Noninvasive Sensors of BCI

Sensors in Brain-Computer Interfaces: Mapping the Mind in Motion

Sensors are the gateway between brain activity and digital commands. These sensors are mainly involved in detecting brain signals.

Broadly, sensors range from invasive, partially invasive, and noninvasive. This blog post is limited to discussing noninvasive sensors.

Noninvasive BCIs record brain activity without surgery, making them safer and more accessible. The most common sensor is EEG, which uses scalp electrodes to measure voltage changes from large groups of neurons.

Two Worlds of Noninvasive Sensors: Metabolic vs. Electrophysiological

• Brain sensor techniques used in BCIs fall into two broad categories:

Metabolic (Hemodynamic): PET, fMRI, fNIRS

Electrophysiological: EEG, MEG

Metabolic Sensors: Following the Oxygen Trail

When you plan to move your right hand, the corresponding brain region involved in planning and executing the movement becomes active and undergoes electrochemical changes. This activity raises its need for oxygen, leading to an increased flow of oxygen‑rich blood into that area. Metabolic sensors detect the oxygen‑rich areas.

1. fMRI (Functional Magnetic Resonance Imaging)

fMRI uses magnetic resonance imaging (MRI) technology to detect which brain regions are "asking" for more oxygen compared to dormant areas. This ability to pinpoint where activity happens is called spatial resolution.

fMRI image of the brain

But sensors in BCI, including fMRI, don’t just measure where activity happens, they also measure when the activity happens. This ability to track the timing of brain activity is known as temporal resolution, and in fMRI it is relatively poor compared to other types of sensors, about 1–2 seconds.

In simpler terms, fMRI takes “snapshots” of brain activity every couple of seconds, so it can miss very fast processes that happen in milliseconds. That’s why it’s excellent for mapping brain regions but less suited for tracking rapid thought patterns or split‑second reactions.

fMRI is noninvasive, safe, and does not use radiation. However, because it requires heavy MRI machinery, its use is largely restricted to research labs rather than everyday applications.

2. fNIRS (Functional Near-Infrared Spectroscopy)

fNIRS works similarly by measuring oxygenated blood flow using near‑infrared light. It is noninvasive and often referred to as optical topography or near‑infrared imaging (NIRI).

fNIRS works by shining near‑infrared light (650–950 nm wavelength) into tissue and measuring how much light is absorbed or scattered. Human tissues are relatively transparent in this range, but hemoglobin strongly absorbs NIR light. Because oxygenated (O₂Hb) and deoxygenated hemoglobin (HHb) have distinct absorption spectra, fNIRS can estimate their concentrations and thus track blood oxygenation. When a brain region becomes active, oxygen‑rich blood flows in, changing the balance of O₂Hb and HHb—these shifts are detected by fNIRS sensors placed on the scalp.

Near‑infrared spectroscopy (fNIRS) provides a more practical alternative to the hardware limitation of fMRI: it achieves comparable spatial resolution (though limited to cortical areas 1–3 cm deep) with far less technical effort and cost compared to other metabolic sensors. Importantly, fNIRS systems are portable, making them suitable for use in everyday environments, even at home.

3. PET (Positron Emission Tomography)

PET uses radioactive compounds like Flurodeoxyglucose (FDG) tagged to red blood cells. These compounds accumulate in active brain regions, allowing researchers to measure concentration and infer activity.

PET scan image of the brain

One drawback of PET scans is that they can feel invasive for patients. To start, a small amount of radiotracer has to be injected into the body either all at once or slowly over time. While this is safe and routine, it can still be uncomfortable, especially for people who dislike needles.

For the most precise results, doctors often need to take blood samples during the scan. This adds extra steps, makes the process more complex, and can increase discomfort.

Nicolas-Alonso LF et al., 2012. Brain-computer interfaces, a review

Electrophysiological Sensors: Capture Brain Electromagnetic Activity

Electrophysiological methods detect the electrical or magnetic signals generated by neurons during activity.

1. EEG (Electroencephalography)

EEG measures electrical potential differences across the scalp, which are generated when a neuron fires. It records the activity in milliseconds, making it ideal for studying quick cognitive processes. For example, when a person sees a written word, their brain produces an electrical response within a few hundred milliseconds. EEG captures this rapid response, giving researchers a precise window into how the brain recognizes and processes language almost in real time.

Brain signals are hard to measure because the skull and scalp block and weaken them, so only the firing of large groups of neurons shows up clearly. On top of that, signals can get messy from outside electronics, muscle movements, eye blinks, heartbeats, or even oily hair and other skin conditions that affect electrode contact. Hence, its spatial resolution is lower than that of metabolic sensors.

To keep the data clean, researchers use special tools like shielding, amplifiers, and noise‑reduction techniques to boost the signal and cut down interference.

2. MEG (Magnetoencephalography)

When a neuron fires, both magnetic and electric waves are produced. Magnetic waves pass through the skull with minimal distortion, unlike electrical activity. This magnetic transparency provides an advantage over EEG in having slightly finer spatial resolution and capturing distinct signal properties, making them synergistic tools in brain research.

Therefore, to detect cerebral activity that creates tiny magnetic fields, the scanners must be closer to the brain’s surface. As a result, specific sensors are required for MEG, such as superconducting quantum interference (SQUID) sensors.

Other Sensory Modalities:

3. Electromyography (EMG)

EMG is another sensor type used in neurotechnology. It involves inserting fine needles into nerves that supply muscles. By stimulating these nerves and recording the resulting muscle contractions, EMG helps map neuromuscular function, which is especially useful in motor rehabilitation and prosthetics.

Why Combine Techniques?

No single sensor gives a complete picture. Metabolic sensors tell us where activity happens, while electrophysiological sensors tell us when. By combining tools like EEG and MRI, researchers can better understand the spatiotemporal dynamics of how thoughts unfold across both space and time.

------------------------------------------------------------------------------------------------------------------------

FOOTNOTES:

Invasive Sensors: Electrodes are implanted directly into or near brain tissue. They capture signals at the level of individual neurons (single‑unit activity), groups of neurons (multi‑unit activity), and local field potentials. This proximity provides high accuracy and resolution, but comes with surgical risks, immune responses, and higher costs.

Noninvasive Sensors: Methods like EEG, MEG, and fNIRS record brain activity from outside the skull. EEG, MEG tracks electrical signals in milliseconds, ideal for fast cognitive processes like reading a word. fNIRS and fMRI measure blood‑flow changes, revealing which regions activate during tasks such as motor imagery. Metabolic techniques offer high spatial resolution. But because vascular changes occur a certain number of seconds after their associated neural activity, they have poor temporal resolution compared to electrophysiological sensors.These approaches are safer and more accessible, though less precise due to lower spatial resolution and susceptibility to noise.