanding human speech coding and developing brain-computer interfaces (BCIs). However, existing methods trade off acoustic richness (pitch, prosody) for linguistic intelligibility (words, phonemes). To overcome this limitation, we propose a dual-path framework to concurrently decode acoustic and linguistic representations. The acoustic pathway uses a long-short term memory (LSTM) decoder and a high-fidelity generative adversarial network (HiFi-GAN) to reconstruct spectrotemporal features. The linguistic pathway employs a transformer adaptor and text-to-speech (TTS) generator for word tokens. These two pathways merge via voice cloning to combine both acoustic and linguistic validity. Using only 20 min of electrocorticography (ECoG) data per human subject, our approach achieves highly intelligible synthesized speech (mean opinion score = 4.0/5.0, word error rate = 18.9%). Our dual-path framework reconstructs natural and intelligible speech from ECoG, resolving the acoustic-linguistic trade-off.

Abstract Neural interfaces for monitoring and modulating spinal nerve activity are increasingly being designed to be flexible and stretchable to enhance their biomechanical compatibility and integration. However, excessive flexibility introduces practical limitations such as difficulty in insertion into narrow spinal spaces and long-term electrical instability, hindering real-world applications. In this study, we developed a spinal nerve interface by incorporating a liquid-metal conductor and dynamic stiffness-based variable-compliance structure, which can address the challenges of current flexible neural interface technologies. During insertion, the dynamic stiffness enhancer minimizes unintended buckling and ensures minimally invasive implantation into the intended target. The proximity of the proposed device to the spinal cord increases as it flexes automatically and rapidly in a humid environment. The liquid-metal conductor maintained stable electrical properties in freely moving rats, ensuring reliable and sustained functionality. This study lays the foundation for practical, fully implantable spinal bioelectronics designed with a focus on ease of implantation and long-term functionality.

Motor imagery-based brain computer interface (MI-BCI) have been increasingly adopted in neurorehabilitation and related fields. The performance of MI-electroencephalogram (MI-EEG) decoding algorithms is central to the advancement of MI-BCI. However, current studies often lack rigorous investigation into the brain's complex network organization. Moreover, most existing methods do not incorporate the cross-frequency coupling (CFC) phenomena that occur during MI into their algorithmic designs, nor do they adequately account for how temporal dynamics across different MI stages influence decoding outcomes. To address these limitations, we propose the Dynamic Spectral-Spatial Interaction Convolution Neural Network (DSSICNN), a parameter-efficient MI-EEG decoding framework that jointly extracts temporal-spectral-spatial features. DSSICNN adopts a dual-branch parallel architecture to concurrently learn spatial representations in both Euclidean and non-Euclidean domains. It further integrates a CFC-inspired attention module to model cross-spectral interactions, followed by an additional attention mechanism that quantifies the contributions of distinct MI stages to decoding performance. DSSICNN achieves decoding performance on two public datasets that surpasses the current state-of-the-art (SOTA) under both session-dependent and session-independent settings. Beyond its empirical advantages, DSSICNN offers design insights for developing Graph Neural Network (GNN)-based MI-EEG decoding algorithms and provides a network neuroscience-inspired perspective for understanding the neurophysiological mechanisms underlying MI.

Speech brain-computer interfaces (BCIs) can restore speech features like articulatory movements from brain activity. However, for individuals with vocal tract paralysis, lack of articulatory movements can pose a challenge for speech BCI development. To address this challenge, our study aims at extracting generalizable articulatory features from a group of native Dutch speakers and reconstructing these features from brain data of a separate group of able-bodied individuals. We applied a tensor component analysis (TCA) model to extract generalisable articulatory features from a publicly available articulatory movement dataset. To reconstruct articulatory features from the brain, we analyzed data of three able-bodied participants P1, P2 and P3 with high-density electrocorticography (HD-ECoG) electrode arrays implanted over the sensorimotor cortex. For each participant, a separate TCA model was applied to extract neural features. A gradient boosting regression model was used to reconstruct articulatory features from neural features. Reconstruction performance was measured as Pearson's correlation coefficient (PCC) between the reconstructed and the generalizable articulatory features. The extracted articulatory features showed even contributions across participants, indicating that these features captured generalizable articulatory kinematic patterns. By using these features, we were able to reconstruct articulatory features from brain data. PCC between the reconstructed and original articulatory features were significantly above chance for all three participants, with mean PCCs of 0.80, 0.75 and 0.76 for P1, P2 and P3 respectively. With the rapid development of speech BCI, our research demonstrates that speech-related articulatory features can be restored from HD-ECoG signal using generalizable articulatory features derived from able-bodied individuals. With the potential to reconstruct audio or speech-related facial movements from the reconstructed articulatory features, our framework may provide a new way for developing speech BCIs for people unable to produce mouth movements.

