Brain-computer interface (BCI) technology utilizing electroencephalography (EEG) marks a transformative innovation, empowering motor-impaired individuals to engage with their environment on equal footing. Despite its promising potential, developing subject-invariant and session-invariant BCI systems remains a significant challenge due to the inherent complexity and variability of neural activity across individuals and over time, compounded by EEG hardware constraints. While prior studies have sought to develop robust BCI systems, existing approaches remain ineffective in capturing the intricate spatiotemporal dependencies within multichannel EEG signals. This study addresses this gap by introducing the attentive graph-temporal convolutional network (AGTCNet), a novel graph-temporal model for motor imagery EEG (MI-EEG) classification. Specifically, AGTCNet leverages the topographic configuration of EEG electrodes as an inductive bias and integrates graph convolutional attention network (GCAT) to jointly learn expressive spatiotemporal EEG representations. The proposed model significantly outperformed existing MI-EEG classifiers, achieving state-of-the-art performance while utilizing a compact architecture, underscoring its effectiveness and practicality for BCI deployment. With a 49.87% reduction in model size, 64.65% faster inference time, and shorter input EEG signal, AGTCNet achieved a moving average accuracy of 66.82% for subject-independent classification on the BCI Competition IV Dataset 2a, which further improved to 82.88% when fine-tuned for subject-specific classification. On the EEG Motor Movement/Imagery Dataset, AGTCNet achieved moving average accuracies of 64.14% and 85.22% for 4-class and 2-class subject-independent classifications, respectively, with further improvements to 72.13% and 90.54% for subject-specific classifications.
Brain-computer interface enables direct interaction between brain and device. However, common brain-computer interfaces often employ one-step strategy that rely on non-natural paradigms, such as SSVEP-BCI and MI-BCI, are limited to specific scenarios, restricting their broader application. This paper first proposes a two-step strategic brain-muscular-hand interface (BMHI) based on biological evolutionary selection mechanism, by integrating the brain-muscle (BM) interface with the muscle-hand (MH) interface through crosstalk ("BMHI = BM + MH"). To verify the effectiveness of BMHI and the advantages of a two-step strategy inspired by natural evolution, we conducted offline, comparison (comparing BMHI (two-step) and brain-hand interface (one-step)), and online experiments (using BMHI to control a virtual/machine hand for daily tasks). The results show that: (1) BMHI is feasible and the prediction accuracy is 0.79; (2) Unlike traditional multi-layer neural networks that attempt to establish a direct brain-signal-to-action mapping through a single end-to-end process (brain-hand interface), BMHI incorporates the neuro-muscular transmission mechanisms evolved in biological systems as an intermediate constraint layer. This phased decoding strategy can reduce training time by approximately 18-fold and improve decoding accuracy; (3) In the online control experiment, both the virtual hand and the manipulator were able to successfully complete tasks, like moving objects such as boxes or plates and holding water glasses. The results show that BMHI adopts a two-step decoding strategy that mimics natural human neural motor pathways, improves training efficiency and prediction accuracy, and promotes the development of BCI technology to a more natural interaction mode.
Individuals with severe physical disabilities often experience diminished quality of life stemming from limited ability to engage with their surroundings. Brain-Computer Interface (BCI) technology aims to bridge this gap by enabling direct technology interaction. However, current BCI systems require invasive procedures, such as craniotomy or implantation of electrodes through blood vessels, posing significant risks to patients. Sub-scalp electroencephalography (EEG) offers a lower risk alternative. This study investigates the signal quality of sub-scalp EEG recordings from various depths in a sheep model, and compares results with other methods: ECoG and endovascular arrays. A computational model was also constructed to investigate the factors underlying variations in electrode performance. We demonstrate that peg electrodes placed within the sub-scalp space can achieve visual evoked potential signal-to-noise ratios (SNRs) approaching that of ECoG. Endovascular arrays exhibited SNR comparable to electrodes positioned on the periosteum. Furthermore, sub-scalp recordings captured high gamma neural activity, with maximum bandwidth ranging from 120 Hz to 180 Hz depending on electrode depth. These findings support the use of sub-scalp EEG for BCI applications, and provide valuable insights for future sub-scalp electrode design. This data lays the groundwork for human trials, ultimately paving the way for chronic, in-home BCIs that empower individuals with physical disabilities.
