Table of contents
1. Definition: What is a brain-computer interface?
2.3 Non-invasive (non-invasive)
2.3.2 Positron Emission Tomography PET
2.3.3 Functional Magnetic Resonance Imaging fMRI
2.3.4 Near Infrared Spectroscopy fNIRS
1. Definition: What is a brain-computer interface?
A brain-computer interface (BCI) is a system that allows communication between the brain and various machines.
They work in three main steps: collecting brain signals, interpreting them , and outputting commands to connected machines based on the received brain signals . BCIs can be applied to a variety of tasks, including but not limited to neurofeedback, restoring motor function in paralyzed patients, restoring communication function in patients, and improving sensory processing. Depending on the method used to collect brain signals, BCIs can be divided into three categories.
2. What are the types of BCI?
There are many different techniques to measure brain signals.
We can divide them into Non-invasive , Semi- invasive and Invasive .
The diagram below shows the different layers of the brain and where signals are taken from them.
Non-invasive: collect EEG signals, and place electrodes on the scalp, which is the outermost part.
Semi-invasive: ECoG signals are taken from electrodes on the dura or arachnoid.
Invasive (invasive): electrodes are implanted in the cortex to directly obtain signals in the brain parenchyma.
2.1 Invasive (invasive)
In neurosurgery, an invasive (invasive) type of BCI is implanted directly in the brain. One is a single-unit BCI, which detects signals from brain cells in a single region, and the other is a multi-unit BCI, which detects signals from multiple regions. The electrodes are of different lengths, eg a maximum length of 1.5 mm (Utah, Blackrock Microsystems) or 10 mm (FMA, MicroProbes) in MEA [1]. The quality of the signal is the highest, but there are some issues with the procedure, such as the risk of scar tissue formation. The body reacts to the foreign object and forms scars around the electrodes, which can lead to poor signal. Because neurosurgery can be a risky and expensive procedure, the target population for invasive BCIs is primarily the blind and paralyzed.
2.2 Semi-invasive
ECoG
Electrocorticography uses electrodes placed on the exposed surface of the brain to measure the electrical activity of the cerebral cortex. It was first used at the Montreal Institute of Neurology in the 1950s. This is called semi-invasive, but still requires a craniotomy to implant the electrodes. Therefore, it is only used if surgery is required for medical reasons such as epilepsy.
The electrodes can be placed either outside the dura mater (epidural mater) or under the dura mater mater (subdural mater). Strip or grid electrodes cover large areas of the cortex (from 4 to 256 electrodes) [2], allowing a wide variety of cognitive studies.
Advantages of ECGs:
High spatial resolution and signal fidelity
Anti-noise
Low clinical risk and robustness in long-term records[1]
high amplitude
spatial resolution
The benefit of ECoG over EEG is higher spatial resolution because the signal does not have to travel to the scalp. The spatial resolution in ECoG is one tenth of a millimeter, while that in EEG is centimeters [4].
What does spatial resolution mean here? We can compare the sharpness of an image. A picture with a higher spatial resolution is "sharper"; in other words, it appears more accurate because it contains more pixels per inch, showing more detail. Images with lower spatial resolution appear less sharp or blurry because each inch is made up of fewer pixels. Better spatial resolution allows us to understand more precisely where the signal is coming from. For an EEG, as electrical signals travel through the skull, they are attenuated due to the low conductivity of bone.
Anti-noise
ECoG signal is immune to noise and artifacts, e.g., EMG (electromyogram - caused by muscle movement) and EOG (electrooculogram - caused by eye movement)
low clinical risk
Electrode arrays do not need to penetrate the cortex, which is safer than invasive recordings
high amplitude
ECoG records up to 50–100 µV and 10–20 µV
There have been many studies on the use of ECoG in BCI, but they have been limited to cases requiring surgical removal of epileptic foci.
