Chapter: Functional Imaging

[gdevenyi]

No description provided.

[danyluik]

Some ideas -

Goal - identify spatially localized changes in tissue properties related to brain activity

Background

• HRF - how does blood flow correlate with brain activity? Why does this change T2*?
• EPI - what fast imaging techniques can we use to measure this over time?

Preprocessing

• Reference image for head motion - why is head motion a problem? How can we estimate it?
• Slice timing correction - why is it a problem that each slice of an image is gathered at a different point in time? How can we fix this?
• Susceptibility distortion correction - what's a field inhomogeneity? How does it impact our signal? How can we fix it, using field mapping sequences?
• Registration - why are EPI images low-contrast and low-resolution? How can we fix this, using structural scans?
• Detrending - why remove the mean signal from each voxel?
• Filtering - what frequency bands does the hemodynamic response occur in? Why might we remove signals outside of this range?
• Smoothing - how can smoothing boost statistical power and improve signal/noise?

Confound correction

• What do people regress from their time series to correct for confounds? (head motion parameters, global signal, CSF/WM signals, components classified as motion)
• What are the drawbacks of removing, e.g., global signal?
• Misc considerations - should all be done at once/should be orthogonal to filter (Lindquist et al., 2019)

Analysis (in the context of rs-fMRI)

• Registration - why express everything in a common space when doing cross-subject comparisons?
• What are some measures of brain activity that people use (fALFF, functional connectivity)? What do we think these represent, biologically? How can they depend on our preprocessing/confound correction choices?
• What's a brain network (Yeo), and how can we derive these from our time series?

Some of this would already be introduced in an earlier structural chapter (registration, etc.)