====== Brain-Computer Interface (BCI) ======
A **Brain-Computer Interface (BCI)** is a technology system that establishes direct communication pathways between the brain and external computing devices by detecting, processing, and interpreting neural signals. BCIs translate patterns of electrical activity across the brain into actionable commands, enabling users to control external systems without traditional neuromuscular pathways. This technology bridges neuroscience, biomedical engineering, and computer science to create novel interaction modalities for human-computer communication (([[https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3691921/|Wolpaw et al. - Brain-Computer Interfaces for Communication and Control (2002]])).

===== Technical Foundations =====
BCIs operate through several interconnected components working in concert. **Signal acquisition** captures neural activity using various recording methods, including electroencephalography (EEG) for non-invasive surface recordings, electrocorticography (ECoG) for recordings from the cortical surface, or single-unit microelectrode arrays for direct neuronal recordings. The acquired signals contain inherent noise and artifacts that require sophisticated **signal processing** techniques including filtering, amplification, and artifact rejection to isolate meaningful neural patterns.

The **feature extraction** phase identifies distinctive patterns within processed signals that correlate with specific neural states or intentions. Machine learning algorithms analyze these features to build **classification models** that map neural patterns to desired outputs or commands. Modern BCIs increasingly employ deep learning architectures to identify complex, high-dimensional patterns within neural data (([[https://arxiv.org/abs/1611.02174|Bashivan et al. - Learning Representations from EEG with Deep Recurrent-Convolutional Neural Networks (2016]])).

The **feedback loop** provides essential sensory information to users about system performance, enabling calibration and adaptation. This closed-loop interaction allows users to refine their neural control strategies through practice, leveraging neural plasticity to improve performance over time. The latency between neural signal and system response critically affects usability—lower latency systems enable more natural, intuitive control (([[https://www.nature.com/articles/nature12373|Hochberg et al. - Reach and Grasp by People with Tetraplegia Using a Neurally Controlled Robotic Arm (2012]])).

===== Recording Modalities and Biosensing Approaches =====
Different BCI implementations employ distinct recording technologies with varying trade-offs between invasiveness, signal quality, spatial resolution, and practical applicability. **Electroencephalography (EEG)** remains the most accessible BCI modality, utilizing non-invasive scalp electrodes to measure aggregate electrical potentials. While EEG offers portability and safety advantages, its spatial resolution remains limited due to signal dispersion through skull and tissue.

**Functional Near-Infrared Spectroscopy (fNIRS)** measures changes in blood oxygenation across cortical regions, providing better spatial localization than EEG while maintaining non-invasive deployment. **Electrocorticography (ECoG)** and **microelectrode arrays** provide superior signal quality and spatial resolution by positioning recording electrodes closer to neural tissue, but require surgical implantation. Recent advances in **flexible electronics** and **biocompatible materials** continue improving the long-term stability and performance of implanted devices (([[https://arxiv.org/abs/1905.06115|Musk and Neuralink - An Integrated Brain-Machine Interface Platform with Thousands of Channels (2019]])).

===== Clinical Applications and Current Implementations =====
BCIs demonstrate substantial clinical utility for individuals with severe motor impairments from spinal cord injury, amyotrophic lateral sclerosis (ALS), or brainstem stroke. **Motor BCIs** enable paralyzed users to control robotic arms or computer cursors with neural signals, restoring communication and manipulation capabilities. Documented case studies show users achieving real-time control of multi-jointed robotic limbs with natural, coordinated movement patterns.

**Communication BCIs** decode intended speech from neural activity in motor and premotor cortex, enabling severely paralyzed individuals to compose text or speech at clinically meaningful rates. Recent implementations have achieved character communication rates approaching natural conversation speeds. **Sensory BCIs** deliver tactile or visual information through direct brain stimulation, restoring rudimentary sensory perception to users with sensory loss.

**Neurofeedback BCIs** employ real-time brain activity visualization to help users self-regulate neural states for therapeutic applications including stroke rehabilitation, attention disorders, and pain management. These systems typically provide visual or auditory feedback correlated with target neural patterns, leveraging operant conditioning principles for therapeutic purposes (([[https://www.thelancet.com/journals/laneur/article/PIIS1474-4422(20)30322-3/fulltext|Musk - An Integrated Brain-Machine Interface Platform (2020]])).

===== Challenges and Limitations =====
Current BCI systems face substantial technical and practical constraints limiting widespread deployment. **Non-stationarity** of neural signals—gradual shifts in signal properties over hours or days—necessitates frequent recalibration and substantially impairs long-term usability. **Signal degradation** from electrode drift, inflammation, or material breakdown degrades performance in chronically implanted systems.

The **decoding complexity** and **individual variability** across users require extensive personalized calibration, limiting rapid adoption. **Cognitive load** of operating BCIs remains substantial, with users requiring sustained attention and mental effort for extended operation. **Limited bandwidth** relative to conventional input methods constrains information transfer rates, typically ranging from 5-50 bits per minute compared to hundreds of bits per minute for natural typing or speech.

===== Future Directions =====
Emerging research addresses fundamental limitations through improved **neural decoding algorithms**, **adaptive learning systems**, and **implant biocompatibility**. Multi-[[modal|modal]] approaches combining BCIs with other sensing modalities show promise for robust, high-performance systems. **Wireless power delivery** and **miniaturized electronics** continue enabling less invasive, more practical implant designs for therapeutic and consumer applications.

===== See Also =====
  * [[system_level_ai|System-Level AI / Ambient Intelligence]]
  * [[bio_inspired_hardware|Bio-Inspired Hardware]]
  * [[galaxy_brain|Galaxy Brain]]

===== References =====