What is BCI?

This blog was updated on March 26, 2026.

When considering the topic of speech, we tend to think of it as something the body does: air rises from the lungs, takes shape inside the larynx, and words leave our body through our lips, carried out into the world for all to hear.

Brain-computer interfaces are built upon a different premise: that communication is fundamentally a brain process, and with the right technology, the body’s neural pathways become less of a limitation. For individuals with impairments – where these pathways have become damaged – BCI technology can be utilized to help them regain lost functionality.

As a growing field with decades of clinical viability, there are now several different types of BCIs out there, each with distinctive qualities, features, and use cases. Let’s explore how they compare.

Disclaimer: While BCIs have the ability to write information into the brain as well as decode from it, this blog will focus solely on the recording and sensing functionality.

In this article:

• What is a brain-computer interface?
• Different types of BCIs
• Current and future BCI use case
• Potential challenges within the field of BCI
• What’s next for BCI technology?

A brain-computer interface is a system that reads, interprets, and acts upon recorded neural signals in order to establish a link between the brain and an external device.

Establishing this connection creates a channel for passing brain data from the biological to the technological realm, and vice versa. 

For someone living with paralysis, for instance, the interface might gather signals from electrodes near neurons in the motor cortex, translate them into intended actions or words, and use this data to help restore the ability to speak, type, or connect with others.

Components of a BCI

While BCI systems vary greatly in terms of design and capability, they share a similar structure: five distinct components that work together to turn intent into outcome.

1. Signal Acquisition: Sensors that record neural activity. The placement of these sensors, whether on the scalp, on the brain’s surface, or within the brain tissue itself, directly determines signal quality – and what the BCI is actually capable of doing or restoring for the user.
2. Preprocessing: Raw neural signals are amplified, filtered to remove electrical noise and artifacts (interference from muscle movement, environmental electronics, etc.), and digitized – an important step, as signal cleanliness directly relates to reliable decoding.
3. Feature Extraction: Software analyzes the cleaned brain signal and identifies meaningful patterns in the form of specific neural signatures that relate to intended movements, speech, or other outputs.
4. Neural Decoding: An algorithm typically powered by machine learning maps extracted patterns to specific outputs. Here, intent becomes instruction: the system is recording what the brain is attempting to do and producing a related output.

Device Output: The translated information is then securely transferred to an external device, which may be a computer cursor, a speech synthesizer, a robotic limb, or a communication software, executing the user’s intended action in the real world.

It’s important to note that some modern BCI systems perform preprocessing, feature extraction, and neural decoding on dedicated external hardware downstream from the implant. However, next-generation systems, including future iterations of the Connexus BCI, aim to move these functions on-chip, further reducing latency and external processing demands.

Not all technology is created equal, and the same goes for BCIs. The type of device and which area of the brain it records information from determines everything from signal quality to performance, creating huge discrepancies between what users are able to do with each device.

Electroencephalography (EEG)

Electroencephalography describes a type of non-invasive method of recording electrical activity from the surface of the scalp, providing an indirect measure of brain activity. Types include eye-tracking devices, consumer headsets, head caps with electrodes, and other wearable technology.

EEG has been around for decades, but its utility for complex real-time control can be limited due to comparatively poorer spatial resolution – meaning EEG struggles to understand where brain activity originates, which reduces signal specificity.

While EEG does offer excellent temporal resolution in that it can track brain activity within milliseconds, there is a poor signal-to-noise ratio due to its placement outside of the skull, making it difficult for EEGs to detect or process data.

What this means for users is that while EEG devices are accessible and relatively easy to set up, they simply do not have the capability to support or restore complex, low-latency communication. 

Surface Electrocorticography (ECoG)

Surface electrocorticography (ECoG) devices record from the brain’s surface – underneath the skull but not inside the cortex itself. Even though ECoG devices are closer to neurons than EEGs, they still have some limitations as they are not located directly near individual neurons. 

The information they are recording from the brain are referred to as Field Potentials, which are average sums of neuronal activity. Action Potentials (APs) – captured in intracortical BCIs described below – track more precise, individual neuron firings to produce more specific, real-time outputs.

For users, ECoG offers a middle ground: it offers improved signal quality, but without the richness of individual neural data. Communication restoration demands speed and specificity, and the signal averaging inherent to ECoG signals can be quite limiting.

For a deeper look on ECoG, head to our blog: Electrocorticography (ECoG) surveys the landscape.

