Brain-computer interfaces (BCI) are a quickly evolving subject of research. BCI is used to process brain signals to control a machine, generate speech or control a prosthetic. So far, creating an accurate BCI model has undergone plenty of research, however, a less researched area is how to reduce the price. This is problematic for lower-income countries and people in need of a Brain-computer interface. My topic of study is trying to reduce the price while keeping the accuracy at an acceptable level. There are two types of BCI, invasive and non-invasive. I will be studying non-invasive because it is not harmful and does not require an implant. The main issue related to the price is the amplifier, which amplifies the weak signals received from non-invasive BCI and removes the artifacts (unwanted signals in the EEG recording). This could make a huge difference for people with ALS or other neurological and spinal disorders. This could help them live a more normal life by restoring mobility and speech.

This project used 3 non-invasive electrodes placed in positions Fp1, Fp2, and Fpz to monitor the prefrontal cortex. These positions are at the front of the forehead. Because of the amplifier's design, the recordings from the electrodes were combined into one channel. It is better to use less expensive hardware in exchange for more advanced software because if we wanted to supply a group of people with this technology, the code can be made open source, but everyone has to buy their hardware. We used ChatGPT to generate the code because we are looking for the most affordable way to make a brain-computer interface model. Since ChatGPT 3.5 is open source, people who would not otherwise be able to acquire the code, the people we are trying to help, can access the code and alter it to fit their needs. To make the model more affordable, we have replaced the Ganglion board with an AD8232 heart monitor. This is possible because they are both used to process electrical signals acquired via electrodes; one from the brain and the other from the heart. We also used single-use EMG/ECG electrodes, which are a fair bit cheaper than the reusable ones. If we wanted to save even more money, the electrodes could have been reused by putting a rubber ring around the sides and replacing the electrode gel after each use. To prove the model's accuracy, we observed the brain waves while performing various activities, such as chess (which uses the prefrontal cortex) and listening to calming music while relaxing. While playing chess, we noticed patterns such as a higher Hz while thinking about a chess move. After some trial and error, we figured out that the sampling rate should be at least twice the Hz of the brain waves you are measuring. In this case, we measured it at a sampling rate of nearly 200. After recording the brain waves, we took a one-second snapshot of the data to analyze and determine the type of brain wave it was. Instead of using the serial plotter provided by the Arduino app, we put the EEG data on a spreadsheet and made a graph of the data. While observing brain waves, we limited external stimulation other than the activities we were testing to ensure accurate results. Our last action was to buy a Bluetooth module to transmit the EEG data wirelessly to a display screen, however, we lost the instructions, which contained a code that was vital to its use. 

What Are Neurons

Neurons are the tiny building blocks of our brains that receive external stimuli (the soma are 4-100 micrometres). With that, large groups of neurons send electrical signals to activate our muscles or generate a thought. Think of a neuron like a tree; they have 3 main parts, the soma, (AKA the cell body) that controls the cell and contains its genetics, the dendrites, that receive signals from other neurons and transmit them to the cell body. The third part of the neuron is the axon. The axons are encased by a sheath called myelin that improves the potential of the signals that travel along them, called action potentials. The axon generates action potentials and sends them to other neurons as an action potential. This message passes through synapse (the gap between 1 cell’s dendrites and another’s axon) through chemical neurotransmitters. An action potential is made by changes in the neuron's electrical charge. Action potentials travel down the axon and release neurotransmitters (chemical messages.) Another type of brain cell, neuroglia provides support and structure in the brain. The electrodes sense the oscillations of voltage coming from millions of neurons releasing action potentials and can decipher them into useful insights. 

 

(Woodruff, n.d.)

Regions of the Brain

There are 3 primary brain regions: the midbrain, the hindbrain, and the forebrain. Inside of each, there are various subdivisions. 

• The Forebrain: Responsible for thinking, planning, language processing, and decoding information from our senses. It accounts for 85 percent of our total brain mass.

• Thalamus: The thalamus has two oval masses connected by a bridge. It processes information about all the senses except smell and sends them 
• to the cerebral cortex.
• Hypothalamus: The hypothalamus controls the smooth muscle on the sides of the intestines, stomach, and blood vessels. It is the first to detect body changes and stimulates glands and organs to release hormones. It also is the part of the brain that turns emotions into physical responses and controls body temperature, eating and drinking, and falling asleep.
• Amygdala: The amygdala is a small almond-shaped region of the brain that feeds into the limbic system. 
• Pituitary and pineal glands: These glands work closely with the hypothalamus and release several hormones. The pineal gland releases the hormone that causes skin pigmentation (color), and many of the hormones released by the pituitary gland regulate the activity of other glands.
• Cerebral cortex: The cerebral cortex is by far the largest region of the brain. It is named after its surface which looks like bark. It is responsible for making sense of and understanding the signals sent from sensory input and each sense seems to have a separate area dedicated to receiving and processing the sense. Inside the cerebral cortex, there are several subdivisions:

 - The association cortex is responsible for long-term planning, interpretation, and organizing ideas. 