Objective: To establish sub-scalp electroencephalography (EEG) as a viable option for brain-computer interface (BCI) applications, particularly for chronic use, by demonstrating its effectiveness in recording and classifying sensorimotor neural activity. Approach: Two experiments were conducted in this study. The first aim was to demonstrate the high spatial resolution of sub-scalp EEG through analysis of somatosensory evoked potentials in sheep models. The second focused on the practical application of sub-scalp EEG, classifying motor execution using data collected during a sheep behavioural experiment. Main Results: We successfully demonstrated the recording of sensorimotor rhythms using sub-scalp EEG in sheep models. Important spatial, temporal, and spectral features of these signals were identified, and we were able to classify motor execution with above-chance performance. These results are comparable to previous work that investigated signal quality and motor execution classification using ECoG and endovascular arrays in sheep models. Significance: These results suggest that sub-scalp EEG may provide signal quality that approaches that of more invasive neural recording methods such as ECoG and endovascular arrays, and support the use of sub-scalp EEG for chronic BCI applications.
Brain computer interface (BCI) research, as well as increasing portions of the field of neuroscience, have found success deploying large-scale artificial intelligence (AI) pre-training methods in conjunction with vast public repositories of data. This approach of pre-training foundation models using label-free, self-supervised objectives offers the potential to learn robust representations of neurophysiology, potentially addressing longstanding challenges in neural decoding. However, to date, much of this work has focused explicitly on standard BCI benchmarks and tasks, which likely overlooks the multitude of features these powerful methods might learn about brain function as well as other electrophysiological information. We introduce a new method for self-supervised BCI foundation model pre-training for EEG inspired by a transformer-based approach adapted from the HuBERT framework originally developed for speech processing. Our pipeline is specifically focused on low-profile, real-time usage, involving minimally pre-processed data and just eight EEG channels on the scalp. We show that our foundation model learned a representation of EEG that supports standard BCI tasks (P300, motor imagery), but also that this model learns features of neural data related to individual variability, and other salient electrophysiological components (e.g., alpha rhythms). In addition to describing and evaluating a novel approach to pre-training BCI models and neural decoding, this work opens the aperture for what kind of tasks and use-cases might exist for neural data in concert with powerful AI methods.
The advance of speech decoding from non-invasive brain data holds the potential for profound societal impact. Among its most promising applications is the restoration of communication to paralysed individuals affected by speech deficits such as dysarthria, without the need for high-risk surgical interventions. The ultimate aim of the 2025 PNPL competition is to produce the conditions for an "ImageNet moment" or breakthrough in non-invasive neural decoding, by harnessing the collective power of the machine learning community. To facilitate this vision we present the largest within-subject MEG dataset recorded to date (LibriBrain) together with a user-friendly Python library (pnpl) for easy data access and integration with deep learning frameworks. For the competition we define two foundational tasks (i.e. Speech Detection and Phoneme Classification from brain data), complete with standardised data splits and evaluation metrics, illustrative benchmark models, online tutorial code, a community discussion board, and public leaderboard for submissions. To promote accessibility and participation the competition features a Standard track that emphasises algorithmic innovation, as well as an Extended track that is expected to reward larger-scale computing, accelerating progress toward a non-invasive brain-computer interface for speech.