For example, in one study [5], researchers used ECoG to control a computer cursor in two-dimensional space. Five patients had subdural electrodes implanted for 7-14 days in preparation for surgery for epilepsy. After a brief training of less than 30 minutes, patients have been able to control the cursor in two-dimensional space, with an average success rate of 53-73%.
2.3 Non-invasive (non-invasive)
In the next section, we briefly review the main non-invasive techniques. There are several non-invasive techniques used to study the brain, of which EEG is the most commonly used technique due to cost and portability of the hardware.
MEG magnetoencephalography
PET Positron Emission Tomography
fMRI fMRI
fNIRS near-infrared spectroscopy
EEG
A comparison of different brain imaging techniques, in terms of spatial and temporal resolution, can be seen in the image below:
2.3.1 MEG
What are MEGs? (magnetoencephalography)
It is a functional neuroimaging technique that maps brain activity by using very sensitive magnetometers to record magnetic fields generated by the brain's naturally occurring electrical currents. "
How does it work?
MEG measures the magnetic field induced by electrical currents in the brain, which provides better spatial resolution compared to EEG [7]. Why? Magnetic fields suffer much less than electric fields because of the spatial blurring effects of the skull and brain fluid [8]. "MEG is most sensitive to tangential sources and less sensitive to radial sources". "MEG is superior to EEG at detecting high-frequency activity (for example, above 60 Hz). This is because magnetic fields pass through the skull and scalp, while electric fields are conducted through these tissues, reducing the signal-to-noise ratio at high frequencies."
2.3.2 Positron Emission Tomography PET
PET(positron emission tomography)
What is PET?
PET is a nuclear imaging technique used in medicine to observe different processes such as blood flow, metabolism, neurotransmitters.
How does it work?
A small amount of radioactive material (called a radiotracer) is injected into the bloodstream to reach the brain. In the brain, the radiotracer attaches to glucose and produces a radionuclide called fluorodeoxyglucose (FDG) [10]. The brain uses glucose, which shows different levels depending on the level of activity in different areas. The images from the PET scan are in color, where areas with more activity appear as yellows and reds in warmer colors. PET scans of the brain are often used to detect cancer or other diseases.
2.3.3 Functional Magnetic Resonance Imaging fMRI
fMRI(functional magnetic resonance imaging)
What is fMRI?
Functional magnetic resonance imaging, or functional MRI (fMRI), is a functional neuroimaging procedure that uses MRI technology to measure brain activity by detecting changes associated with blood flow. The technique relies on the fact that cerebral blood flow and neuronal activation are coupled. When an area of the brain is in use, blood flow to that area also increases.
Functional magnetic resonance imaging was developed in the 1990s. It is a non-invasive and safe technique that does not use radiation, is easy to use, and has excellent spatial and good temporal resolution.
2.3.4 Near infrared spectroscopy fNIRS
fNIRS(near-infrared spectroscopy)
What is fNIRS?
Functional near-infrared spectroscopy (fNIR or fNIRS) is the use of NIRS (near-infrared spectroscopy) for functional neuroimaging purposes. Using fNIR, brain activity can be measured through hemodynamic responses that correlate with neuronal behavior.
An optical technique to measure regional cerebral cortical activity
How does it work?
fNIRS measures changes in blood flow like fMRI, but uses a different technique, infrared light and a magnetic field.
When a mission begins, oxygen is consumed, and as complexity increases, so does the need for oxygen. fMRI measures how much oxygen is consumed. fNIRS can also measure how much oxygen is in the area.
However, the temporal resolution of fNIRS is inferior to that of EEG. fNIRS takes 10 samples per second, while EEG takes 500 to 1000 samples per second. And the spatial resolution is not as good as fMRI. For example, fMRI can image subcortical brain regions, while fNIRS cannot analyze images passing through the cortex and cannot capture any subcortical activation. In fact, many researchers at SfN have demonstrated their use of the instrument as a supplement to their EEG or fMRI data.
