Functional Magnetic Resonance Imaging (fMRI)
Introduction
Structural imaging reveals the static physical characteristics of the brain. It makes it useful in diagnosing disease. Functional imaging reveals dynamic changes in brain physiology that might correlate with cognitive functioning, for example. Neural activity consumes oxygen from the blood. This triggers an increase in blood flow to that region and a change for deoxyhemoglobin in that region. As the brain is always physiologically active, functional imaging needs to measure relative changes in physiological activity.
The most basic experimental design in functional imaging research is to subtract the activity in each part of the brain whilst doing one task away from the activity in each part of the brain whilst doing a slightly unfamiliar task. We call this cognitive subtraction.
Other methods, including parametric and factorial designs, can minimize many of the problems associated with cognitive subtraction. There is no foolproof way of mapping a point on one brain onto the putatively same point on another brain because of individual differences in structural and functional anatomy.
Current imaging methods cope with this problem by mapping individual data onto a common standard brain (that is, stereotactic normalization) and by diffusing regions of significance (a process we call smoothing).
A region of activity refers to a local increase in metabolism in the experimental task compared to the baseline, but it does not mean that the region is essential for performing the task. Lesion studies might provide evidence concerning the necessity of a region for a task.
Functional imaging can make crude discriminations about what someone is thinking and feeling and might outperform the traditional lie detectors. However, it is highly unlikely that they will ever be able to produce detailed accounts of another person’s thoughts or memories.
An fMRI measures regional cerebral blood flow. Cognitive functions are region-specific, if a task involves a certain cognitive function, the areas involved will become more active, need more oxygen and more blood. fMRI measures regional levels of blood oxygen by detecting magnetic changes in red blood cells when they become deoxygenated
Magnetic Resonance Imaging
Magnetic resonance imaging creates images of soft tissue in the body, which x-rays pass through undistorted (so computerized tomography would not capture well). The density/intensity of the images is water-based, with different amounts of water for different tissues. It enables a 3D image of the layout of these tissues. Structural MRI produces a static image of the brain structure. It has a high spatial resolution. It is used to overlap functional images on to.
Metal items and MRI
We remove metal items before the functional imaging because the strong electromagnetic fields will attract them. Patients who use pacemakers cannot have magnetic resonance imaging or its functional variant.
Underlying Mechanisms of fMRI
The fMRI uses a strong magnetic field to line up protons. It measures oxygenated blood by recording the spin of protons, which have a magnetic charge. After aligning protons in fMRI it sends a radio pulse through the lined-up protons, to record how they resonate.
Different proton resonance patterns.
Different protons (different tissues) resonate differently (magnetic susceptibility), allowing the composition of a tissue image. fMRI uses the differential response of oxygenated and deoxygenated blood for the imaging. Oxygenated blood resonates differently from deoxygenated blood, allowing the composition of an (indirect) image of the brain activity.
It is T1-contrast (measures a different magnetic property to functional scans).
The spatial resolution of fMRI
Although fMRI is not as spatially resolute as MRI, it can record 3x3x3 mm and more detail with a 7T (stronger tesla coil strength) scanner.
Both spatial and temporal T2 contrast rely on tesla strength. Temporal T2 contrast measures a different magnetic property to structural scans.
In structural MRI, the magnetic field aligns protons. It aligns protons in water molecules that have weak magnetic fields, initially randomly oriented, but some align with the external field. A radio pulse knocks orientation by 90 degrees, which leads to a change in the magnetic field. After this change in the magnetic field, the protons become stead and we can repeat the procedure for fresh slices of the brain. A whole-brain image in 2 seconds (3 mm slices) 1: relaxation time. T1-images structural scans.
It relies on the brain to store a large amount of oxygen and glucose. It does not store oxygen though still consumes around 20% of the body’s oxygen supplies. The brain tissue does not store oxygen; oxygen must be supplied from the fresh blood supply. Active tissue consumes more oxygen compared to less active brain tissue. Oxygen-rich blood is lost in areas of higher brain activity.