- The occipital lobe, responsible for visual functions, has an area that perceives vision, a section that is responsible for the visual association, and a region that is responsible for visual memory.

- The prefrontal cortex is technically part of the limbic system (the emotional brain). It is connected to all other parts of the cortex with association fibers (axons). It is responsible for predicting consequences, limiting bad behavior, keeping the mind focused on the task and goal at hand, and making the thought process continuous.

- The precentral gyrus/motor cortex is responsible for conscious movement. There is a “map” in this area devoted to controlling the body. Areas of the body like hands and mouth have a disproportionately large representation in the brain. They require a lot of control, whereas areas like the torso are smaller because they need less control.

- The postcentral gyrus/ somatosensory cortex has a map similar to the one described in the motor cortex, however, it is instead dedicated to receiving and processing sensory signals from the body.

 

 

 

• The Hindbrain: The hindbrain coordinates movements in the body, as well as vital body functions like heartbeat and breathing.

• Brainstem: the brainstem controls the heart, breathing, and blood vessel size. At the bottom of the brainstem, there is another region responsible for passing signals between one side of the brain and the other side of the body. It controls head and shoulder movement, the mouth, and hearing. At the top, there is something called pons that passes information on to other parts of the brain.
• Cerebellum: The cerebellum means “little brain” in Latin. It is the second largest region of the brain, beat by only the cerebral cortex. It takes signals from the motor cortex and sends them down the spinal cord. It also receives information about the movement from the spine and compares it to the original command so that the thalamus can make modifications. Because the left side of the brain controls the right side of the body and the right side controls the left side of the body, the cerebellum receives information from the opposite side of the brain than the side of the body it is intended for, the cerebellum switches the signal over to the opposite side of the body, and when it receives feedback from the body it sends the signals back to the original side of the brain.
• Hippocampus: The hippocampus is named for its shape as a seahorse. It is responsible for retaining short-term memory information and communicates with the entire cerebral cortex, which is responsible for long-term information. 

• The Midbrain: The midbrain transmits information between the hindbrain and the midbrain.

(National Library of Medicine, 1992) 

 

Types of Brain Waves

Delta: Delta waves are under 4 Hz, seen in developing baby and adult dreamlessly sleeping brains.  They are easy to confuse with artifacts (features in the EEG signals that are not important, like movement and noise.)

Theta: Theta waves are from 4-7 Hz, and are seen in the developing brain, drowsiness, meditation or sleep. They are also found in meditative concentration, including mental calculations and conscious awareness.

Alpha: Alpha waves are seen from 8-12 Hz, and are found in mental effort and remembering things. 

Beta: Beta waves are seen from 12-30 Hz and generally occur with motor movement. 

Gamma: Gamma waves are seen from 30-100 Hz and are associated mostly with body movement more actively than beta waves. Some studies also associate them with visual and auditory reception. They are mostly not used in BCI because there are too many artifacts.

 

(Nicolas, 2012)

 

Uses of BCI

Some uses of BCI are:

• People with motor impairments

If people have a motor impairment or spinal disorder, BCI can help. Instead of the brain signals having to travel down the spinal neurons, the brain signals could be directly captured with BCI and put into use with prosthetics, wheelchairs, or other assistive technology. One large impairment is a neurological disorder called ALS

• The broader market

BCI can also be useful in the broader market. It is overall an easier way to make things work; from drones to computer mice and even military technology, it has more uses that can be thought of at a glance.

 

(University of Calgary, n.d.)

 

BCI Stages

Signal Acquisition

In this step, electrodes are placed on the surface of the scalp and gather brain signals based on the oscillations recorded (in this context “oscillation" means the fluctuations of voltage coming from the brain; the amount of times that the brain signal chart reaches a peak within a timeframe). 

Feature Extraction

In this stage, the important signals are extracted from the brain signals, hence determining the user’s intent. The “artifacts” (signals in the mix that come from noise or body movement.) are also removed in this stage. 

Feature Classification

This stage uses machine learning to classify the signals and match them to the command associated with that pattern of brain waves. For example, when AI recognizes a signal, it matches it with the corresponding machine command, like bending a prosthetic elbow.

Feature Translation 

After the intended commands are recognized in the previous stage, this stage matches the command to the brain pattern and sends it to the application.

Application 

This stage involves controlling the application with the brain signals sent in from the processing stage. This can be anything from a prosthetic to a mini video game. 

Feedback

This stage takes the sensory input from the prosthetic and sends it back for a more immersive and better experience with a prosthetic. It is optional and adds expenses. 

 

(National Library of Medicine, 2023)

 

Parts of BCI

EEG Electrodes:

The EEG electrodes are the part of the design that executes the signal acquisition. It records the fluctuation of voltage that the brain emits. Although sometimes electrodes are implanted onto the surface of the brain via surgery, the ones that we are using are non-invasive, meaning that they are placed on the outside of the head. 

EEG Amplifiers:

One of the issues with non-invasive BCI is that the signals are very weak. Because of this, expensive amplifiers are needed. These amplifiers, as well as amplifying the signals, often work to remove artifacts (brain signal patterns that come from movement or noise.) This is the most expensive part of BCI.

