Human Neuroscience Methods

Goals

• To provide an overview of the qualities of different methods
• To provide a brief introduction to non-invasive brain stimulation
• To provide a brief introduction to brain electrical recording
• To provide a brief introduction to brain imaging.

Topic Slide

Oersted (1820) discovered that a compass needle was deflected by a current running in a wire. This observation was later investigated by Michael Faraday and the laws governing electrical and magnetic fields were established by James Clerk Maxwell. However, that issue is the province of a physics class. Why does it interest us here? One is simply to establish that charges that move in space create both electrical and magnetic fields, something we will exploit. The other is for the practical technical aspect. This is a picture of a Thomson galvanometer. You can lead its input to the brain and see if anything passes through the wire that causes the “compass” needle to deflect. As we will learn, the brain activity recorded was so feeble that it needed amplification. Here, the amplification comes from illuminating the needle and watching its shadow cast on the measuring tape.

Reading

There are no readings for this supplemental lecture.

Overview

We can categorize methods in human neuroscience along several dimensions.

Temporal resolution

The temporal resolution of a method determines how quickly a method can detect a change in the physiological variable it is measuring. Some methods with high temporal resolution, like the EEG, can measure changes in neuronal brain activity at less than a millisecond resolution. Methods with low temporal resolution, like Positron Emission Tomography (PET) can measure changes in glucose utilization over an interval of about 40 minutes. High temporal resolution is generally desirable, but it depends upon the variable being measured. Respiration, for example, is a relatively slow phenomenon and so it would not be necessary to sample it at a millisecond temporal resolution to capture meaningful changes.

Spatial resolution

The spatial resolution of a method determines how changes can be detected over space. Structural MRI, for example, has relatively high spatial resolution and can image the brain’s structure at less than 1 millimeter resolution. EEG, on the other hand, has relatively low spatial resolution and (when measured with scalp electrodes) is limited to 10s of centimeters resolution. Like with temporal resolution, high spatial resolution is generally desirable, but depends upon the variable being measured.

Correlational or Causal

Some methods permit one to make observations of some brain variable, either in a resting state, during the performance of a task, or during the presentation of a stimulus (such as a sound or a visual object). The changes in the brain variable can be correlated with whatever events are occurring in the subject’s environment. These correlations are then the basis for inferences drawn about changes observed in the brain variable and the events in the environment. For example, if the subject is presented with a face as a visual stimulus, and if a method such as functional MRI detects localized changes in blood oxygenation in a region of the fusiform gyrus (the fusiform face area) during the face presentation, then one might infer that the fusiform face area was responsible for processing the face. While this may be a reasonable inference, it would not be a strong inference because it is correlational. A researcher would need additional data, for example, demonstrating that the surgical removal of the fusiform face area left a subject unable to process faces, to make the stronger claim that this region of the brain was necessary for face processing. Many of the techniques used in human neuroscience are observational or correlational and so suffer from weak inference.

Methods that permit stronger causal inferences are those that manipulate the brain’s activity and observe changes in the subject’s ability to perform a task. In non-human animal research, a typical method for making causal inferences would be to create a surgical lesion, in which part of the brain is removed or destroyed, and to see if task performance is affected. For obvious ethical reasons, experimental lesions cannot be made in humans for scientific studies of this kind. However, investigators of human brain-behavior relationships can take advantage of medically-mandated surgical excisions – for example, by studying patients in whom the hippocampus was removed during surgery to relieve epileptic seizures. Scientist can also take advantage of naturally occurring brain lesions, such as those caused by neurological strokes or brain trauma.

Some methods, such as direct electrical stimulation or transcranial magnetic stimulation (TMS), can reversibly affect brain activity for a short duration (seconds to minutes). In a general sense, these methods can thus be used to generate a temporary ‘functional lesion’, where the stimulated brain region is activated (resulting in behavior) or deactivated (resulting in an inability to perform its normal function). Ethics review boards for human experimentation have concluded on the basis of the extant literature that these stimulation techniques are safe within proscribed limits, and so research using stimulation techniques has become common.

Necessary and sufficient conditions and causality

One frequently encounters the concept of necessary and sufficient conditions for a particular outcome. Although taken from the realm of logic, necessary and sufficient conditions are frequently discussed in the neuroscience literature in the context of causal inference. There is some controversy about the use of these terms in biology, and particularly in genetics and neuroscience.

