Resting State & Event-Related fMRI Flashcards
What are the two experimental conditions?
- Block/Epoch Design
2. Intermixed/Event-Related Design
What is the relevance of event-related fMRI?
Post hoc categorization of trial-types according to the subject’s behavioure.g. recognition memory task
Modelling events whose occurrence is beyondexperimental control, such as those that canonly be indicated by the subjecte.g. perceptual transitions in the vase-faceillusion
Some trials cannot be blockede.g. “Oddball” paradigms, epileptic events (on EEG)
What are the disadvantages of event-related designs?
- Less efficient for detecting effects than are block designs
- Some psychological processes may be better blocked e.g. task-switching, attentional instructions
What are advantages of event-related fMRI
• Randomised trials order
• Post-hoc subjective classification of trials
• Some events can only be indicated by participants
- Some events cannot be blocked due to stimulus context
- More accurate model even for epoch/block designs
Example of Event-related Resting-state FMRI: BOLD correlates of the Alpha rhythm:
- Put electrodes on the back of the head
- Ask subjects to close eyes – block alpha
- Record EEG and calculate the power in the alpha band – fluctuate naturally
- Convolute with the canonical HRF and get the desired post-hoc design
What part of the brain have this kind of blood flow variations?
- predicting based on the ongoing EEG
- Put in GLM and SPM analyses
- region in the brain in which signal correlates strongly to predictor
BOLD of Alpha rhythm (continued):
- The brain regions were more active when the alpha power was reduced – get a negative correlation
- EEG will tell you about the different waves of the scalp, but it is not easy to infer from that which part of the brain will be more active
What is alternative design for BOLD of Alpha rhythm?
- Berger effect
2. Alternating blocks of eyes closed and open
What is the spike-triggered FMRI?
- Patient with epilepsy
- 43 spikes on EEG during FMRI scan
- Record EEG during FMRI and look back at EEG recorded in the scanner
- Identify which scan occurred during the spike and which scan occurred when there was no spike
- Build regressors/predictors based on that
- Classify scans as being in the resting or active state
- If there is an epileptic spike at time T=0 – which scans do you expect to be different from the baseline
The BOLD Impulse Response
• Represents the BOLD response to a brief stimulus
• Is a function of blood oxygenation, flow, volume?
• Its typical features are:
- Peak at 4-6s post-stimulus
- Undershoot at ~15s post-stimulus
- Return to baseline after 20-30s
• Its common mathematical representation is the Canonical hemodynamic response function (Canonical HRF)
• It varies across:
- Brain regions
- Individuals
What is a good experimental design?
A design that gives you results that are straight-forwaed to interpret
What can designs be?
- Blocked or intermixed
2. But models for blocked designs can be epoch - or event-related
What are epochs?
Periods of sustained stimulation (e.g. box-car functions)
- Events are impulses (delta-function)
What can near-identical regressors be created by?
- Sustained epochs
2. Rapid series of events
What are all conditions specified in SPM9?
- Onset
- Durations
Epochs = variable or constant duration
Events = Zero duration
What can blocks of trials be modelled as?
Boxcars or runs of events
Interpretation of the parameter estimate may differ
What does epoch model estimate?
Parameter that increases with rate because the parameter reflects response per block
What does event model estimate?
Parameter that decreases with rate because the parameter reflects responses per word
What is the BOLD impulse response?
- Function of blood oxygenation, flow, volume
- Peak (max. oxygenation) 4-6 seconds post-stimulus: baseline after 20-30s
- Initial undershoot can be observed
- Similar across V1, A1, S1
- But possible differences across:
- Other regions
- Individuals
What is a typical haemodyanmic response observed?
Primary sensory areas such as visual cortex/ auditory cortex
What can the shape of the BOLD impulse response impose?
Constraints on design efficiency
What is the linear convolution model?
The predicted fMRI series is obtained by convolving a neural function (e.g. stimulus function) with an assumed IR
What is the basic idea behind maximising efficiency for linear convolution model?
Maximise energy of the predicted fMRI timeseries
What is the benefit of maximising the variability of signal?
Detect the signal in the presence of background noise
What is the temporal basis function?
- Fourier Set
- Windowed sines & Cosines
- Any shape (up to frequency limit)
- Inference via F-test - Finite impulse Response
- Mini ‘timebins’ (selective averaging)
- Any shape (up to bin-width)
- Inference via F-test
What is the timing issues: sampling?
Typical TR for 60 slice EPI at 3mm spacing is ~3-4seconds
Sampling at [0,4,8,12] post-stimulus may miss peak signal
What can be done to improve timing issues: sampling?
Higher effective sampling by:
- Asynchrony e.g. SOA=1.5TR
- Random Jitter e.g. SOA= (2+/- 0.5)TR
What is timing issues: slice timing?
The model of the haemodynamic response that is built in the convolution model is of a very high temporal resolution
Every measure that you take of the whole brain takes seconds
in the model of the haemodynamic response - only 1 value is assigned to each of the slices
Which particular time point during volume acqusition is the time point during which i want to sample - Reference slice
What is the slice-timing problem?
- Slices are acquired at different times, yet model is the same for all slices
- Different results (using canonical HRF) for different reference slices
What is the acqusition (partial) solutions?
- Middle slice as reference
2. Short TR
What are the modelling solutions
Temporal interpolation of data: “Slice timing correction”
Use a more general basis set (e.g., with temporal derivatives)
What is the design effficency?
- HRF can be viewed as a filter
- We want to maximise the signal passed by this filter
- Dominant frequency of canonical HRF is ~0.04 Hz
What is the most efficient design?
Sinusoidal modulation of neural activity with period ~24 seconds
(e.g. boxcar with 12s on/ 12s off)
What is block design
Alternate between blocks of two types of stimuli - A rest and a stimulus with a temporal property that has a dominant frequency
What is design efficiency?
A measure of how well an experiment can be expected to reveal the effect of interest
Block designs
Generally efficient but often not appropriate.
Optimal block length 16s(beware of high-pass filter).
Event-related designs:
Efficiency depends on the contrast of interest
With short SOAs the inclusion of ‘null events’ (jittered ITI) can help optimise efficiency across multiple contrasts.
Non-linear effects start to become problematic at SOA<2s