Spontaneous intracerebral hemorrhage (ICH) is associated with substantial mortality and morbidity. Current management paradigms rely heavily on the rapid interpretation of neuroimaging and clinical data, yet are frequently constrained by limitations in processing speed, diagnostic accuracy, and prognostic precision. Artificial intelligence (AI), specifically machine learning (ML) and deep learning (DL), offers transformative potential to circumvent these challenges across the entire continuum of ICH care. This comprehensive review synthesizes the rapidly evolving landscape of AI applications in ICH management. Through a systematic evaluation of recent literature, we examine studies focused on the development, validation, or critical appraisal of AI-driven technologies for ICH care. Our analysis encompasses automated neuroimaging, computer-assisted surgical navigation, brain-computer interfaces (BCIs), prognostic modeling, and fundamental research into disease mechanisms. AI has demonstrated performance comparable to that of clinical experts in automating hematoma segmentation, predicting complications such as hematoma expansion, and refining surgical planning via augmented reality. Furthermore, BCIs present innovative therapeutic avenues for motor rehabilitation. However, the translation of these technological advances into routine clinical practice is impeded by substantial challenges, including data heterogeneity, model opacity ("black-box" issues), workflow integration barriers, regulatory ambiguities, and ethical concerns surrounding accountability and algorithmic bias. The integration of AI into ICH care signifies a paradigm shift from standardized treatment protocols toward dynamic, precision medicine. Realizing this vision necessitates interdisciplinary collaboration to engineer robust, generalizable, and interpretable AI systems. Key priorities include the establishment of large-scale multimodal data repositories, the advancement of explainable AI (XAI) frameworks, the execution of rigorous prospective clinical trials to validate efficacy, and the implementation of adaptive regulatory and ethical guidelines. By systematically addressing these barriers, AI can evolve from a mere analytical tool into an indispensable clinical partner, ultimately optimizing patient outcomes.

Brain-computer interface (BCI) control inefficiency often occurs in stroke survivors due to insufficient sensorimotor activity generated during motor imagery. Previous studies focused on upregulating excitability of primary motor cortex (M1) alone. Dorsolateral prefrontal cortex (DLPFC), an important region for motor imagery, may be effective for improving BCI performance. This study is aimed at investigating how intermittent theta burst stimulation (iTBS) targeted on M1 and DLPFC influences BCI performance and its neural mechanisms.25 healthy subjects (9 males) received four types of iTBS (i.e., M1 iTBS, DLPFC iTBS, combination of M1 and DLPFC iTBS and sham iTBS) on separate days. BCI control testing, functional near-infrared spectroscopy assessment and single-pulse transcranial magnetic stimulation were performed before and immediately after iTBS in each session. Corticospinal excitability, brain activation, and functional connectivity were calculated. Our results revealed that corticospinal excitability was significantly increased after M1 iTBS (P = 0.016), with the magnitude of increase positively correlated with BCI performance (P = 0.013). Frontoparietal network functional connectivity was significantly increased after DLPFC iTBS (P's<0.05), with the magnitude of increase positively correlated with changes in BCI performance (P's<0.05). In conclusion, M1 iTBS and DLPFC iTBS alone influences BCI performance through specific neural mechanisms, and the combination of M1 and DLPFC iTBS did not induce any significant results. M1 iTBS could influence BCI performance by enhancing corticospinal excitability, while DLPFC iTBS could influence BCI performance by increasing frontoparietal network connectivity. These findings could contribute to the advancement of novel therapeutic strategies aimed at enhancing BCI effectiveness for neurological populations. Trial registration: The study was retrospectively registered in the Chinese Clinical Trial Registry (ChiCTR2500097678). Registration Date: 2025-02-24.