Arduino:

The Arduino is used either for signal processing to convert the brain signals into commands or to send them on to the computer to view visually. The software is uploaded onto the Arduino.

Software: 

The software to run the program is either acquired from the Python library, a portfolio of a massive number of Python programs, or the openBCI GUI which is used to display the brain waves on a screen. They are both free and open source.  

Electrode Gel: 

The electrode gel is used to conduct the brain signal through the skin to the electrode. 

Wires: 

Although quite obvious, the cables connect the different components and transmit information along the system. Some of the wires can be replaced with Bluetooth modules, making it easier to transport and less clunky.  

 

(Roy, 2023)

 

Pros and Cons of BCI

 

 

(U.S. Government Accountability Office, 2022)

 

These diagrams show how brain wave frequency changes when different external stimuli are used. The rock music, chess, and typing graphs show what looks like gamma waves, and the relaxing with music one shows what looks like theta waves.

 

Brain-computer interface can help people with paralysis, brain stem stroke, or spinal cord injury live a better life. Creating a more affordable version benefits low-income people who would otherwise be unable to access the technology. As well as making it more accessible to more people, it is also useful for people wishing to study the brain but not having enough funds. As financial inequality grows, having access to an affordable brain-computer interface model will become even more important. Although affordability is important, accuracy is even more important.

For this reason, we have created a model that, as well as being accessible to all, retains an adequate accuracy level. To do this, the OpenBCI ganglion board was replaced with an AD8232 heart monitor, which serves a similar purpose. We have also used fewer electrodes than usual, resulting in less advanced hardware and software. Lastly, we have used ChatGPT 3.5 to generate the code, demonstrating that people who need a BCI have a relatively easy and open-source way to acquire the software. To test our design, we recorded our brain waves while performing activities such as playing chess, relaxing, and typing. We found that as the alertness of the user increases, the frequency also rises. For playing chess, listening to rock music, and typing, the user’s brain emits gamma waves. For relaxing with music, the signals seemed more in the theta wave range.

Despite its advantages, our model only has one channel, so turning signals into computer commands may be difficult. The results of this study will change slightly across different people, so knowing how brain waves differentiate across changing populations will be vital in creating a commercial BCI that is universally accessible.

 

References

National Institute of Neurological Disorders and Stroke. (2023, November 15). Brain Basics: Know Your Brain | National Institute of Neurological Disorders and Stroke. National Institute of Neurological Disorders and Stroke. Retrieved January 11, 2024, from https://www.ninds.nih.gov/health-information/public-education/brain-basics/brain-basics-know-your-brain

National Library of Medicine. (1992). Major Structures and Functions of the Brain - Discovering the Brain. NCBI. Retrieved January 11, 2024, from https://www.ncbi.nlm.nih.gov/books/NBK234157/

National Library of Medicine. (2023, August 4). Brain computer interface: trend, challenges, and threats. NCBI. Retrieved January 11, 2024, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10403483/

Nicolas, L. F. (2012). Sensors | Free Full-Text | Brain Computer Interfaces, a Review. MDPI. Retrieved December 20, 2023, from https://www.mdpi.com/1424-8220/12/2/1211

Queensland Brain Institute. (n.d.). Types of neurons - Queensland Brain Institute - University of Queensland. Queensland Brain Institute. Retrieved January 11, 2024, from https://qbi.uq.edu.au/brain/brain-anatomy/types-neurons

Roy, A. (2023, July 3). Everything you need to start building HCI & BCI Projects | Launching DIY Neuroscience Kits. YouTube. Retrieved January 3, 2024, from https://www.youtube.com/watch?v=_dAHZdoh_kg

University of Calgary. (n.d.). What is BCI? | Calgary Pediatric Brain-Computer Interface Program | Cumming School of Medicine | University of Calgary. Cumming School of Medicine. Retrieved November 30, 2023, from https://cumming.ucalgary.ca/research/pediatric-bci/bci-program/what-bci

U.S. Government Accountability Office. (2022, September 8). Science & Tech Spotlight: Brain-Computer Interfaces. Government Accountability Office. Retrieved January 11, 2024, from https://www.gao.gov/products/gao-22-106118

Woodruff, A. (n.d.). What is a neuron? - Queensland Brain Institute - University of Queensland. Queensland Brain Institute. Retrieved November 25, 2023, from https://qbi.uq.edu.au/brain/brain-anatomy/what-neuron

OpenAI. (2024). ChatGPT (3.5) [Large language model]. https://chat.openai.com

 

Thank you to the following people for helping me with the science fair project: 

• Ms. Webber &  Ms. Summerscales for teaching the science fair, and paving the way for future science fairs at banded peak.
• My dad, for helping to write the code that visualizes the data on Google Sheets
• My mom did several things, such as helping me edit the text.
• ChatGPT 3.5, for writing the code
• Hudson Allred is my partner in this project, but he did not get onto the platform. 

(We did this to show people who might need a BCI that you don’t need to be a software expert to create one)