Science behind Brain-Imaging: A journey in the tools of neuroscience

Brain-imaging as a way to measure human brain activity started with the EEG technique as early as the decade of 1930 (Michel & Murray, 2012). It has since only gotten better, with the rise of more advanced techniques, such as PET, expanding on the idea of tracing “labeled” molecules into the body (Greenberg et al., 1981) that are used in the metabolism of neurons, highlighting highly metabolic states. Finally, the fMRI was created, as a paradigm shift from tracing “foreign entities” to recognizing differences caused to the body itself on increased metabolism (Roalf & Gur, 2017).

The brain-imaging scanners utilize revelations in fields like mathematics (e.g: Signal Processing in EEG/MEG) and physics (e.g: Positron annihilation in PET), not to mention the vast change in data processing capabilities on the field of computer engineering (prevalent in fMRI). While cognitive neuroscientists do not directly need the vast scientific background used in brain-imaging, understanding some under-the-hood details of their tools can only benefit them, helping them form proper expectations for their measurements. Also, deeper knowledge enables them to treat their tools less as black-boxes, while interpreting their measurements with needed perspective and avoiding misjudgments, such as the ones mentioned by Michel & Murray about the “reference dependency” of EEG (2012). The rest of this essay aims to unfold some technical background on EEG, PET and fMRI techniques

Electroencephalography (EEG) #

Background on (Electrical) signals and waves #

Electrical signals are fluctuations of electrical potential through time. It is mathematically shown that every signal can be expressed as a sum of waves, a procedure called “signal processing”, with the most notable example being the “Fourier Transform” (Grant, 2018). Signal processing “untangle” signals to a sum of waves letting us take a look at their “ingredients”.

Waves are also signals, with the sole difference that their fluctuations repeat themselves in time, creating a pattern. The frequency of appearance of this repetition is measured in Hz or Cycles Per Second (CPS). Also, the potential (how high or low a signal goes) itself is called “amplitude” and for electrical waves it is measured in Volts (Kaur & Kaur, 2015).

What a scanner does #

Electroencephalography aims to read electrical currents that escape the brain through the scalp. The electrical currents are generated by the firing of neurons perpendicular to the scalp that reside at the top-most cerebral cortex (Louis et al., 2016). The currents are captured by electrodes, placed on the subject’s head surface. The number of electrodes determines the EEG resolution, while the placement of them on the scalp is standardized through the 10-20 system (Homan, 1988).

During an EEG, electrode pairs capture electrical potentials, measured in microVolts (μV) (Kaur & Kaur, 2015). These potentials are very volatile, as they constantly change from positive to negative electrical potential, forming signals. These signals are normalized to minimize external influence, analyzed to frequency components through signal processing, and spatially grouped, to provide information about the dominant wave frequencies of the brain cortices (Kaur & Kaur, 2015).

The brain through its frequencies #

Processing the EEG signals provides waves with typical frequencies from 5 to over 30 Hz (Kaur & Kaur, 2015). It is described by Stevens & Zabelina, that the EEG wave frequencies are grouped into “bands”, just like electromagnetic frequencies are split into colors (2019). Also, they enumerate through the different frequency bands: “delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30+ Hz)”, as well as the brain state associated with each band, through the vast literature on them: “Delta – deep sleep; theta – error monitoring, cognitive control; alpha – eye closed, relaxation, internally-directed attention, network inhibition; beta – being awake, affective and cognitive processing; and gamma – active processing of sensory information, perceptual binding” (Stevens & Zabelina, 2019). An almost identical categorization can be found in the book “The Body Keeps the Score” (van der Kolk, 1994).

Temporality and Event Related Potentials #

As the scholar Nancy Kanweiser put it, in a 2018 lecture on brain-imaging techniques: “EEG is like sticking a microphone over a football stadium”. She used this analogy to show that while a tremendous amount of things might happen simultaneusly in a brain (as in a stadium), with EEG we can only pick up noisy events, like “goals” and “kick offs”.

Yet, EEG picks up brain events within only fractions of a second, making it an excelent technique to analyze brain activity caused by external stimuli, safely assuming causation through closeness in time and repeatability (Stevens & Zabelina, 2019). Specifically, the well-studied field of Event Related Potentials (ERPs) focuses on signals generated by brains after being exposed to specific kinds of stimuli. The most famous such phenomenon is the Oddball Effect (Squires et al., 1975). A Positive Potential emerging around 300 miliseconds (P300) after any stimulus not following a previously established pattern.

Positron Emission Imaging (PET) #

Background on Nuclear Physics #

While PET was invented in the decade of 1960 (Otte & Halsband, 2006), the principles of PET go back as far as the 1910s. At that time, Sir Ernest Rutherford discovered the alpha, beta and gamma radiations of the radioactive materials. Just after James Chadwick discovered the neutron (1932), Rutherford’s model became prevalent, suggesting that the atom consists of protons and neutrons that reside in its center, forming the “nucleus”, and a number of electrons, equal to the number of protons, spinning around that nucleus (Rutherford, 1914).