Advantage:
non-invasive
Carry
barrier free
Less sensitive to artifacts than fMRI and EEG
With a temporal resolution more similar to EEG fMRI, fMRI can record one sample every 2 seconds, and fNIRS can record 10 samples every 1 second.
fNIRS has better spatial resolution than EEG and better temporal resolution than fMRI.
2.3.5 EEG
definition
An EEG records the electrical activity of the brain on the surface of the scalp.
How does it work?
Electrodes are placed on the scalp to pick up the electrical current produced by the brain.
When a neuron fires, a dipole is formed, with a lower voltage at the synapse and a higher voltage at the axon. In the case of an inhibitory neuron, the dipole flips, and the voltage at the axon is lower and the voltage at the synapse is higher. What causes this voltage shift inside the neuron? Sodium ion (Na+) channels open along the dendrites, generating a large number of positrons, this positive charge moves along the axon, opens more sodium ion (Na+) channels, and causes the charge to move along the axon and discharge at the synapse , and release neurotransmitters. When groups of neurons fire together, they provide us with enough signal to measure from the scalp. We were only able to measure neuronal clusters using EEG (about the size of a quarter of the diameter).
Pros: It's portable and fits in a small suitcase (compared to MEG, which requires a dedicated room to be built). Lab-grade EEG systems can be expensive, but they are less expensive than other BCI methods. In recent years, an increasing number of commercial EEG systems have been released.
EEG data contain rhythmic activity that reflects neural oscillations. Oscillations are described by frequency, power and phase. Oscillations occur at a specific frequency (ie, at a certain rate). These variables include delta, theta, alpha, meta, and gamma. The study found associations between these rhythms and different brain states. For example, commercial EEG headsets, often used for purposes such as meditation, typically measure the amount of brain activity that occurs at alpha frequencies.
Spatial resolution:
The spatial resolution of EEG depends on the number of electrodes used. In research, a minimum of 32 electrodes and a maximum of 256 electrodes are usually used when higher spatial resolution is required. In general, EEG has a lower spatial resolution (eg, compared to ECoG and fMRI) because the signal needs to travel up through different layers to the skull. However, resolution can be increased using certain types of filters or by combining EEG with other tools such as fMRI. (image of electrode placement..)
More electrodes cost more time (eg, setup), bandwidth (for data collection and analysis), and money (for materials). Commercial headsets typically use fewer electrodes because high spatial resolution (ie, localizing the precise brain region producing the signal) is not necessarily required.
"The spatial accuracy of EEG is rather low, but can be improved by spatial filters such as surface Laplacian or adaptive source spatial imaging techniques"
Likewise, spatial precision is low because the activity recorded by the electrodes is a mix of different signals generated by different brain regions that are close to and distant from the brain region placed beneath the electrodes. Microscopic scale (less than a few cubic millimeters) = invisible to EEG, insufficient potential to reach the scalp. EEG can be used, but with more than 64 electrodes and spatial filtering techniques, mesoscopic scale (cortical plaques of a few cubic millimeters to several cubic centimeters) = can be detected with EEG. Macroscopic scale (large cortical area of many cubic centimeters) = easy to measure with EEG.
Time resolution
The advantage of EEG is that it has good temporal resolution. In one second, thousands of snapshots of electronic activity can be taken on different sensors. Depending on the experiment, as many as 500 electrodes can be used in EEG. They are used to be mounted on caps to collect data from the same scalp area.
Compare
Brain-computer interfaces can use any type of brain imaging. These include fMRI, PET and near-infrared spectroscopy (NIRS), which rely on changes in blood flow, and magnetoencephalography (MEG) and electroencephalography (EEG), which measure the brain's magnetic and electrical activity, respectively. fMRI and NIRS have high spatial resolution but poor temporal resolution; MEG and PET have high spatial and temporal resolution; EEG has low spatial and high temporal resolution. Currently, fMRI and MEG rely on expensive and cumbersome equipment; PET requires injecting radioactive substances into the blood. Therefore, methods relying on near-infrared spectroscopy (NIRS) and especially electroencephalography (EEG) are the most commonly used.