Endovascular BCIs

Endovascular BCIs are inserted via the jugular vein and sit inside a blood vessel within the brain, recording data without a direct connection to brain tissue. While it does not require entry to the brain, there are clear tradeoffs: Once inserted, the device cannot be removed, and users are prescribed lifelong blood thinners.

With limited electrode count, limited signal temporal resolution and a very limited data output, it isn’t clear how different endovascular BCIs are compared to EEGs in terms of capability and performance.

Regardless, for users who need more control, more accuracy, and more capability, the answer lies deeper.

Intracortical BCIs

One of the most promising types of brain-computer interfaces is an intracortical BCI, which involves the placement of microelectrodes 1 to 2 mm into the cortex – into the surface of brain – in order to record the most detailed measure of neuronal activity possible (action potentials).

Examples include microelectrode arrays like the Utah Array, the Neuralink N1, and Paradromics’ Connexus® BCI, each placed next to individual neurons to record single-unit or multi-unit APs, in contrast to the average sums of activity captured by surface devices. Paradromics and Neuralink are currently the top BCI companies developing fully implantable high-data-rate intracortical BCIs that enable real-time speech restoration.

In addition to location, the number of electrodes also matters when it comes to producing complex outputs, as high-channel devices provide higher data rates and better outputs. For a more detailed perspective on why this matters, head to our blog: Why brain speed doesn’t limit BCI potential: The case for high-data-rate BCIs.

Engineering breakthroughs power intracortical BCI technology

As the human body is a harsh environment, significant technical and engineering advancements must be built for intracortical or implantable BCI devices. This requires BCI technology that bridges the gap between cutting-edge semiconductor technology – which often is not designed to be inside the body – and traditional medical device manufacturing. 

Other design considerations include miniaturization, power consumption, data rates, and the ability to operate near or within the brain without causing tissue damage. In tandem, machine learning and AI are crucial for analyzing these vast datasets and identifying relevant brain states and biomarkers in order to produce the right outputs.

Whereas wearable BCI devices have plateaued in their communication capabilities, implantable BCIs have continued to show increasing performance as hardware and machine learning algorithms have improved.

The primary focus of many BCI applications is on restoring function for individuals with neurological impairments. For example, assistive and augmentative communication medical devices allow individuals who have lost the ability to speak or type to communicate through text or speech, or to directly operate a computer cursor or mouse. 

Pioneering efforts like the BrainGate clinical trial program helped lay the groundwork for today’s intracortical BCIs, showing impressive feats such as decoding imagined handwriting, which has set new records for BCI-based communication speed. More recent work, including ongoing clinical trials from Paradromics and others, is pushing these benchmarks even further.

Beyond communication, BCIs are enabling precise control of robotic arms and cursors – and some are even restoring movement through muscle stimulation. 

‍Therapeutic applications for mental health are also a powerful possibility, potentially enabling advanced, closed-loop treatment for conditions like depression, addiction, and chronic pain.

BCIs introduce a complex ethical landscape. Questions of data privacy are paramount, as neural data contains intimate information, including potential biomarkers (that is, measurable processes within the body, like blood pressure or heart rate) for mental states or even medication use. 

The immediacy of this data – revealing intent milliseconds before an action – raises concerns beyond those typically associated with other health data, such as genetic information.

The line between restoring function and enhancing human capabilities is a topic that sparks both excitement and apprehension, making it yet another important consideration for the field of neurotechnology.

Overall, BCIs represent a huge potential for healthcare, mental health treatment, and society at large. Experts foresee a future where neurotechnology is as commonplace and life-saving as  a cardiac pacemaker. 

To make it happen, expect to see collaborations across diverse disciplines like neuroscience, engineering, computer science, manufacturing, regulatory, and even ethics, as these areas are all crucial in translating research into clinically viable medical devices.

Companies and research institutions are increasingly focusing on user-centric design, aiming for solutions that are not only technologically advanced but also intuitive, durable, and genuinely improve quality of life. 

As funding for neurotechnology expands and the neurotech market continues to grow, we anticipate a wave of innovations that will make currently unimaginable interventions a reality, fundamentally transforming the relationship between brain and machine.

While the science of brain-computer interfaces is moving fast, the people driving it forward don’t exist solely in the world of academia. The BCI community is full of patients, caregivers, clinicians, and advocates who believe that losing the ability to verbally communicate doesn’t need to mean losing the ability to connect with others.

If you're interested in following this work, joining the conversation, or learning more about what's possible with today's BCI technology, we'd love to welcome you in. Join our community or learn more about our communication restoration clinical study.