Beta Series Correlation:
What do they measure?
Goals
- understand beta series correlations
- compare task and rest data
- first level comparison
- graph theory measures
- multiple parcellations
How do we model brain activity?
The Brain during a Task
How do we model brain activity?
Condition A (Orange)
How do we model brain activity?
Condition A (Orange)
How do we model brain activity?
Condition B (Blurple)
How do we model brain activity?
Condition B (Blurple)
How do we model brain activity?
- Brain Regions Conditionally Correlated
- Activated vs Activated Together
How do we model brain activity?
But Data are Noisy
How do we model brain activity?
And Trials Occur Close Together
How do we model brain activity?
LSA
(least squares all)
- Computationally Cheap
- Works when intertrial interval is long
LSS
(least squares separate)
- Computationally Expensive
- Works regardless of inter trial interval
Estimate Activation (Betas) of Individual Trials
How do we model brain activity?
Task Switching
Task 1: High or Low
Task 2: Odd or Even
4
Task Switching
Task 1: High or Low
Task 2: Odd or Even
4
Validation of Beta Series
Treat resting state fMRI AS IF task switching was performed.
Hypothesis: There should be a normal distribution of correlations between regions with a mean of 0.
Methods
- Run fmriprep on all participants for both the task and rest data
- calculate beta series for both task and rest
- use the Schaefer (2018) atlas to define regions
- generate correlations between regions
Compare within/between network correlation
Compare within/between network correlation
Switch - Repeat
Task - Rest
Task - Rest
Task - Rest
Graph Theory Measures
Participation Coefficient:
how much a parcel connects to all other parcels, with 1 meaning the parcel is connected to all other parcels equally.
Result:
Default network has increased participation coefficient during task relative to rest.
Task Positive Regions
Instead of using the Schaefer atlas, use task activations.