Brain Imaging

MRI studies brain antomy
fMRI studies brain function

BOLD (blood oxygen level dependent signal)
More neural activity = more energy needs = more O2 needed = more blood flow = higher fMRI signal

Metabolic demands of neurons
Not neural activity
Neural activity is infered

+ High spatial resolution, very precise (although low spatial resolution in comparison to invasive techniques used in non-human animals)
+ Non invasive
+ Can look at large structure of the brain
+ Whole brain coverage to uncover networks in health and disease
- Low temporal resolution, 2-6 seconds (a lot slower than the brain works), not the technique which is slow, but the variable which is being measured
- Indirect measure of neural activity
- Correlational not causal relationship between BOLD and behaviour

Slower neuron signal that reflects post-synaptic potentials, and reflects info that comes from nearby neurons (population activity)

Reflects AP
Similar to what most electrophysiology measures

Logothetis et al 2001
Combined BOLD fMRI and electrophysiological recordings
BOLD correlates with spike output (exact info the neuron is currently processing)
BUT correlates better with local field potentials (LFP) compared to multi-unit activity (MUA)

AP
Excitatory/inhibitry input to an area
Inputs from same layer
Feedforward/feedback projections
Small change in neural activity of many neurons may show larger activation differences than large change in only a few neurons
Ergo, fMRI results may not always match with neurophysiological results

Protons in body line up with magnetic field and spin
Protons in deoxy blood not as well organised with their spinning
When brain area is active = AP = increased blood flow and O2 = displacement of deoxyHb from capillaries, venules, small veins
MRI scanner measure difference in proton spinning bewteen oxy and deoxy blood

Diffusion tensor imaging (DTI)
Electroencephalogram (EEG)
Magnetoencephalography (MEG)
Brain lesions
Transcranial magnetic stimulation (TMS)

Measures anatomical connectivity between brain regions
Provides info about location and orientation of white matter tracts based on magnitude and direction of molecular diffusion
No restrictions = diffuse equally in all directions (isotropic)
Restrictions = diffusion may occur along one axis (anisotropic)

Measures electrical potentials of neurons through electrodes placed on scalp

Evoked = field potentials that occur in response to sensory stimulus
Event-related = field potentials that occur in response to perceptual/cognitive event

+ Non-invasive
+ High temporal resolution
+ Allows study of interactions between cortical circuits
- Low spatial resolution
- Correlational not causal relationship between potentials and behaviour

Measures very small changes in magnetic field caused by electrical activity using superconducting coils

+ Non-invasive
+ High temporal resolution
+ Higher spatial resolution than EEG, as magnetic fields are unimpeded by high resistance of skull
- Inverse problem: which neural population is related to the behaviour?
- Only detects current tangential to the surface of the skull
- Correlational not causal relationship between potentials and behaviour

Provides info about necessity of brain region for a particular function
When a process evokes fMRI activity in brain region and the process is disrupted when the same brain area is lesioned, potentially conclude that execution of the process relies on that brain region

+ Useful for large scale systems
+ Can be used to investigate plasticity and recovery of function
- Lesions likely to be large and extensive rather than specific to one area
- Large variability across patients

Uses electromagnetic coil close to scalp
Current passes through coil, generating magnetic field in nearby brain tissue producing localised electrical currents
Creates temporary lesions on the brain
Disrupts function of a specific brain region

Determine whether regions activated in fMRI are essential for task performace (cause and effect)
Potential treatment of disorders (eg. increase dopamine levels in frontal cortex to treat depression)

Determining whether activity of a given voxel (region) changes in response to experimental manipulation

Measuring the time course within an active voxel (region) in response to the experimental manipulation

Block design
Slow event-related (ER) design
Rapid counterbalanced ER design
Rapid jittered ER design

Trials from different conditions grouped into separate blocks
Allows fMRI signal to accumulate over trials of the same kind (same condition)
Used when network of areas involved in a task is unknown and whole brain needs to be searched to identify them

+ High detection power
+ Most widely used approach for fMRI studies
- Poor estimation power (only knows what happens in this block, not in every single presentation)
- Subjects get into a mental set for a block, brain may adapt resulting in decreased fMRI signals due to repetition of stimuli presentation
- Very predictable for subject
- Can't look at effects of single events (ie. presentation of single stimuli)

Trials from different conditions randomised across whole experiment

When areas involved in a task have been identified
Want to explore functions of a brain region in more detail

Blank intervals of variable duration introduced to avoid overlapping signals between trials of different conditions cancelling each other out

Control trial history by making it the same for all trial types
Jitter the trials

The peak of 1 and the trough of another can cancel each other out

+ Can have many trial types in an unpredictable order
+ Can do post hoc sorting of trial type (eg. correct vs incorrect, remembered vs forgotten)
+ Better estimation power than block
- Reduced detection compared to block
- Accurate detection requires accurate modelling

Block = don't know where to look in the brain
ER = when looking at how the brain processes things, compares different stimuli, know where it is but want to understand the processes involved

T tests
Correlations, GLM (general linear models)

Compare activation levels between 2 conditions
Uses a time shift to account for haemodynamic lag

Model activation and see whether any areas show a similar pattern

Voxelwise analysis
Region of interest (ROI) analysis

Search whole brain for voxels (regions) that show significantly higher responses to a task/stimulus compaed to control conditions

+Requires no prior knowledge about role of different brain regions, searches whole brain
- Neglects individual differences
- Can result in compromised spatial resolution due to intersubject averaging

Region selected from previous studies using other techniques
fMRI used to only study signals from voxels in this region

+ Reduced multiple comparison problems seen in whole brain analyses (only studies one region), fewer degrees of freedom, fewer t-tests needed
+ Higher statistical power
+ Avoids problems with functional mislocalisation due to intersubject averaging
- Neglects other areas that could also play a fundamental role in the task

Grey = synapses and dendrites
White = axons

Partial volume effects
Head motion
Vessels

Larger voxels = stronger signal to noise ratio (SNR) (high SNR means signal higher than noise, so clearer image), except when activated region doesn't fill entire voxel - do not refer to image
fMRI signal often occurs in grey matter
If voxel includes white matter/fluid/space outside brain, signal is decreased

Scan at higher resolution (smaller voxels)

Results in activations shifting over time
Difficult to align brain images

Large vessels line sulci
Produce BOLD activation further from true site of activation (where has the activity come from?)
Even bigger problem when comparing between pps (may have different vasculature)

Choose appropriate control condition
Avoid statistical errors

Can't compare a task to nothing (eg. blank screen) - activation in test condition could be due to increased attention rather than the stimulus itself
Also need to make sure task difficulty is equal across controls and trials

Occurs when stats tests performed are not independent from the means used to select the brain region
Selected the brain area than did stats test using that brain area

Use independent data from the original data (used to identify the brain areas involved) when performing stats analyses
Allows conclusions to be made about how this area processes info etc.