In Brain-Computer Interface (BCI) research, the detailed study of blinks is crucial. They can be considered as noise, affecting the efficiency and accuracy of decoding users' cognitive states and intentions, or as potential features, providing valuable insights into users' behavior and interaction patterns. We introduce a large dataset capturing electroencephalogram (EEG) signals, eye-tracking, high-speed camera recordings, as well as subjects' mental states and characteristics, to provide a multifactor analysis of eye-related movements. Four paradigms -- motor imagery, motor execution, steady-state visually evoked potentials, and P300 spellers -- are selected due to their capacity to evoke various sensory-motor responses and potential influence on ocular activity. This online-available dataset contains over 46 hours of data from 31 subjects across 63 sessions, totaling 2520 trials for each of the first three paradigms, and 5670 for P300. This multimodal and multi-paradigms dataset is expected to allow the development of algorithms capable of efficiently handling eye-induced artifacts and enhancing task-specific classification. Furthermore, it offers the opportunity to evaluate the cross-paradigm robustness involving the same participants.
An Oculomotor Brain-Computer Interface (BCI) records neural activity from regions of the brain involved in planning eye movements and translates this activity into control commands. While previous successful oculomotor BCI studies primarily relied on invasive microelectrode implants in non-human primates, this study investigates the feasibility of an oculomotor BCI using a minimally invasive endovascular Stentrode device implanted near the supplementary motor area in a patient with amyotrophic lateral sclerosis (ALS). To achieve this, self-paced visually-guided and free-viewing saccade tasks were designed, in which the participant performed saccades in four directions (left, right, up, down), with simultaneous recording of endovascular EEG and eye gaze. The visually guided saccades were cued with visual stimuli, whereas the free-viewing saccades were self-directed without explicit cues. The results showed that while the neural responses of visually guided saccades overlapped with the cue-evoked potentials, the free-viewing saccades exhibited distinct saccade-related potentials that began shortly before eye movement, peaked approximately 50 ms after saccade onset, and persisted for around 200 ms. In the frequency domain, these responses appeared as a low-frequency synchronisation below 15 Hz. Classification of 'fixation vs. saccade' was robust, achieving mean area under the receiver operating characteristic curve (AUC) scores of 0.88 within sessions and 0.86 between sessions. In contrast, classifying saccade direction proved more challenging, yielding within-session AUC scores of 0.67 for four-class decoding and up to 0.75 for the best-performing binary comparisons (left vs. up and left vs. down). This proof-of-concept study demonstrates the feasibility of an endovascular oculomotor BCI in an ALS patient, establishing a foundation for future oculomotor BCI studies in human subjects.
In recent years, deep learning (DL) approaches have demonstrated promising results in decoding hemodynamic responses captured by functional near-infrared spectroscopy (fNIRS), particularly in the context of brain-computer interface (BCI) applications. This work introduces AbsoluteNet, a novel deep learning architecture designed to classify auditory event-related responses recorded using fNIRS. The proposed network is built upon principles of spatio-temporal convolution and customized activation functions. Our model was compared against several models, namely fNIRSNET, MDNN, DeepConvNet, and ShallowConvNet. The results showed that AbsoluteNet outperforms existing models, reaching 87.0% accuracy, 84.8% sensitivity, and 89.2% specificity in binary classification, surpassing fNIRSNET, the second-best model, by 3.8% in accuracy. These findings underscore the effectiveness of our proposed deep learning model in decoding hemodynamic responses related to auditory processing and highlight the importance of spatio-temporal feature aggregation and customized activation functions to better fit fNIRS dynamics.
Recent advancements in Large Language Models have inspired the development of foundation models across various domains. In this study, we evaluate the efficacy of Large EEG Models (LEMs) by fine-tuning LaBraM, a state-of-the-art foundation EEG model, on a real-world stress classification dataset collected in a graduate classroom. Unlike previous studies that primarily evaluate LEMs using data from controlled clinical settings, our work assesses their applicability to real-world environments. We train a binary classifier that distinguishes between normal and elevated stress states using resting-state EEG data recorded from 18 graduate students during a class session. The best-performing fine-tuned model achieves a balanced accuracy of 90.47% with a 5-second window, significantly outperforming traditional stress classifiers in both accuracy and inference efficiency. We further evaluate the robustness of the fine-tuned LEM under random data shuffling and reduced channel counts. These results demonstrate the capability of LEMs to effectively process real-world EEG data and highlight their potential to revolutionize brain-computer interface applications by shifting the focus from model-centric to data-centric design.
