[Eeglablist] Issues processing renamed .mff files

[Chris Rose]

Hello EEGLAB devs and users,

I've been working on a project recording on EGI hardware and thus utilizing
.mff folders/files as our initial inputs loaded in for pre-processing. For
the past several months I've used a couple of our early subjects as pilot
data to standardize our workflow and analysis procedure before taking a
look at our full subject cohort with no issues.

Recently my colleague passed along data for those initial subjects as well
as those we hadn't looked at yet with cleaned up .mff filenames. Running
our finalized preprocess on all files(~120 for context) generally went
well, but ran into issues with two recordings where NaN values were being
introduced into EEG.data causing warnings and ultimately a failure to
process. What was most concerning was that one of those two errors was the
rest recording for one of the pilot subjects that I had ran through the
same preprocess before without issue. Our preprocess steps our largely in
line with Makoto's pipeline: file input > remove unused sensor channels >
select a time range based on event markers > resample > cleanLineNoise >
hi/low filters > clean_channels(clean_rawdata, not ASR) > pop_interp >
pop_reref > ICA.

Trying to debug the issue I found EEG.data NaNs were being introduced at
various points with our new data, from a handful of random instances
arising after the resample step to several seconds of data being NaN for
all channels(buffered by extreme voltage values before/after) after running
reref, which ultimately caused the rank function to crash. I could find no
apparent reason for this in this listserv or elsewhere and ultimately
updating EEGLAB/related plugins didn't resolve anything. Since I had the
original unaltered pilot file for the subject with a rest recording issue I
reran that from .mff through ICA and had no problem with the data. Is it
possible that renaming .mff files could cause some sort of data corruption?
>From what I can tell at a glance everything within the original and new
.mff folder contents for that rest recording are identical.

If there have been similar issues reported that I didn't find in my search
feel free to link them to me or if you've resolved tangentially related
problems successfully I would love to hear about it. As always, feedback on
best practice in preprocessing is appreciated, the code I started with was
originally written some time ago and looking at the Makoto's Pipeline page
now it seems like there's been many updates to considerations for best
practice in that time.

Best,

Chris