We have discussed necessary conditions already with respect to removal or disabling of a brain region. For example, if we surgical remove the fusiform area that is activated by face processing, and the individual can no longer identify faces, then that fusiform area is necessary for face processing. This is pretty straightforward.

More problematic is the concept of sufficiency. Is having an intact fusiform face area sufficient for the ability to recognize faces? At one level, this question is absurd. The fusiform area is dependent upon visual input. If visual cortex is damaged, an intact fusiform face area will receive no visual input to process. So the strict answer to such a question will always be no.

But I think there is a way in which the concept of sufficiency is relevant to neuroscience. We know that there are other brain structures involved in face recognition besides the fusiform , and they form connected network of face processing brain regions. Some of these regions appear to code different aspects of face processing – such as facial expression in the posterior superior temporal sulcus (pSTS) and biographical person-knowledge associated with the face in the ventral anterior temporal lobe (vATL). It seems quite reasonable to me to employ the sufficiency criteria when constrained to this network. For example, we can ask if the fusiform face area is sufficient for face recognition when the vATL has been surgically removed.

Brain stimulation

In my lecture of Functional Localization in the Brain, I discussed the use of electricity to stimulate the exposed brain surface in humans and experimental animals. This included the work of Fritsch and Hitzig, David Ferrier, and neurosurgeon Wilder Penfield. Non-physiological electrical stimulation can depolarize neurons in the brain. This can evoke can evoke movements, sensations, and other phenomena. However, because the electrical stimulation is not organized as neural input/output would be, it can also disrupt or prevent the normal function of a brain area. Consider a brain area involved in face perception. Stimulation could evoke a hallucination of a face. However, it could also make it difficult or impossible for a subject to recognize a face. The latter phenomena would not be noted by an experimenter unless the experimenter was testing faces during stimulation.

Transcranial magnetic stimulation (TMS)

In our discussion of electrical stimulation, current was applied to the exposed brain. Can electrical stimulation be applied through the skull? The skull is not a good conductor of electricity. However, with a sufficient potential difference, electrical current can be passed through the skull and, consequently, through the brain. For example, if a cathode is placed on the top of the head, and an anode is place in the mouth against the upper palate, an electrical current will run through the skull and brain between these electrodes. This method is used today – particularly with respect to electroconvulsive shock therapy (ECT). However, because the skull is a poor conductor, much of the electric current will run around the skull through the muscles, which can be painful.

A less noxious method for stimulating the brain through the skull is to induce an electric current in the brain using a magnetic field. This is the basis of Transcranial Magnetic Stimulation (TMS). In TMS a coil outside of skull is rapidly charged with electricity. This creates an oscillating magnetic field that easily penetrates the skull, because magnetic fields are not influenced by the highly resistive skull. An electric field is induced in the brain by the oscillating magnetic field, and this electric field stimulates the neurons.

TMS is used in many psychology labs as a method for interrupting or evoking sensations, motor activity, or cognitive activity. The following YouTube videos provide examples of TMS.

Here is one example

Here is another.

TMS is also being increasingly used in therapy – particularly to treat depression. In this sense, it is used as a less noxious version of electroconvulsive therapy.

Compared to direct stimulation of the exposed brain, there is less certainty about where, exactly, in the brain the neurons are being stimulated.

Electrophysiology

In my lecture about postsynaptic potentials and the action potential, I focused on the current flows occurring within the neuron. However, the movement of charged ions create electrical and magnetic fields that can be recorded throughout the brain, and even penetrate the skull to the scalp. The magnitude of these extracellular potential differences will be small relative to the differences near the active transmembrane channels. Rather than recording a -70 millivolt potential across the membrane, we may only record a 10 microvolt signal at the scalp. Nevertheless, with sufficiently sensitive recording devices and sophisticated signal processing techniques, we can make such measurements.

Recording of action potentials

If a micro electrode is advanced into the brain to a point right next to a neuron, the action potential of that neuron can be recorded as rapid change in potential. On a graph, these very brief potential changes look like spikes, and so scientists often refer to action potentials as 'spikes'. Recording action potentials is sometimes referred to as 'single unit recording'.

Because the micro electrode must be advanced into the brain to record the action potential from a neuron, it is an experimental technique that is not typically performed in humans. However, in some rare instances, it has been performed in humans undergoing evaluation for epilepsy.

