ds003645/sub006 rotation issues (Scott)

[arnodelorme]

• one clearly bad channel got through the cleaning
• the maps are all turned 90 deg to the left
• the browser option does not work
  whole dataset:
• fix the montage orientation
• make the browser run on these data - ask Yahya how to do this
• check and adjust the data cleaning parameters (see above)
• please run the pipeline again using the new AMICA call Jason provided
• although Dung told me that the runs are concatenated before running ICA decomposition
  in this dataset that is not the case - each run has a separate decomposition.
  Please concatenate the runs and run a single decomposition -- as you thought had been done already.
• when you run the pipeline, please add the PMI (pairwise mutual information) and save the output and the component subspace plots for review.

[dungscout96]

• Which cleaning parameters should I use?
• @neuromechanist Can you point me to the updated AMICA code? Also I don't remember if the PMI scripts have been fixed per our last discussion?
• @arnodelorme We discussed last time that it might not be necessary to concatenate the runs. Should we discuss this again?

[neuromechanist]

Hi Dung,

You can use sccn/amica/runamica17_nsg.m to run efficiently on Expanse. You should use [batch, 1] in the input so that the job is submitted as a job. Other parameters seem to be ok and would not need any change.

Let me know how it goes.

P.S.: It seems that we are once again at the end of NSG allocation. I hope that you can submit new tasks.

[arnodelorme]

Too short for ICA would be if the number of samples is less than nchannel^2*20 samples.

[neuromechanist]

BTW, the amica bag code is available from:

sccn/amica/runamica17_nsg.m

[dungscout96]

@neuromechanist the AMICA call ran yet seemed to exit with error. There's no output in the amicaout directory (see screenshot) and there's no message to indicate that the process finished in the log file (attached)

amica.30062187.txt

[neuromechanist]

Interesting, pinging @japalmer29 for potential insights about the error:
forrtl: severe (66): output statement overflows record

[japalmer29]

@dungscout96 Could you post the matlab command you used to run it?

[dungscout96]

Thanks for helping @japalmer29! I ran the batch mode of https://github.com/sccn/amica/blob/e0db55f07c8c410641cd5b0df5498a5a53129ce6/runamica17_nsg.m#L849

[japalmer29]

@dungscout96 could you post the runamica17_nsg(...) command with all the arguments?

[dungscout96]

The exact command was runamica17_nsg(EEG, 'batch', 1)

[neuromechanist]

It seems that the problem is that runamica17_nsg does not return anything in batch mode. I just added a mirror return (EEG=dat) in sccn/amica@3257aba, so it returns dat (which, most of the case, is the EEG structure) in the batch mode as well.

However, to load data later, you should use EEG = eeg_loadamica(EEG, outdir, 1) for single model data.

[dungscout96]

EEG=pop_chanedit(EEG, 'forcelocs',[],'transform',{'TMP = X; X = Y; Y = -TMP;'});

[arnodelorme]

Add HED to https://nemar.org/dataexplorer/detail?dataset_id=ds002718

Add new columns to NEMAR.json to disable ICA for that dataset