fdelta vs sdelta for N3 classification #209
-
|
I have been doing sleep staging experiments on EEG data (collected via Muse S). One thing that I notice is that for N3 classification, YASA is more sensitive to fast delta (1-4hz) as compared to slow delta (0.5-1hz). While playing around with the features, I saw that even when the total delta power is the same (say 0.7), the probability of N3 is higher if the delta is 0.5 fdelta and 0.2 sdelta, as compared to when it is 0.5 sdelta and 0.2 fdelta. This confuses me as my understanding was that slow oscillations are core to N3 stage. To some extent, this does make some sense to me as one could think that a lot of noise from EMG usually falls in the sdelta frequency range. But curious if there are other neuroscience reasons for this. PS: Wasn't sure if I should raise this question here. Let me know if I should ask this somewhere else |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Hi, I have converted this issue to a discussion. I cannot speak about data collected with the Muse system, but you can see below the top 20 most important features in the model on PSG data (from the Vallat et al 2021 paper). In line with your observation, fast—but not slow—delta is among the top features. In my own experience, I have observed that the prototypical N3 slow-waves are often around ~1Hz (note: confounded by many factors) and therefore more likely to be captured by the fast delta band.
|
Beta Was this translation helpful? Give feedback.
Hi,
I have converted this issue to a discussion.
I cannot speak about data collected with the Muse system, but you can see below the top 20 most important features in the model on PSG data (from the Vallat et al 2021 paper). In line with your observation, fast—but not slow—delta is among the top features. In my own experience, I have observed that the prototypical N3 slow-waves are often around ~1Hz (note: confounded by many factors) and therefore more likely to be captured by the fast delta band.