With improvements in circuit miniaturization and signal processing, many neurons can be recorded at once from multi-barrel electrodes or embedded chips that contain electrodes and electronics.

Here is a picture of a 32*5 array of electrodes for recording single unit activity from 160 neurons simultaneously in the rat.

Recording of summated postsynaptic potentials

This is not a course in electrophysiology or biophysics, but a little background information about electric fields may help you understand these methods. Two charges of opposite sign separated in space create an electric field. An electric field will influence charges that occur in the space. Field lines extend from the source (anode) to the sink (cathode), with the density of the lines indicating the strength of the field. The density of the lines increase nearer to the source and sink, and fall off with roughly an inverse square with distance.

At 90 degrees to the field lines, we draw equipotential lines. There is no difference in potential (measured in volts, or voltage) between any two points on the same equipotential line. However, there is a potential difference between any points on two different equipotential lines. Electrical current will want to run between points on different equipotential lines.

I used the example of a topographic map in lecture. You can substitute altitude for electric field, and the topographic lines indicating the same altitude (equi-altitude, or contour lines) would substitute for the equipotential lines. Every point on the same contour line is at the same altitude, and water (current) will not run down along the same contour line. However, a stream of water will run between two places on different contour lines (i.e., water will run downhill).

On my topographic map, if I place altimeters on different contour lines, I can measure a difference in altitude. In my electric field map, if I put a sensor on different equipotential lines, I can record a voltage difference. Thus, in measuring postsynaptic potentials, I need to use two electrodes, and make sure that they are on different equipotential lines. The voltage measured between those lines will reflect neuronal activity.

Recall from my lecture of neuronal excitability that a PSP creates an inward flow of positive or negative current (depending upon it is an EPSP or IPSP) through the membrane. However, due to the Law of Conservation of Charge, if we force a current into the neuron, it will be matched by an outward current flow. This outward current flows into the extracellular space, and is associated with potential field lines that occur instantaneously throughout the volume of the brain and extending through the skull to the scalp. As mentioned above, the field is strongest nearest the point in the membrane where depolarization or hyperpolarization occurs.

The electroencephalogram (EEG)

The PSP of an individual neuron is a brief event that occurs against the background of millions of other PSPs occurring throughout the brain. However, the synchronous occurrence of PSPs on many neurons located closely together, and with ideal geometric arrangement, can sum to a large population PSP. This is the basis of so-called extracellular field potentials.

This field will extend through the volume conductor of the brain, and through the skull. The high electrical resistivity of the skull will smooth the field somewhat (which reduces spatial resolution) but the field will still be measurable.

The fields recorded at the scalp are called the EEG, or electroencephalogram. The ‘spontaneous’ EEG is usually described in terms of particular frequency bands, such as delta, theta, alpha, beta, gamma. These fields are sensitive to physiological states, such as sleep, or attentiveness. The EEG was first described by Hans Berger in the 1930s.

In lecture, I gave an example of the sensorimotor, or mu rhythm. The mu rhythm is recorded from motor cortex at rest, and it is suppressed by an individual's movement. However, it is also suppressed when an individual watches another make a movement (mirror neurons).

Evoked potentials (or, Event-related potentials)

Embedded in the spontaneous EEG can be minute signals that are associated with particular events such as the onset of sensory stimuli (a flash of light, a tone pip), a motor act (such a a finger movement), or a perceptual or cognitive act (such as the recognition of a face). We call these signals evoked potentials, or event-related potentials.

EPs are too small to see in the raw EEG record, but can be enhanced through signal averaging. I illustrated signal averaging in lecture.

The magnetoencephalogram (MEG)

A changing electrical current also creates a magnetic field (according to the right hand rule) that can be recorded with an exquisitely sensitive recording device called a super quantum interference device (SQUID). Thus the same PSPs that are manifest in the EEG are also manifest as the magnetoencephalogram, or MEG. Magnetic fields are not distorted by the skull, and so MEG has advantages in spatial resolution over EEG.

In lecture, I provided an example of the MEG ‘alpha’ rhythm.

The inverse problem

Recall that the EEG and MEG record summated PSPs from neurons throughout the brain. If all you have is a sum, you can’t work backwards to what was added together to create that sum – the possibilities are infinite. This inverse problem makes it impossible in principle to determine where in the brain the sinks and sources were located that produced a give potential difference on the skull. This limits the usefulness of EEG and MEG. However, one can use simplifying assumptions to approximate the locations of sinks and sources.

