sleepscoremaster run on interval set instead of continuous time

[DavidTingley]

This way we could remove 'noisy' segments of recording that often mess up the scoring results...

To do this, we'll probably need to add functionality to LoadBinary.m as well, to load interval sets

[evilrobotxoxo]

Are you talking about using the intervalSet class? You can just save those in a .mat file, so you don't need to modify LoadBinary.m

[DavidTingley]

No, we've stayed away from TStoolbox with this repo. intervals would be an Nx2 matrix of intervals that can be given as in input to SleepScoreMaster and LoadBinary

[evilrobotxoxo]

Ok, that's too bad. I wish you guys would reconsider before you commit to repeating the mistakes of FMAtoolbox, but I don't have time now to contribute in a meaningful way myself.

On Mon, May 8, 2017 at 5:44 PM David Tingley ***@***.***> wrote: No, we've stayed away from TStoolbox with this repo. intervals would be an Nx2 matrix of intervals that can be given as in input to SleepScoreMaster and LoadBinary — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#38 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AE3LJYsj_F8lc-dtkMXhzSLcLS7YYObIks5r34zSgaJpZM4NUhg0> .

-- Sent from my iPhone

[dlevenstein]

similarly, this will allow us to Ignore Behavior epochs during state scoring

[dlevenstein]

From email chain with @rhuszar (how do we add an assignee?):

@brendonw1
I think masking out certain epochs when generating the histograms is easiest.

In this function there is already something called “badtimes”

https://github.com/buzsakilab/buzcode/blob/master/detectors/detectStates/SleepScoreMaster/ClusterStates_GetMetrics.m

Maybe you can take that concept and expand it. That’d let you generate histsandthreshs without bad data in them thereby making the right cutoff thresholds. Then once you have that you’d need to separately decide how to actually score the bad times. In this case i assume they should all be called wake. But I’d not hard code them as wake all the time. I’m thinking you could assign them all state 0 (unscored) or can let the user specify the state to assign to those unused times.

I’m imagining you’d give badtimes or some similar name as an input to sleepscoremster. You can also have an input called something like “defaultscoreforbadtimes”.

The final issue is if/how to carry forward the badtimes for if you re-score the file using the window that comes up when you press “a” (auto) in TheStateEditor. I’d think you can store badtimes in the main states.mat and then whenever ClusterStates_DetermineStates is called it checks for those times after applying the histsandthreshes and overwrites the badtimes with the default state - whether it’s called the before TheStateEditor (ie from SleepScoreMaster) or from inside TheStateEditor.

@dlevenstein
Looks like there’s also an option ‘scoretime’ in SleepScoreMaster that’s passed through to bz_EMGFromLFP and PickSWTHChannel!

I think scoretime and badtimes are both old options that you’ll have to make sure still work as intended, but if we’re lucky it will just be a matter of carrying through the right times.

[brendonw1]

Thanks for transcribing/copy-pasting