Dear community,
I would really appreciate your input on how to best treat my experimental design.
For some background: I have resting-state fMRI data collected during natural sleep inside the MRI scanner, with simultaneous EEG recordings for sleep staging. My main interest is in the two REM sleep substates — phasic and tonic REM.
In EEG research, phasic REM is usually defined as short windows (3–5 s) that contain multiple rapid eye movement (REM) events (each < 500 ms). Typically, windows containing at least two such REMs are classified as phasic, while windows without them are considered tonic. Across REM sleep, the phasic periods usually account for about 20–25% of total REM time, and a single phasic window may contain 2–8 REM events.
Following this convention, I plan to segment my REM sleep periods into 5-s windows (corresponding to two TRs, as TR = 2.5 s).
My main question is:
If I want to compare brain activation between phasic and tonic REM substates using fMRI, how should I model these conditions — as a block design, an event-related design, or perhaps a mixed/microstate-like approach?
Any insights, practical suggestions, or references to similar analyses (especially for spontaneous or sleep-related paradigms) would be greatly appreciated.
Thank you very much for your time and advice!
Best,
Xinyuan