Magnetic properties of blood
Oxyhemoglobin is diamagnetic while deoxyhemoglobin is paramagnetic. Hemoglobin molecules resonate differently in these different magnetic states.
Diamagnetic substance
A diamagnetic is magnetic when exposed to the external magnetic field for example oxyhemoglobin.
Paramagnetic substance
A paramagnetic substance is normally magnetic for example deoxyhemoglobin.
Blood Oxygenation Level Dependent (BOLD) Signal?
It compares the level of oxygenated with deoxygenated blood derived from the magnetic properties of blood. It is an indirect measure of brain activity.
Factors on while BOLD Signal depends:
1) Cerebral metabolic rate of oxygen (goes up when tissue is active *of genuine interest* more oxygen when spending energy, so de-oxygen goes down)
2) Cerebral blood flow
3) Cerebral blood volume
fMRI compares the differences between magnetic spins of protons in oxygenated blood and deoxygenated
Hemodynamic Response Function.
Initial Dip
Neurons consume oxygen leading to a small rise for deoxyhemoglobin causing reduction of BOLD signal.
Overcompensation
In response to the increased consumption of oxygen, blood flow to the region increases. Increased blood flow is greater than increased consumption >> BOLD signal increased
Undershoot
Blood flow and oxygen consumption dip before returning to original levels. This may reflect a relaxation of the venous system.
Active' areas
Active areas in fMRI refer to a physiological response that is greater relative to some other conditions. To label active areas, we need a baseline response, well-matched to the experimental task. Example: Petersen, Fox, Posner, Mintun, and Raichle (1988) Study brain activity involved in word recognition, phonology, and retrieval of word meaning, cognitive subtraction.
Research designs can exploit this difference by finding two tasks, an experimental task and a baseline task, which differ in terms of a few cognitive components.
Subtraction design
Subtraction is taking a task with the cognitive component in it, and then subtract another task with only that component is taken out
Neuronal structures underlying a single process P
Contrast: [Task with P] [control task without P].
Conjunction
Conjunction requires a set of orthogonal tasks that has a particular component in common. Look for regions of activation that are shared across several subtractions. A test for such activation common to several independent contrasts is called a conjunction. It resembles a factorial design in ANOVA.
Issues with subtraction design
1) The assumption of pure insertion is the assumption that we can insert a single cognitive process into another set of cognitive processes without affecting the functioning of the rest.
2) At baseline the brain is always active, and the level of activity is not consistent which makes it challenge where to make comparisons.
Donders coined the term pure insertion as a criticism of reaction time methods. One way to minimize the baseline/pure insertion problem is to isolate the same process by two or more separate comparisons and inspect the resulting simple effects for commonalities.
Example of this cognitive subtraction in Petersen, Posner 1998
Brain activity involved in word recognition, phonology and retrieval of word meaning cognitive subtraction e.g. contrasts passive viewing of (words vs fixation cross) e.g. (read aloud word vs look at the word) e.g. generate (a word associated with viewed word vs read aloud a written word)
The issue with pure insertion is that adding an extra component does not affect the operation of earlier ones in the sequence. BUT: interactions are likely to occur– Baseline task: should be as like the experimental task as possible.
Examples of conjunctions and factorial designs by Frith:
1) Why cannot we tickle ourselves (Blakemore, Rees, and Frith,1998).
2) Factors touch (felt/not) self-movements (moved/not)
Parametric fMRI design
To get around baseline continuous manipulation of the factor of interest. We treat the variable of interest as a continuous dimension rather than a categorical distinction. Associations between brain activity rather than differences between two or more conditions. passive listening to spoken words at six different rates. Different brain regions show different response profiles to different rates of word presentation. Adapted from Price et al. (1992), and Friston (1997). no baseline necessary.
