Neural engineering is an exciting and highly interdisciplinary area of biomedical engineering that combines engineering, neuroscience, psychology, and machine learning to understand, repair, replace and enhance the properties of neural systems.
There are several ways we can measure neural activity. Either invasively by placing electrodes inside the body or non-invasively from outside the body.
We measure the bioelectrical activity of the neurons in the brain non-invasively from the scalp by using electrodes. The recorded signal is also called an electroencephalogram or EEG. You may know this method from a hospital stay.
The electrodes were specially developed for electrochemical measurements. They are often conveniently integrated into caps so that we only must wear them. Signal recording can be messy because we use an electrolytic gel to increase conductivity and thus get better signal quality.
Signal amplitudes are typically very small. More specifically, in the hundred microvolts (+-0.0001) range. Similarly, a way to measure action potentials that neurons send to muscles are adhesive electrodes that we place on the skin over the muscle. This signal is called the electromyogram or short EMG.
You may ask how it is that we can measure action potentials non-invasively from outside the brain? The reason for this is volume conduction.
Ion currents in neurons propagate through biological tissue, just like waves spread through water when you throw a stone into it. The waves arriving at the shore will be weaker and provide a distorted image, but you can still see them.
The same is true for neural signals. Electrodes can pick up some activity from outside the head.
This naturally raises the question of how accurate and specific our measurements are? I would like to explain this with a Gedankenexperiment (thought experiment).
Let’s assume there is a football game in a stadium near you that is filled with 90,000 fans. Let’s also assume that you are sitting comfortably at home reading a book.
You can't hear a fan shouting at the referee in the stadium. But when a team scores a goal, you can hear the cheering.
This means, we cannot listen to the activity of an individual neuron. We “only” get a bird eyes view of what is going on because the EEG measures the superposition (sum) of the activity of millions of neurons.
The evoked potential and the neural oscillation are the most studied neural patterns that can be found in the non-invasively recorded signals such as the EEG or the EMG.
An evoked potential is an electrical potential of a specific pattern generated by a specific part of the nervous system after the presentation of a perceptual stimulus.
An example of this is the flash of light we perceive when the light of a traffic light changes. If the stimulus has a meaning for us and we attend it, for example when the traffic light turns green, then the pattern of the EP changes.
Since we involve different networks that process the sensory information, different action potential sequences are generated and as a result also the EP changes. However, EP amplitudes are very small and typically only a few microvolts. So, these signals are hidden in the EEG, and we need to figure out ways on how to find and extract them.
Neural oscillations are rhythmic or repetitive patterns of neural activity in the central nervous system.
Oscillations look like waves that propagate through water. If large number of neurons synchronise their activity, then their patterns add up and generates oscillations with large amplitudes, which can be observed in the EEG with the naked eye. Hans Berger in 1924 was the first to see such oscillations in the EEG of Humans.
A simple way to see such oscillations is to measure the EEG over cortical visual areas (in on the back of the head). When the eyes are closed, the neurons that process the information sent by the eyes to the brain have nothing to do. This means they all do the same and their activity synchronizes.
The brain is very complex and challenging to study. Available neuroscientific knowledge, and mathematical and computational models, do not describe the causal relationships between nervous system activity and measured activity patterns in sufficient detail and explain what those relationships are.
The challenge is that superimposing the activity of different neurons or neural networks can produce almost identical patterns in non-invasive signals.
Researchers therefore design experiments that isolate patterns and enable us to learn more about the function of the brain. This knowledge is then used to develop systems that can correctly recognise and interpret neural activity patterns. Advanced machine learning and pattern recognition methods are commonly employed to detect neural patterns.
You now have a basic understanding of the neural patterns that exist and the challenges we face in reading them correctly. It is time for you to switch into the role of a junior scientist and researcher EPs and oscillations, with our virtual lab.