Editor’s Note: Originally published as a retrospective in the journal Device.
Non-invasive brain-computer interfaces (BCIs) are at the forefront of neurotechnology, enabling a direct link between the brain and external devices. BCIs hold great potential for clinical and consumer applications, ranging from cognitive enhancement and entertainment to assistive and rehabilitative devices for motor impairments. While invasive BCIs, such as those implanting intracortical electrodes, offer higher signal fidelity, non-invasive BCIs are better suited for widespread use due to their safety, ease of use, and cost-effectiveness. Recent advancements in material design, device miniaturization, and algorithmic tools have significantly improved the performance of non-invasive BCIs, making them viable for real-world applications. This commentary explores the state of the art in non-invasive BCIs, focusing on use cases in clinical rehabilitation and the advancements in related supporting technologies. It also sheds light on emerging trends and future directions for BCIs, highlighting their potential adoption as an accessible consumer technology for both clinical and non-clinical settings.
State of the Art
Technological Overview
Non-invasive BCIs primarily use electroencephalography (EEG), functional near-infrared spectroscopy (fNIRS), functional magnetic resonance imaging (fMRI), and magnetoencephalography (MEG) to record brain activity. Among all, EEG is the most widely used due to its portability, affordability, and ability to capture brain activity in real time. However, EEG signals are often noisy, unstable, and have low spatial resolution, making it challenging to decode complex neural activity reliably. Conversely, fMRI and MEG provide higher spatial resolution but are less practical for everyday use due to their high cost and bulkiness. fNIRS, which measures hemodynamic responses in the brain, provides another lens to study brain activity by recording blood flow changes in response to neural activity. However, its slow temporal response limits its usability in real-time applications.1 Given the potential of BCIs, ongoing research is mitigating the challenges of EEG-BCIs with material designs that support stable and high-fidelity EEG acquisition and with machine learning (ML) tools to enhance the performance of BCI decoders for EEG classification.
Applications in Neurorehabilitation
EEG-BCIs have found an important niche in neurorehabilitation, where they can help patients recover motor and cognitive functions. In motor rehabilitation, after insults to the central nervous system, such as stroke or spinal cord injury (SCI), BCIs uncover residual neural activity that is responsible for motion intents. Even if paralyzed, patients can volitionally activate their sensorimotor cortex by attempting movements, and BCIs can detect such activation to drive external devices like neuroprosthetics or functional electrical stimulation (FES) and facilitate motor recovery. Such BCI neurorehabilitation interventions have demonstrated significant and lasting neurophysiological improvements in stroke patients.
In cognitive rehabilitation, BCIs are being used to treat conditions like attention-deficit/hyperactivity disorder (ADHD) through neurofeedback.4
Mutual Learning Paradigms
The effectiveness of a BCI system relies heavily on the ability to decode users’ intentions accurately. This hinges on two active agents—the human and the machine—who work collaboratively to achieve reliable control. With extended use and suitable feedback provided by accurate decoders, the user can learn to modulate brain patterns that are easier to decode. On the other hand, as new data come from the user, ML tools can adaptively personalize the decoding models to the emerging modulations generated by the user. Recent work has explored transferring BCI decoders from experts at controlling BCIs to naive users, which allows the latter to readily use a BCI while still learning personalized control strategies accounted for through incremental adaptation of model parameters. Such transfer learning methods not only allow transferability across subjects but also across days and similar tasks.
Trends
Wearable BCIs
Wearable BCIs are a growing trend. Advances in soft and stretchable electronics are being explored to improve the adaptability and ergonomics of BCIs. Additionally, wireless data transmission is increasingly important in providing a better BCI user experience for a wider range of user scenarios, opening up opportunities for the next phase of the flourishing of brain-computer interaction. Notably, recent progress in material science has substantially enhanced electrode technology, with considerable efforts directed toward enabling long-term stability and high signal fidelity in wearable BCIs.
Closed-loop BCIs and Neuroplasticity
Closed-loop BCIs provide real-time feedback based on the user’s brain activity, which is critical in rehabilitation. These systems leverage neuroplasticity to help patients recover motor and cognitive functions. The use of BCIs induces activity-dependent plasticity because BCI-directed volitional modulation of activity in a neural population triggers actions that activate another functionally connected population, thus facilitating recovery in rehabilitation and learning BCI control in general. In the future, closed-loop systems could also enhance cognitive enhancement, helping users improve attention, memory, or emotional regulation.