Decoding speech directly from neural activity is a central goal in brain-computer interface (BCI) research. In recent years, exciting advances have been made through the growing use of intracranial field potential recordings, such as stereo-ElectroEncephaloGraphy (sEEG) and ElectroCorticoGraphy (ECoG). These neural signals capture rich population-level activity but present key challenges: (i) task-relevant neural signals are sparsely distributed across sEEG electrodes, and (ii) they are often entangled with task-irrelevant neural signals in both sEEG and ECoG. To address these challenges, we introduce a unified Coarse-to-Fine neural disentanglement framework, BrainStratify, which includes (i) identifying functional groups through spatial-context-guided temporal-spatial modeling, and (ii) disentangling distinct neural dynamics within the target functional group using Decoupled Product Quantization (DPQ). We evaluate BrainStratify on two open-source sEEG datasets and one (epidural) ECoG dataset, spanning tasks like vocal production and speech perception. Extensive experiments show that BrainStratify, as a unified framework for decoding speech from intracranial neural signals, significantly outperforms previous decoding methods. Overall, by combining data-driven stratification with neuroscience-inspired modularity, BrainStratify offers a robust and interpretable solution for speech decoding from intracranial recordings.
Human preference research is a significant domain in psychology and psychophysiology, with broad applications in psychiatric evaluation and daily life quality enhancement. This study explores the neural mechanisms of human preference judgments through the analysis of event-related potentials (ERPs), specifically focusing on the early N1 component and the late positive potential (LPP). Using a mixed-image dataset covering items such as hats, fruits, snacks, scarves, drinks, and pets, we elicited a range of emotional responses from participants while recording their brain activity via EEG. Our work innovatively combines the N1 and LPP components to reveal distinct patterns across different preference levels. The N1 component, particularly in frontal regions, showed increased amplitude for preferred items, indicating heightened early visual attention. Similarly, the LPP component exhibited larger amplitudes for both preferred and non-preferred items, reflecting deeper emotional engagement and cognitive evaluation. In addition, we introduced a relationship model that integrates these ERP components to assess the intensity and direction of preferences, providing a novel method for interpreting EEG data in the context of emotional responses. These findings offer valuable insights into the cognitive and emotional processes underlying human preferences and present new possibilities for brain-computer interface applications, personalized marketing, and product design.
A brain-computer interface (BCI) system enables direct communication between the brain and external devices, offering significant potential for assistive technologies and advanced human-computer interaction. Despite progress, BCI systems face persistent challenges, including signal variability, classification inefficiency, and difficulty adapting to individual users in real time. In this study, we propose a novel hybrid quantum learning model, termed QSVM-QNN, which integrates a Quantum Support Vector Machine (QSVM) with a Quantum Neural Network (QNN), to improve classification accuracy and robustness in EEG-based BCI tasks. Unlike existing models, QSVM-QNN combines the decision boundary capabilities of QSVM with the expressive learning power of QNN, leading to superior generalization performance. The proposed model is evaluated on two benchmark EEG datasets, achieving high accuracies of 0.990 and 0.950, outperforming both classical and standalone quantum models. To demonstrate real-world viability, we further validated the robustness of QNN, QSVM, and QSVM-QNN against six realistic quantum noise models, including bit flip and phase damping. These experiments reveal that QSVM-QNN maintains stable performance under noisy conditions, establishing its applicability for deployment in practical, noisy quantum environments. Beyond BCI, the proposed hybrid quantum architecture is generalizable to other biomedical and time-series classification tasks, offering a scalable and noise-resilient solution for next-generation neurotechnological systems.
