6.2 fMRI 2 Flashcards
(35 cards)
What is an important step after fitting a model to your fMRI data?
(after convolving + adding linear ramp)
creating a statistic which tells you how well the model fits the data (model fit goodness)
What is the amplitude parameter called in a fMRI model of the fMRI data?
What sort of parameter should it be?
How is it calculated? what model is it in?
-Beta = amplitude parameter
-free parameter so you can adjust it based on the data
-Beta estimates by a computer algorithm (usually a GLM)
What is the residual noise?
residual noise is the calculated error between the model function and the acquired data (at each data point)
What is the equation for the GLM model? What does each variable mean?
What does the GLM show?
y = XBeta + e
y=voxel time series data
X=design matrix (red line)
Beta = regression parameter (eg amplitude)
e = Gaussisan noise/residual noise
shows the pattern that the fMRI data follows
What does the t statistic tell?
How do you calculate it?
-the RATIO of the fitted amplitude to the residual noise
- t stat = Beta/residual noise (e)
What does it tell you about model fit when the t stat is low or high?
What values give a high t stat?
low = bad model fit
high = good model fit
higher the t stat, the better the model fit
high amplitude, low residual error => high t stat, good model fit
What is the H0 null hypothesis for an fMRI experiment?
H0 = there is no activation
What is the t stat formally defined as in an fMRI experiment?
What are each term in the definition mean? 3 terms
the ratio of the departure of the estimated value of a parameter from its hypothesised value to its standard error
hypothesised value = no activation
departure = Beta
standard error = residual noise
Is fMRI data qualitative of quantitative?
What does fMRI data/images rely on to give it meaning
fMRI is qualitative! arbitrary signal changes
the signal contrast generated between different states eg at rest, task, different tasks
What is a block design?
Is the same trial type used in each block? WHY?
What is block design in layman’s terms?
-alternates between rest and task intervals of time called blocks. Task block consists of many closely spaced successive trials (over a short interval of time).
-utilises blocks of identical trial types to establish a task-specific condition.
you have a task block and a rest block with each a length of 20s
What experiment was block fMRI derived from?
PET trial
What is the advantage of using block design for fMRI?
-best design for detecting BOLD signal amplitude differences between states
-fairly robust when there is uncertainty to the timing/shape of haemodynamic response because block duration is usually longer than the haemodynamic response
-most efficient for acquiring more trials in less time than other designs because you don’t have to worry about spacing the individual trials apart to get an estimate of each individual event
you can put a lot of trials into each block eg you dont have to worry about the spacing between each finger tap in a block
Why must you repeat tasks many times in the task block for block design fMRI experiments? eg tap finger repeatedly for 20s
lots of baseline noise and the BOLD effect is only a small percentage change in signal maybe 3% -> repeated action to get a significant result
For fMRI terminology what is a trial?
stimulus presentation followed by a response.
What are the disadvantages of using block design for fMRI?
-stimuli are highly predictable and may affect patient’s response strategy/ no surprise element
-inflexible for more complex tasks: can’t use oddball stimuli within a block as you can’t distinguish between trial types in a block
-does not account for transient responses at the start/end of the task: takes subject a second to process the task
-block trials can also change the psychological process you are interested in
-determining appropriate baseline condition can be challenging - what is the best way to design experiment to show significance?
What are the main types of fMRI experimental design?
block
slow event related
fast event related
mixed design
What is slow event related (ER) design?
a short stimuli separated by an inter-stimulus interval (ISI) which allows allows enough time for the haemodynamic response to fall back to baseline before the next trial
usually about 10-20s
What are the advantages and disadvantages of slow event related design?
-can isolate individual BOLD responses to a single trial because of the long ISI
-more exploratory - good if you dont know what the haemodynamic response will be for your experiment -> you can estimate it from slow event
-it makes experiments longer -> so only used when necessary
What is unique about the shape of the GLM model which fits slow event related data?
once the model is convolved with the haemodynamic response function (HRF) -> the signal always return back to baseline after a stimulus because of the long ISI
What is fast event related design?
What is the advantage of this design?
What are the disadvantages?
-short stimuli presented in a random order separated by short and jittered ISI
-you can put odd ball stimuli, events can be trully randomised
-faster than slow ER
-less sensitive to signal change = 1% whereas block design is 3%
-issues with truly randomising the order of the trials
-more complicated model fit and statistics
What does the GLM model look like for a fast event related study?
some peaks merge together, a mix of ISI so not all peaks go back to baseline, (not usually colour coded as its difficult to clearly separate stimuli in a convolved model
What is the most commonly used experimental design for complicated tasks?
fast event related design
What is mixed design?
builds on block design but with different stimuli within same blocks
What are some examples of good practice in fMRI?
-use appropriate stimulus
-collect as much fMRI data as possible and data from as many participants
-if you can do a block design -> do it to maximise BOLD signal
-get a measure of the subject’s bahaviour in the scanner: are they actually doing the task?