Challenges
Signal Noise and Artifact Removal
One of the biggest challenges facing non-invasive BCIs is the low signal-to-noise ratio of EEG signals. EEG is highly susceptible to interference from muscle movements, eye blinks, and environmental noise, which can significantly reduce the accuracy of BCI systems. Although modern signal processing techniques and advancements in hardware have improved the ability to filter out noise, the constant high signal fidelity remains a key impediment to the widespread use of non-invasive devices.
Materials — Longevity and Wearability
Non-invasive BCIs often rely on electrodes making direct contact with the scalp. However, ensuring consistent, high-quality contact can be challenging, especially in dynamic scenarios. As a result, most high-performance EEG-BCIs use wet electrodes, which provide the best signal quality but need frequent reapplication to prevent signal degradation, making them unsuitable for long-term or daily use. On the other hand, dry electrodes offer more convenience but often result in poorer signal quality, increased susceptibility to noise, and lower wearing comfort. Furthermore, the wearability of current BCIs in the consumer market remains a significant issue, particularly those designed for daily use outside of clinical environments. Many EEG-based BCIs rely on rigid electrode caps or uncomfortable headsets that restrict movement and lead to discomfort over extended periods.
Scalability — User Training Barriers
Another challenge is scalability, particularly in terms of user training. Many BCI systems require users to undergo extensive training before they can effectively control devices. Furthermore, individual differences in brain activity can make it difficult to develop systems that work well for a wide range of users. How to facilitate the fact that subjects can rapidly acquire BCI control is a critical issue.
Future Directions
Advancements in AI and ML
Artificial intelligence (AI) is poised to play a transformative role in the evolution of non-invasive BCIs. ML algorithms have already demonstrated their potential to improve the accuracy of decoding brain signals, making BCIs more responsive and user friendly. As ML advances, BCIs could become more personalized, adapting to individual users’ unique neural patterns and requiring less training time. This will be particularly crucial in making BCIs more accessible for both clinical and consumer applications.
Wearable BCI Devices: Enhancing Accessibility and Daily Integration
Future BCIs should be lightweight, portable, transparent, and user friendly, facilitating their daily use outside clinical settings. Wearable BCIs could become a staple for assisting in activities of daily living and improving cognitive health as users engage in neurofeedback or mindfulness exercises on the go. Moreover, with the advancements in material science, these devices will likely become more comfortable and less obtrusive, resembling everyday accessories like headbands or caps. Such innovations would make BCIs a practical neurorehabilitation tool in clinics or at home as well as continuous cognitive monitoring.
Exploring AR/VR Integration
A key emerging trend in the future of BCIs is the integration of virtual reality (VR) and augmented reality (AR) into wearable BCI systems, especially in neurorehabilitation and everyday applications. VR- and AR-enhanced BCIs create immersive environments where users can interact with virtual objects or navigate digital spaces using their brain signals. For example, a stroke patient could use a VR-BCI system to imagine grasping virtual objects, with the VR environment providing visual and haptic feedback. Such an approach may accelerate neuroplasticity by making the rehabilitation process more interactive and motivating, leading to a higher willingness to engage in the stroke rehabilitation process, thus leading to better long-term outcomes.
Driving Neuroplasticity with Neuromodulation
A major bottleneck in BCIs is having humans within the interaction loop. As is typical of learning new skills, subjects often require a considerable period of training to achieve reliable BCI control. This is due to the fact that activity-dependent neuroplasticity underlying intuitive BCI control happens on relatively long time scales throughout the training phase. Therefore, there has been growing interest in using neuromodulation to induce targeted neuroplasticity that accelerates learning of BCI control skills. In this respect, the role of neuromodulation can be seen as 2-fold. First, it can enrich the feedback received by the user while operating the BCI to accelerate learning—an example is delivering FES on target muscles contingent on volitional activation of relevant motor brain areas as shown in BCI neurorehabilitation. Second, neuromodulation can condition the brain and prepare it for learning. Recent work has uncovered an overlooked role of inhibitory brain neuromodulation—before feedback training—in promoting faster skill learning by constraining neural dynamics to task-relevant brain areas.
Conclusion
Looking ahead, interdisciplinary research will be essential to overcome these challenges and unlock the full potential of BCIs. Ethical considerations will also be critical as the technology becomes more widely adopted, ensuring that BCIs are developed to benefit all users. As non-invasive BCIs continue to evolve, they can revolutionize healthcare, human-computer interaction, and cognitive enhancement, offering new opportunities for individuals with disabilities and the general population.
Written by Ju-Chun Hsieh, Hussein Alawieh, José del R. Millán, Huiliang “Evan” Wang