Despite major advances in surgical brain-to-text (B2T), i.e. transcribing speech from invasive brain recordings, non-invasive alternatives have yet to surpass even chance on standard metrics. This remains a barrier to building a non-invasive brain-computer interface (BCI) capable of restoring communication in paralysed individuals without surgery. Here, we present the first non-invasive B2T result that significantly exceeds these critical baselines, raising BLEU by $1.4\mathrm{-}2.6\times$ over prior work. This result is driven by three contributions: (1) we extend recent word-classification models with LLM-based rescoring, transforming single-word predictors into closed-vocabulary B2T systems; (2) we introduce a predictive in-filling approach to handle out-of-vocabulary (OOV) words, substantially expanding the effective vocabulary; and (3) we demonstrate, for the first time, how to scale non-invasive B2T models across datasets, unlocking deep learning at scale and improving accuracy by $2.1\mathrm{-}2.3\times$. Through these contributions, we offer new insights into the roles of data quality and vocabulary size. Together, our results remove a major obstacle to realising practical non-invasive B2T systems.
Graph neural networks have re-defined how we model and predict on network data but there lacks a consensus on choosing the correct underlying graph structure on which to model signals. CoVariance Neural Networks (VNN) address this issue by using the sample covariance matrix as a Graph Shift Operator (GSO). Here, we improve on the performance of VNNs by constructing a Density Matrix where we consider the sample Covariance matrix as a quasi-Hamiltonian of the system in the space of random variables. Crucially, using this density matrix as the GSO allows components of the data to be extracted at different scales, allowing enhanced discriminability and performance. We show that this approach allows explicit control of the stability-discriminability trade-off of the network, provides enhanced robustness to noise compared to VNNs, and outperforms them in useful real-life applications where the underlying covariance matrix is informative. In particular, we show that our model can achieve strong performance in subject-independent Brain Computer Interface EEG motor imagery classification, outperforming EEGnet while being faster. This shows how covariance density neural networks provide a basis for the notoriously difficult task of transferability of BCIs when evaluated on unseen individuals.
Emotion recognition is essential across numerous fields, including medical applications and brain-computer interface (BCI). Emotional responses include behavioral reactions, such as tone of voice and body movement, and changes in physiological signals, such as the electroencephalogram (EEG). The latter are involuntary, thus they provide a reliable input for identifying emotions, in contrast to the former which individuals can consciously control. These signals reveal true emotional states without intentional alteration, thus increasing the accuracy of emotion recognition models. However, multimodal deep learning methods from physiological signals have not been significantly investigated. In this paper, we introduce PHemoNet, a fully hypercomplex network for multimodal emotion recognition from physiological signals. In detail, the architecture comprises modality-specific encoders and a fusion module. Both encoders and fusion modules are defined in the hypercomplex domain through parameterized hypercomplex multiplications (PHMs) that can capture latent relations between the different dimensions of each modality and between the modalities themselves. The proposed method outperforms current state-of-the-art models on the MAHNOB-HCI dataset in classifying valence and arousal using electroencephalograms (EEGs) and peripheral physiological signals. The code for this work is available at https://github.com/ispamm/MHyEEG.
Invasive brain-computer interface (BCI) technology has demonstrated the possibility of restoring brain-controlled walking in paraplegic spinal cord injury patients. However, current implementations of BCI-controlled walking still have significant drawbacks. In particular, prior systems are unidirectional and lack sensory feedback for insensate patients, have suboptimal reliance on brain signals from the bilateral arm areas of the motor cortex, and depend on external systems for signal processing. Motivated by these shortcomings, this study is the first time a bidirectional brain-computer interface (BDBCI) has demonstrated the restoration of both brain-controlled walking and leg sensory feedback while utilizing the bilateral leg motor and sensory cortices. Here, a subject undergoing subdural electrocorticogram electrode implantation for epilepsy surgery evaluation leveraged the leg representation areas of the bilateral interhemispheric primary motor and sensory cortices to operate a BDBCI with high performance. Although electrode implantation in the interhemispheric region is uncommon, electrodes can be safely implanted in this region to access rich leg motor information and deliver bilateral leg sensory feedback. Finally, we demonstrated that all BDBCI operations can be executed on a dedicated, portable embedded system. These results indicate that BDBCIs can potentially provide brain-controlled ambulation and artificial leg sensation to people with paraplegia after spinal cord injury in a manner that emulates full-implantability and is untethered from any external systems.