Rutherford concluded that, the radioactive materials are unstable in nature and need to find a more stable form by shedding protons and neutrons (alpha rays), electrons (beta- rays) and energy (gamma rays), a process called “radioactive decaying”. This process, given time, decomposes the material, as it sheds its particles. A material’s “half-life” is the time needed for a given radioactive material’s mass to get halved by decaying.

Some radioactive materials emit positrons instead of electrons (beta+ rays). It is notable that the positron is the “anti-matter” counterpart of the electron, meaning that they have identical mass, but opposite electrical charge. Finally, when two anti-matter counterparts collide, energy is emitted in the form of light, and when positrons are involved, the phenomenon is called “positron annihilation”. The emitted light is exactly what is detected by a PET scanner (Phelps et al., 2000).

What a scanner does #

For a PET scanner to work, radioactive material is needed to exist in the brain. Such materials can be created by adding neutrons to atoms that can naturally be found in the human body, creating radioactive “isotopes” of them (Otte & Halsband, 2006). These isotopes are then used to create molecules that can be inserted in the human body and are carried in the blood flow or used in metabolism. This process is called “radiolabeling” and the materials are called “tracers”. An example is Oxygen, and its 15O isotope, that has a half-life of 123 seconds (Halsband et al., 2002) that is routinely used to “label” water (H2O) and carbon dioxide (CO2). Another isotope used to label glucose that also has a role in metabolism, is Fluorine 18F (Otte & Halsband, 2006).

Finally, the scanner is a device surrounding the head, detecting photons created by the positron annihilation. The assumption that PET is based on is that position annihilation is common where energy metabolism takes place in the brain, as the labeled materials tend to concentrate there (Phelps et al., 2000). Given, that the created photons take opposite directions along a “coincidence line” and have a known speed (the speed of light), the machine can calculate where in the 3D plain the electron-positron collision happened and pinpoint the part of the brain that was metabolicly active at the time (Phelps et al., 2000), with a spatial accuracy as precise as 4mm X 4mm (Otte & Halsband, 2006).

Experiments using PET #

PET has been used for measuring neuroactivity in cognitive tasks, as it can directly measure the regional Cerebral Blood Flow (rCBF) by using labeled water (15O-H-O) (Otte & Halsband, 2006). One such study was performed by Halsband, et al. in 2002, examining the retrieval of word pairs in bilingual subjects. The study replicated differential activations of areas such as Broca’s area, when a subject was recalling from memory word pairs in different languages. The spatial accuracy of the study was 8mm X 8mm, which is low, even for the standards of 2002, as better accuracy could be achieved with fMRI.

Functional Magnetic Resonance Imaging (fMRI) #

Background on Chemistry and Magnetic Properties of Blood #

Hemoglobin is the blood ingredient that delivers oxygen to tissues. It does so by binding to oxygen through its molecule’s single iron atom, using up all its electrons (Mier & Mier, 2015). The lesser spare electrons a molecule has, the more “diamagnetic” it becomes, responding less to magnetism. In the other hand, when hemoglobin is not carrying oxygen, its iron atom has 4 spare electrons, rendering it “paramagnetic”, hence susceptible to magnetism and capable of influencing the magnetic field around it (Otte & Halsband, 2006).

What a scanner does #

An fMRI scanner is a chamber equipped with rotating magnets and coils, exerting magnetic fields of 1,5 to 7 Teslas (Roalf & Gur, 2017) on the subject. These magnetic fields are used to measure the blood oxygenation level-dependent (BOLD) differences in time inside the subject’s head, effectively pinpointing the space in the 3D plain, where oxygen is needed by metabolically active neurons (Otte & Halsband, 2006). Specifically, the alterations to the magnetic field, caused by the change of diamagnetic hemoglobin molecules to paramagnetic and back again, are what is picked up by the scanner.

fMRI has a spatial resolution of even 1mm X 1mm and higher temporal resolution than PET (Roalf & Gur, 2017). Also, unlike PET, measurements are relative, made in comparison with a reference point (similar to EEG). Additionally, as the changes in the magnetic field caused by hemoglobin alterations are incredibly small to be statistically significant, a statistical model needs to be deployed in order to predict statistical significance in repeated tasks performed by the subject (Albert Einstein College of Medicine, 2014).

Experiments using fMRI #

fMRI has taken by storm the brain-imaging research. Yet, its data processing complex needs, in order to produce valuable results, have been raised as a prevalent disadvantage. Specifically, in 2009, Bennet et al. showed that it was possible to extract false results from an fMRI study that has not undergone proper statistical corrections, proving that a dead salmon fish had neural activations.

Conclusion #

This optimistically scoped essay aims to cover several fields of science that make brain-imaging possible. Providing this information to neuroscientists can create better understanding of their methods and even lead them to question their tools as the products of science that they are, and as such, are prone to errors.

References #