Lecture 1 Part 2: Functional Magnetic Resonance Imaging (fMRI)
Required Reading
Baars, B., & Gage, N. M. (2013). Fundamentals of cognitive neuroscience: a beginner's guide. Academic Press. Chapter 3.
Savoy, R. (2001). History and future direction of human brain mapping and functional neuroimaging. Acta Psychologica, 107, 9-42.
Further Reading
Ward, J. The student's guide to cognitive neuroscience. Chapter 4. Shelfmark: 612.8233 WAR
Web Resources
Animation 2.1: Magnetic Resonance Imaging (MRI) - SLMENS
MRI animations with narration.
fMRI for Newbies
The Essential Question
What do those blobs (in fMRI images) represent?
Historical Images of the Brain and Mind
Brain Structure
Some images are still in use today.
Often interpretive, labor/skill intensive.
Examples:
Christopher Wren (1664)
Brodmann cytoarchitecture maps
Brain Function
Phrenology (19th Century ).
Electrophysiological mapping (1957).
Magnetic Resonance Imaging (MRI) Physics
Uses a magnetic field (B_0) and radio energy to produce an image.
A large magnet (50,000 times Earth's magnetic field) aligns nuclei that have a net magnetic moment (from odd number of protons/neutrons), e.g., H_2O.
Nuclei absorb and re-emit radio frequency energy.
How is an Image Acquired?
Nuclei spin around the main magnetic field.
RF pulse (oscillating magnetic field) tips M out of alignment with B_0 and synchronizes the phase of spins.
M gradually returns to alignment, and spins lose phase coherence.
These changes are detected as the ‘MRI signal’.
Spins oriented with or against B_0.
M = net magnetization.
M deflected by RF pulse.
M Applied Magnetic Field (B_0).
Mapping Lesions
A new extension of an older technique for mapping structure to function.
Images make complex patterns of data intelligible.
Imaging the Structure of the Brain.
Mapping Normal Variability in Structure
Map second-order structure, such as sulcal asymmetry.
Color-coded maps for statistical significance of gyral/sulcal variability.
Large numbers of neurologically-normal participants.
Thompson, P. M. et al. (1998). Cereb. Cortex, 8, 492–509.
Mapping Change Over Time
Examples:
Grey matter volume.
Commissural myelination.
Myelination by age (months).
Functional MRI (fMRI)
Brain Mapping Publications exhibit an exponential rise.
The Blood Oxygen Level Dependent (BOLD) Response – the basis of fMRI
Relative levels of de/oxyhemoglobin change from regional cortical activity.
Momentary decrease in blood oxygenation immediately after neural activity increases, known as the “initial dip” in the hemodynamic response function (HRF).
Followed by a period where the blood flow increases to a level which overcompensates for the increased demand.
Regional blood oxygenation actually increases following neural activation.
The blood flow peaks after around 6 seconds and then falls back to baseline, often accompanied by a “post-stimulus undershoot”.
Importance of Hemoglobin
De/oxyhemoglobin have different magnetic properties.
Local field strength is affected by relative levels.
This affects the local signal in the image.
Increased signal is obtained from ‘active’ regions.
fMRI Experiment Components
Magnet
Video projector
Video screen
Gradient coil
Radio-frequency coil
Prism glasses
Headphones
Radio-frequency amplifier
Button response box
Amplifiers control magnetic field in three dimensions
Stimulus control computer
Spectrometer control computer
fMRI – Experimental Logic
Cognitive subtraction originated with reaction time experiments (F. C. Donders).
Measure the time for a process to occur by comparing two reaction times, one which has the same components as the other + the process of interest.
Assumption of pure insertion: Can insert a component process into a task without disrupting the other components.
BUT task difficulty (a confound) may differ between conditions, resulting in changed attention.
Examples:
T1: Hit a button when you see a light.
T2: Hit a button when the light is green but not red.
T3: Hit the left button when the light is green and the right button when the light is red.
T2 – T1 = time to make a discrimination between light color.
T3 – T2 = time to make a decision.
Experimental Design 1: Block Design fMRI
Alternating blocks of task and rest.
Experimental Design 2: Event-Related fMRI
Allows (pseudo)random presentation of stimuli.
Allows retrospective coding of events (e.g., memory experiments).
fMRI Preprocessing and Analysis
Steps:
Realignment & motion correction.
Smoothing.
Normalization.
Image Data.
Anatomical Reference.
Parameter Estimation.
Kernel Data.
Design Matrix.
Statistical Parametric Mapping.
Contrasts.
Analysis - Modelling the data fMRI signal (data)
Block model convolved with HRF.
We regress our model of our experiment against the signal-change data.
Least-squares best-fit analysis of the data that best estimates the amplitudes of the predictors (minimize residuals).
A T-test tests whether the slope of the function differs from 0 (i.e., flat).
Voxel-wise Analysis
This analysis is done independently at every voxel (3D pixel/volume) - i.e., thousands of times!
Contrasts allow one to test for voxels where activation in one condition is greater than another.
Voxels with significant T statistics can then be colored in according to the size of T.
Interpreting Blobs
Blobs are clusters of significant statistics for either a main effect or a contrast between two sets of regressors at each voxel.
Shows areas where the signal change was significantly predicted by the model (or where the degree of prediction differed between contrasted conditions).
Importantly, this is the end result after much preprocessing and analysis.
Change in the signal is due to regional hemodynamics.
Thus, activations are distantly related to the underlying neurological events.
What has functional brain imaging told us?
Identified functional areas – e.g., Fusiform face area – stimuli; Anterior Cingulate - attention.
Corroborated findings from other methods (e.g., hippocampal involvement in memory).
Allowed the localization of function from undamaged brains.
Meta-analyses bring some order to the flood of data, but are these maps from multiple different studies any more useful than the 1957 electrophysiology map?
Cabeza & Nyberg. (2001). Journal of Cognitive Neuroscience 12:1, pp. 1–47.
New Directions
Functional-connectivity analyses: calculate correlations between activations in different areas.
Dynamic causal modelling: explicit models of distributed networks are tested to see which best fits the observed data.
Both these techniques investigate distributed processing and overcome some of the limitations of lesion studies and earlier fMRI studies.