fMRI Experimental Design Flashcards
What is fMRI measuring?
The BOLD response
What is experimental control?
removing the effect of confounding variables
e.g. testing the effect of a drug: active drug vs. placebo drug.
What are the types of statistical inference?
Between-group comparison
Within-group comparison
Linear association (correlation)
Mixed designs
What is a between-groups comparison?
Two groups of participants are recruited
Independent samples t-test
What is a within-groups comparison?
Same participants are assigned to two conditions i.e. an experimental condition and a control condition
paired-sample t-test
What is a linear association (correlation)?
Two variables that “behave numerically” in a similar way
i.e. the taller we are the bigger our feet our
What are mixed designs?
Two groups and two conditions
i.e. baseline and follow-up (within), active drug or placebo (between)
What is the haemodynamic signal?
Hemodynamic response of firing neurons
Oxyhaemoglobin and deoxyhaemoglobin (oxygen-rich blood and oxygen-poor blood)
Difference in magnetisation
Blood-oxygenation level dependent (BOLD) signal
What are the two types of fMRI experimental designs?
Task-based fMRI- BOLD signal during a certain task
Task-free fMRI- BOLD signal in the absence of a task
What is an issue with task-free fMRI BOLD?
Often called resting-state
Still debatable what people do during resting state
Wondering, thinking causes an uncontrolled situation
It reflects something unique
What are the two design types for task-based fMRI?
Block design
Event-related design
When doing a sub-vocal word reading experiment, what is a better control condition than sub-vocal nonword reading?
Sub-vocal pseudoword reading
Come from real world words, pronouncable too, they follow the orthographic principles and phonological principles of the language you’re studying e.g. english, these words could be a real world but is not within the english lexicon
This makes them a better control condition as it would activate the same part of the neural machinery used in processing language but they are not existing words in the language
What affects the choice of control group?
Effect of disease - patients with disease vs people without disease (matched)
Effect of genotype - carriers of the allele of interest vs non-carriers of the allele of interest (matched)
Symptom correlates - patients with symptom vs patients without symptoms (matched)
What control group would you use to study the correlates of hate?
Everybody selects stimuli of people they personally hate e.g. could be personal or political etc for the first condition
The control group is people they have neutral feelings for
Some of the brain regions they found were related to emotion perception e.g. insula
This paper was successful in eliciting emotional feeling by asking people to provide their own stimuli
What happens in a task-free design?
Participants are asked to lay still and think about “nothing” (and not fall asleep)
- But when we do nothing we normally engage in introspective and abstract forms of congnition
-The brain is not silent- shown to relate to different psychological processess such a theory of mind
This resting state can also be extracted from task designs when we dont give people overt tasks- often used as a baseline condition
How can we elucidate functional connectivity of brain networks?
Idea is that at rest, these areas activate together and everytime we engage in an overt task these areas de-activate together
These regions are then correlated - these patterns of linear association are known as functional connectivity of brain networks
Different parts of the brain that become active or in-active in synchrony is a robust phenomnon that is consistent not just over time but between different people
So we can extract these regions using some form of correlation, and then these can be used to look at connectivity between regions to form some kind of functional network
How do we assess region-based functional connectivity?
Pick a region of interest - often called seed-based analysis
Calculate a mean time-course of the entire area
What are the areas whose activity correlates with the average time-course of this area
Cross-sectional/longitudinal/correlational studies
What is correlation analysis commonly used in?
Structural analysis
Difficult to do correlational analysis in functional analysis across people
What is a problem with correlation analysis?
People mistake correlation with causation so it is often misused, especially by the media
Spurious correlation = two variables statistically correlated but no causal link
-often caused by a confounding variable
What is an example of a mixed design?
Juggling
Experimental group - baseline scan - 3 months of juggling- retest scan
Control group- baseline scan- 3 months of inactivity- retest scan
This control means you can do a mixed anova to see the effects of practice by comparing baseline and reteset scan- often used for therapeutic effects of drugs
What are the advantages of MRI
Versatility (structure and function) often at the same time
Simple statistics e.g. t-tests, correlation analysis, that are easy to interpret
In vivo methodology
Longitudinal designs- repeat the scan without introducing unnessary harm to participants
“Translational Neuroscience” - allows you to take discovery from the bench to clinical applications
“Safe and cheap”
What are the limitations of MRI?
MRI is rather devoid of biological information- interested in underlying mechanisms that drive behaviour or phenomenon so what people try to interpret and discuss is often underneath the level of what we can measure - caution is needed because we are infering through complex relationships
Temporal resolution is “suboptimal” in comparison with other techniques e.g. EEG - limited by the sampling rate, often the TR, normally 2s, and this impact how we design experiments- fMRI not good for rapidly occuring processes
Counterindications - restrict what kind of patients/participants we can include
Metal
Pregnancy
What are the applications of real-time fMRI?