This paper explores silent speech decoding in active brain-computer interface (BCI) systems, which offer more natural and flexible communication than traditional BCI applications. We collected a new silent speech dataset of over 120 hours of electroencephalogram (EEG) recordings from 12 subjects, capturing 24 commonly used English words for language model pretraining and decoding. Following the recent success of pretraining large models with self-supervised paradigms to enhance EEG classification performance, we propose Large Brain Language Model (LBLM) pretrained to decode silent speech for active BCI. To pretrain LBLM, we propose Future Spectro-Temporal Prediction (FSTP) pretraining paradigm to learn effective representations from unlabeled EEG data. Unlike existing EEG pretraining methods that mainly follow a masked-reconstruction paradigm, our proposed FSTP method employs autoregressive modeling in temporal and frequency domains to capture both temporal and spectral dependencies from EEG signals. After pretraining, we finetune our LBLM on downstream tasks, including word-level and semantic-level classification. Extensive experiments demonstrate significant performance gains of the LBLM over fully-supervised and pretrained baseline models. For instance, in the difficult cross-session setting, our model achieves 47.0\% accuracy on semantic-level classification and 39.6\% in word-level classification, outperforming baseline methods by 5.4\% and 7.3\%, respectively. Our research advances silent speech decoding in active BCI systems, offering an innovative solution for EEG language model pretraining and a new dataset for fundamental research.
Current brain-computer interfaces (BCI) face limitations in signal acquisition. While sub-scalp EEG offers a potential solution, existing devices prioritize chronic seizure monitoring and lack features suited for BCI applications. This work addresses this gap by outlining key specifications for sub-scalp BCI devices, focusing on channel count, sampling rate, power efficiency, and form factor. We present the Set-And-Forget EEG (SAFE) system, a custom-built amplifier and wireless transmitter meeting these criteria. This compact (12x12 mm), six-channel device offers 1024 Hz sampling and Bluetooth Low Energy data transmission. Validation using generated sinusoids and electrocorticography recordings of visual evoked potentials in sheep models demonstrated low noise recording. Future animal studies will assess sub-scalp EEG signal quality for BCI applications. This data lays the groundwork for human trials, ultimately paving the way for chronic, in-home BCIs that empower individuals with physical disabilities.
Robots with wheeled, quadrupedal, or humanoid forms are increasingly integrated into built environments. However, unlike human social learning, they lack a critical pathway for intrinsic cognitive development, namely, learning from human feedback during interaction. To understand human ubiquitous observation, supervision, and shared control in dynamic and uncertain environments, this study presents a brain-computer interface (BCI) framework that enables classification of Electroencephalogram (EEG) signals to detect cognitively demanding and safety-critical events. As a timely and motivating co-robotic engineering application, we simulate a human-in-the-loop scenario to flag risky events in semi-autonomous robotic driving-representative of long-tail cases that pose persistent bottlenecks to the safety performance of smart mobility systems and robotic vehicles. Drawing on recent advances in few-shot learning, we propose a dual-attention Siamese convolutional network paired with Dynamic Time Warping Barycenter Averaging approach to generate robust EEG-encoded signal representations. Inverse source localization reveals activation in Broadman areas 4 and 9, indicating perception-action coupling during task-relevant mental imagery. The model achieves 80% classification accuracy under data-scarce conditions and exhibits a nearly 100% increase in the utility of salient features compared to state-of-the-art methods, as measured through integrated gradient attribution. Beyond performance, this study contributes to our understanding of the cognitive architecture required for BCI agents-particularly the role of attention and memory mechanisms-in categorizing diverse mental states and supporting both inter- and intra-subject adaptation. Overall, this research advances the development of cognitive robotics and socially guided learning for service robots in complex built environments.