Functional specialization
Functional specialization: region responds to a limited range of stimuli/conditions. This distinguishes it from the responsiveness of other neighboring regions (no localization).
Functional integration
How different regions communicate with each other. It models how activity in different regions is interdependent. Effective connectivity or functional connectivity between regions when performing a task. Use techniques like the principal component analysis.
Example:
A word production vs repeating letters in patients with schizophrenia and controls.
Block design
In a block design, stimuli in one condition are grouped. Strong BOLD contrast: higher signal-to-noise ratio simple design and analysis - practice/fatigue effects cannot be used when participants should not know which condition is coming next.
Event-related design
When stimuli are presented completely randomly, we call it event-related (new as temporal difficulties, etc.) design. This design works with infrequent and random stimuli. If conditions defined by the participant-sorting what happened in a trial (e.g. correct/incorrect trials; biostable percept (Necker cube); the presence of a hallucination - see right). Different stimuli or conditions are interspersed with each other (e fMRI). Intermingled conditions are subsequently separated for analysis. no practice/fatigue effects can be used when participants should not know which condition is coming next: randomization can be used when trials can only be classified after the experiment- weaker BOLD contrast: lower signal-to-noise ratio more complex design and analysis
Session
a scanning session, all the data collected from a participant. Usually comprises a structural scan and several runs of functional scans.
Run
A continuous period of scanning consists of a specified number of volumes
Volume
A Set of slices taken in succession: a 3D spatial image, with a temporal dimension. Expressed in TR (Repetition Time): how long does it take to acquire a volume.
Epoch
A period when a certain condition is presented. Conditions (epochs) can be grouped (blocked design) or randomly intermixed (event-related design).
Correcting for head movements:
Spatial resolution >> small spatial distortions–Individual differences in brain size and shape stereotactic normalization (adjust the measurement of overall dimensions to the 'standard brain'– Individual head aligned differently in scanner over time due to movements. Regions are harder to detect False-positive results. Physically restraining head (using foam or something) and participant instructions Correction
Smoothing
Spreads some raw activation level of a voxel to neighboring voxels. Smoothing enhances signal-to-noise ratio Compensates for individual differences in anatomy.
Assumption
Smoothing assumes that Cognition does not occur in single voxels. Increases the spatial extent of the active region. more likely to find overlap between participants
Steps of fMRI Analysis
Individual differences “averaging over many participants–Correction for head movement– Stereotactic normalization–Smoothing–Statistical comparison
Stereotactic normalization
Mapping regions on each brain onto a standard brain (brain template is squashed or stretched until it fits). Tailarach and Tournoux (1988).
Brain Atlas (based on one brain), Tailarach coordinates–X left/right–Y-front/back–Z top/bottom.
Alternative: Montreal Neurological Institute (average of 305 brains)—Voxels (volume elements), 3-D coordinates.
Tens of thousands of voxels “capitalization on chance Lower significance level (Bonferroni). Choosing a statistical threshold based on spatial smoothness (random field theory).
Analyze the pre-determined region. Reported, corrected, or uncorrected statistical parameters (ROI?)
We start a stat comparison by dividing up data according to design-then perform stat comparison.
Three points of interpretation:
1) Inhibition versus excitation
2) Activation versus deactivation
3) Necessity versus sufficiency.
Inhibition versus excitation?
Functional imaging signals are assumed to be related to the metabolic activity of neurons, and synapses. However: activity can be excitatory or inhibitory. The BOLD signal is more sensitive to neuronal input into a region than the output from the region. Unclear whether functional imaging can distinguish between two neural functions.
Activation versus deactivation
Activation/deactivation Merely refers to the difference between the two conditions. Does not say anything about the direction of the difference.
Necessity versus sufficiency
Necessity: Are active regions critical to the task? Sufficiency: functional imaging shows us active regions, but these may not be crucial. Use methods in conjunction with other methods.