Video on EEG and MEG

The video embedded below was excerpted from the lecture on attention in fall, 2018. It provides a brief overview of methods and then focuses on EEG and MEG. It is the lecture referred to which some of the notes above refer.

Neuroimaging

X-rays and CT scans

The simplest imaging technique is to shine visible light on the object of interest, and observe the shadow it casts. X-rays are like shadows, except using much more energetic waves than visible light. Because some X-rays are absorbed by tissues (like bone) and pass through other soft tissues, the shadow left by X-rays on photographic film has contrast.

X-rays are typically compressed into 2-dimensions. However, we can we make three-dimensional images. In class I used the example of a flashlight illuminating tent wall, while the shadow is observed on the opposite wall of the tent. If we shine light and record shadows along front and side walls, we can localize in three-dimensions

Back-projection

In lecture, I demonstrated the back-projection process. In back projection, a ring of sensors are used to record ‘shadows’ along many different directions or orientations. For example, consider the acquisition of X-ray 'shadows' while the X-ray source is rotated 360 degrees around the sample to be imaged. This is called a CT scan, or Computed Tomography scan. Once the sensor ring records the data, an image is reconstructed by projecting lines from each sensor back into image space, where the thickness or intensity of the line is related to the intensity of the signal recorded at that sensor. The quality of the image improves with more sensors. This is illustrated in this image of a 'dot phantom' reconstructed using back-projection.

Positron Emission Tomography (PET)

A CT scanner can create an image of static anatomy based upon tissue density. How can we make an image of dynamic processes in the brain?

You may be surprised to learn that Connecticut once had a thriving industry in steel production. Where was it located? Let’s answer this question in a manner that illustrates PET imaging. First, let’s think of the fuel needed for steel production, iron ore. Imagine that you stand at the NY-CT border along I95 and put a bar of radioactive uranium in all of the trucks carrying iron ore. In this fanciful example, I95 is acting as the carotid artery and the uranium bar is the radioactive ‘tag’. The uranium tagged-iron ore will eventually converge on the blast furnaces located in the Naugatuck Valley. Now imagine we were to fly an airplane in a grid pattern over the state, and the airplane was equipped with a geiger counter. If we plotted the intensity of radiation on this grid, we would see a radioactive ‘hot spot’ in the Naugatuck Valley. This is the principle behind position emission tomography, or PET.

PET is an acronym for Positron Emission Tomography. In PET, a radionuclide is attached to a molecule that is normally present in the brain. The radionuclide is a positron (a negative electrode – i.e., antimatter) emitting isotope. For example, fluorine-18, is an isotope of fluorine that is a potent source of positrons, and can be attached to the glucose molecule. The tagged molecule FDG (fluoro-deoxy-glucose) will partake in the normal molecular processes that untagged glucose would, but will emit positrons while doing so. In other words, the tagged glucose goes wherever glucose goes in the brain. Because the brain needs glucose to create energy, FDG will accumulate wherever energy is needed. If one can make an image of where the FDG has accumulated, then one knows where energy was being consumed in brain.

How is this done?

• A positron is emitted from the radionuclide (F18 in our example) and encounters an electron (a matter-antimatter collision).
• The collision causes mutual annihilation with the emission of two energetic gamma rays that fly off at near the speed of light 180 degrees apart.
• A ring of crystals is positioned around around brain. The gamma rays encounter these crystals (and photomultiplier tubes and scintillation counters) at opposite side of ring nearly simultaneously (a ‘coincidence’).
• Because the gamma rays flew off at 180 degrees separation, the annihilation must have occurred somewhere along a straight line between the two crystals that record the coincidence.
• Back-projection is used to make an image of the 3D distribution of FDG.

Although I used FDG as an example, there are other radionuclides that emit positrons, and they can be attaches to dozens of biologically active molecules. For example, oxygen-15 is an isotope of oxygen that emits positrons. O-15 can be attached to a water molecule and used to tag the water in blood. With O-15, changes blood flow can be imaged, and changes in blood flow can be determined by comparing two images taken at different times.

Some positron emitting isotopes can be attached to neurotransmitters. For example, it is possible to tag dopamine. Recently, a radionuclide (PIB – Pittsburgh Compound B) has been attached to a molecule that shows the distribution of amyloid buildup in the brains of living individuals with Alzheimer's disease.