Novel brain-machine interface
Coupling of the contents of conscious experience and brain activation - e.g. people in locked-in states, can use these methods to assess if people still have consciousness
Planning neurosurgical interventions - if there is real-time information fed back to a neurosurgeon, can be used to guide surgical procedures e.g. epilepsy, removing the areas causing it but not removing areas related to language for example asking people to speak to see which areas are active pre-surgically
Neural feedback - use it to allow neuromodulation, increase or decrease the activation of different neural areas
How can we use the BOLD signal from real-time fMRI ?
We can use the BOLD signal from the brain as feedback to control localised neural activation
Nonpharmacological therapy, e.g. for patients with affective disorders.
Overcomes limitations of other feedback methods, e.g. SCR and EEG.
Directly targeting specific brain regions with superior spatial resolution
What is a neurofeedback paradigm?
Neurofeedback is a method that assists subjects to control their brain waves consciously.
Imaging is recorded during the neurofeedback treatment.
Then, its various components are extracted and fed to subjects using online feedback loop in the form of audio, video or their combination.
What does neurofeedback allow patients to do?
Control brain activity- used to change the amplitude or speed of specific brain waves in particular brain locations
A type of biofeedback that helps you harmonize your brain waves, naturally (without medications).
During the neurofeedback sessions, your brain “learns” how to bring abnormally fast or slow waves into the normal range
treat ADHD, anxiety, and insomnia.
What did Ugur, Ture et al. (1999) find?
Feedback is necessary to impose a consistent increase in control, just practiviing mental strategies may not have any effects
Strong evidecne to demonstrate the paradigm works
What is multivariate pattern analysis?
Motivated by machine learning and AI
Allows us to get more sensitive information from the brain than univariate approaches- most UV analyses unimplicitly or implicitly average acrosss a number of different voxels and try to compute the mean activation across large brain regions
In MV we try to see the activation of different voxels as a multivariate pattern- could be related to parallel distributed representations in the brain
It enables the advance machine learning techniques such as pattern classifiers
Why is MV better than UV analysis?
The main advantage of multivariate analysis is that it considers more than one factor in data analysis.
It looks at the various independent variables that influence the dependent variable.
The conclusions you draw from MVA are also more likely to be accurate
What is RSA?
Representational Similarity Analysis
a computational technique that uses pairwise comparisons of stimuli to reveal their representation in higher-order space
What did Kriegeskorte et al. (2008) find?
Present different stimuli to the brain and from a particular region you get patterns of activation
2x2 matrix of different voxels, every voxel has a different level of activation for e.g. a face or a house
You can compute a correlation matrix in RSA analysis called representational dissimilarity matrix (1- correlatation matrix which gives us a distance matrix where 0 means the two things are almost identical)
Compared monkeys and humans
Strong activations in both for objects that were similar
What does RSA allow you to do in studies?
Allows you to have a wider sampling of the stimulus e.g face or object or animal etc
more naturalistic- would see them everyday in life
Make comparisons in activation between all the different types of stimuli- natually get object categories
For what disease can these exp designs be used in a clinical context?
AD
Memory problems are often the first sign of Alzheimer’s disease.
Decreased activation at medial temporal lobe for Alzheimer’s disease than healthy controls (Golby et al., 2005).
Increased activation at medial temporal lobe for high-risk individuals comparing to low-risk controls during encoding but decreased activation during recall (Bassett et al., 2006).
What is Episodic reinstatement in Medial Temporal Lobe?
Show 2 houses- both create different patterns of activation
If we then give people a delay and then present the same house again, we expect the brain response is a pattern that will be more simialr to the same house they saw earlier than a different house (if you present the same thing the same pattern re-emerges).
If someone is higher risk for AD, the prediction is the two patterns will become less similar than they should have been which may be an underlying neural mechanism for poor memory - high risk individuals have less reliable reinstatement
What is study that assesses episodic reinstatement in MTL?
Learning phase (37 indoor, 38 outdoor)
Testing phase (75 old, 25 new )
Learning these stimuli
Undertake strucutral scans for 30-40 minutes
At the end of the scanning session they didnt know they had to remeber those stimuli or that they would be tested for their memory - so this is incidental memory
Are the images old (seen) or new(unseen)
Computed the correlation matrix - compare brain responses during the encoding phase to repsonses in the retrieval phase based on the same set of stimuli
What did the AD encoding retival study find for a family history of AD?
Found no significant difference in recognition memory performance, but … Spatial location related episodic memory impairments in family history of dementia
In parahippocampus gyrus- in forgotten trials, the similarity between encoding and retrieval is dramatically decreased if you have a parent with AD (high risk)
People with low risk, didnt have this dramatic decrease in comparison
Despite increase in activation in medial temporal lobe in people with high risk for AD, the amount of info that is encoded in the brain is reduced- nice finding, seems to be an early sign of memory impariment and damage caused by the underlying pathology
What did the AD encoding retival study find for carriers of ApoE?
Does not actually change the quality of neuronal representations in these experiements, only family history does
