Trayana Hristova
Author
In the last two decades, brain-computer interfaces (BCIs) have become a widely spread and fascinating topic of research, mainly because of all the possibilities they offer for people with amyotrophic lateral sclerosis (ALS) or patients with spinal cord injuries. Since these conditions lead to severe impairments in communication and motor control, BCI systems are well-suited to help them. Brain-computer interfaces are a combination of hardware and software that enable people to interact with their surroundings without the usage of peripheral nerves and muscles. This communication happens thanks to control signals often generated from electroencephalographic activity. Thus, a BCI is a system that can decode brain activity following five consecutive stages. These are signal acquisition, preprocessing, feature extraction, classification, and control interface. Overall, the functioning of a BCI system consists of extraction of the brain signal and turning it into a more “readable” instruction for the connected device, which could be a computer or a wheelchair.
This idea has been on the rise the past few years after a prolonged period of muted interest, and this is due to a couple of reasons. Deciphering all the information from the brain has been an extremely challenging task because of the complexity of the neuronal dynamics involved along with the limited capacity of the hardware available, but notable advances have been made on both fronts as of late. Interest has also been growing outside of the laboratory precisely because it could prove to be beneficial for people with the aforementioned conditions, and even find other applications not considered previously. Companies such as Emotiv and Neurosky have developed their own initial prototypes oriented towards a broader audience. Thus, in this article, we are going to attempt to explain what the existing approaches to BCIs are, as well as the main types and focus on the difference between invasive and non-invasive ones.
First, to explain how a BCI works, we must look deeper into the neuroimaging approaches that exist. These are the techniques that allow us to record and analyze brain activity which can then be translated into a tractable electrical signal. For BCIs, two main types can be recorded – electrophysiological and hemodynamic. We will focus on the first type, electrical activity generated by electro-chemical transmitters (neurotransmitters) exchanged between neurons for cell communication. The second type of brain activity is a process in which increased neuronal activity leads to a greater localized demand for oxygen, in turn increasing the regional cerebral blood flow (rCBF) via neurovascular coupling. There are a few ways in which electrophysiological responses can be measured – electroencephalography (EEG), electrocorticography (EcoG), magnetoencephalography (MEG) and electrical signal acquisition in single neurons.
The most widespread method of recording electrical brain activity is EEG. This is a non-invasive method, and it presents numerous advantages. Firstly, it has a very low risk for the users since it is a non-invasive method. It is also cheap compared to the others and has a high temporal resolution. The technique itself consists of sets of sensors called electrodes placed on the patient’s scalp to detect electrical activity. These signals are then amplified and recorded as waveforms to give insight about the different brain activities in certain areas. The entire system consists of electrodes, amplifiers, A/D converter and a recording device. The EEG signal is obtained by measuring the potential difference over time between the recording electrode and a reference electrode. There is an additional third electrode, the ground electrode, customarily used for noise reduction. The minimum configuration for an EEG setup therefore consists of 3 electrodes, but multi-channel configurations can comprise up to 256 or more.
The amplitude of the recorded from EEG signals is measured in microvolts and their frequency ranges from below 4 Hz to above 100 Hz. These waves are called delta, theta, alpha, beta, and gamma, starting from the lowest to highest frequency, respectively. Gamma waves, for example, are less likely to be used in EEG-based BCIs because they are easily affected by artefacts such as electromyography (EMG) or electrooculography (EOG) with the former being a way of measuring electrical activity from skeletal muscles and the latter – from muscles controlling eye movements.
However, there are also some inconveniences with this technique. The quality of the recorded signals is highly susceptible to background noise coming from the scalp or other layers, thus making it difficult to extract meaningful information from the recordings. There are other ways of recording brain activity non-invasively, such as magnetoencephalography (MEG), which measures the magnetic field produced by the electrical activity of neurons. This method has some advantages over other neuroimaging techniques: for instance, it picks up activity that EEG has a much harder time picking up.
Invasive methods of recording brain activity rely on implanting microelectrode arrays (MEA) inside the skull. They can be placed either on the surface of the cortex or inside it. The first method, as mentioned earlier, is called electrocorticography (ECoG) and it can place the electrodes either outside the dura mater (the thick membrane surrounding the brain), called epidural electrocorticography, or under the dura mater – subdural electrocorticography. This approach can have numerous benefits for more severely paralyzed patients, but it also hides some potential health risks. Firstly, there are always issues when a foreign body is introduced forcefully, and the microelectrodes are no exception. There have been proposals to introduce neurotrophic substances in these brain implants, which are substances aiming to promote better biocompatibility. The materials used in these mediums’ production aim to promote neuronal growth and regeneration and increase the recording performance of the electrodes over time.
A second problem with the invasive methods is that there must be a connection between microelectrodes and the external device that will be controlled. This connection needs to be wireless to reduce infection risk, but a wireless connection is not as strong as a wired one.
That said, let us discuss the two techniques mentioned above. ECoG is a way of measuring electron activity by placing electrodes directly on the surface of the brain. Compared to EEG, this method has a higher temporal and spatial resolution as well as higher amplitudes and lower vulnerability to artifacts such as blinks and eye movement. The problem with ECoG is that it requires a craniotomy for the grid to be implanted which entailes serious potential health risks. In patients, this technique has been used for analysis of alpha and beta waves or gamma waves during voluntary motor action. An ECoG-based BCI was for instance developed successfully allowing to control a one-dimensional cursor.
The last method we are going to discuss is intracortical neuron recording. This technique allows us to measure electron activity from inside the grey matter of the brain. In this invasive modality, microelectrodes are being implanted directly inside the cortex to capture spikes and local field potentials (LFPs) from neurons. When it comes to BCI implications, Kennedy et al. have constructed a system allowing users ”to control movements and flexion of a cyber-digit finger on a virtual hand”.
Overall, the BCI field is still in its early stages (though one might say reaching an inflection point even!) and as promising as it may be, the invasive nature of some of its applications raises questions about whether comparable results can be achieved without having to open up one’s skull. For this to be true, more improvements must be made in the resolution of EEG systems in combination with the other techniques mentioned earlier to achieve optimal results. And although one of the most prominent applications of these interfaces is to help physically impaired people, more applications are being explored every day.