Limitations to PET

PET is a powerful technique. However, it has some disadvantages. The principal disadvantage is that it exposes participants to high energy gamma ray radioactivity. Gamma rays can break chemical bonds and damage cells. Thus, we are limited in how often a subject can be tested for non-medical purposes. This rules out longitudinal tests in which an individual is repeated tested at different time points. It is also not ethical to use in developing brains without a strong medical justification.

A second limitation is that the positron ‘range’ limits spatial resolution. In PET, we image the location where the the annihilation of the electron and positron occurred, not where the molecule of interest was located. The positron range can be several millimeters. This has the effect of blurring the image, resulting in lower spatial resolution.

Magnetic Resonance Imaging (MRI)

Magnetic Resonance Imaging provides an alternative to both CT scans for static anatomy and PET scans for imaging dynamic brain changes associated with energy use.

The principles of MR imaging are more complex than PET and so will only be explained in a simplified way here. Students interested in learning more about MRI may consider taking my course Psyc 260, Methods in Human Neuroscience.

MRI uses very strong magnets and low energy radio-frequency waves to make images. MRI depends upon upon the properties of hydrogen nuclei, which are single protons. Protons have charge and angular momentum (spin), thus each is analogous to a little bar magnet spinning in space. Normally, these proton ‘bar magnets’ are randomly oriented, and thus there is no net magnetic field (due to cancelations). However, in strong magnetic field (like in a MRI machine), an excess of protons line up with the field lines of the MRI magnet, and thus have a net magnetic moment.

The protons spin around their axis like a top at a frequency that is dependent upon the strength of the external MRI magnet field. By delivering radio frequency waves at the frequency of the proton’s spin, we can reorient the spin of the proton. When the RF waves are turned off, the protons return (or relax) to their original axis of spin; however, they emit a RF wave back which can be recorded by a coil.

By changing the strength of the magnetic field across the brain (i.e., by imposing a magnetic field gradient), we can change the frequency at which the protons spin in different areas of the brain, and thus the frequency of the RF wave they emit when they relax. This is the trick used in MRI to localize the spins in 3 dimensions, and thus enable the creation of an image

MRI has advantages relative to PET. The RF waves have low energy, and do not break chemical bonds – thus they are safe for repeated imaging. This permits test-retest, the testing of infants and children, and the acquisition of longitudinal studies. MRI also does not suffer from the positron-range effect, and thus has higher spatial resolution. MRI can also be used to make different images sensitive to different biophysical phenomena.

Submodalities of MRI

Structural MRI:

In structural MRI, we image the density of hydrogen nuclei (protons) within the sample. In biological tissues, such as the brain, the hydrogen nuclei are mainly located in water molecules. Because protons are differentially represented in different brain tissues, and because different tissues exert different constraints upon proton relaxation, we can obtain very good contrast between gray matter, white matter, and cerebrospinal fluid, and thus make high quality anatomical images of the brain.

Diffusion MRI:

In diffusion MRI, we image the 3D movement of water molecules between successive scans. Because water diffusion is constrained by the orientation of white matter tracts, diffusion images can provide information about the directions in which white matter tracts distribute in the brain.

Functional MRI

Most relevant to our interests in brain structure-function relationship is functional MRI (fMRI). In fMRI, a series of 3D brain images are acquired on a regular time base (for example, a 3D image every 1-2 seconds) and thus results in a 4D data set of images (3 dimensions of space, one dimension of time).

Mechanism of Functional MRI

The contrast in fMRI is related to hemoglobin molecules

• Hemoglobin molecules carries oxygen to tissue. They are on red blood cells.
• Hemoglobin becomes strongly paramagnetic after it gives up its oxygen to energy hungry neurons.
• Paramagnetic molecules distort the local magnetic field, and thus influence the image made by MRI.
• Because hemoglobin delivers oxygen to active neurons, we can measure the location of the distortions in the magnetic field and infer that is where neuronal activity is occurring.

I provided examples of fMRI in lecture.

• Reorganization of sensory cortex in a young woman with a brain lesion.
• Example of time course of a fMRI signal recorded from motor cortex as a subject alternated between squeezing with his hand and relaxing his hand. A near sinusoidal change in signal intensity was detected by fMRI in his motor cortex contralateral to the hand.

Video

The embedded video below was excerpted from a fall 2017 lecture and focuses on neuroimaging methods.