Cochlear implant artifact reduction:

This dataset is from a 37 female with bilateral CIs. We tested her RIGHT CI while the LEFT CI was turned

off. No electrode gel was applied over electrodes: e46, e49, e57, e58 and e63.

The stimulus was presented in free-field and was a continuous tone (250 Hz) that changed to 500 Hz

(100% change, lasting 400 ms) every 2 secs. This is trigger S1.

Raw files:

37F_CI022_visit5_DF_250_100p_p 37F_CI022_visit5_DF_250_100p_p_run2 37F_CI022_visit5_DF_250_100p_p_run3 37F_CI022_visit5_DF_250_100p_p_run4 37F_CI022_visit5_DF_250_100p_p_run5

Step 1 is to append all these files together

Step 2 is to remove to electrodes that were directly over the CI (i.e., no electrode gel as applied to these)

Step 3 is to resample the data to 500 Hz AD

Step 4 is to filter the data 1 to 50 Hz

Step 5 is to run ICA

Step 6 is to remove “standard artifacts” (i.e., eye blink, horizontal eye movements, EKG…)

Step 7 is to identify and remove CI artifacts

Step 8 is to perform an average reference

Step 9 is to interpolate the missing electrodes

Files:

Appended file: 37F CI022 DF 250 runs1to5.set

CI electrodes removed, resampled and filtered: 37F CI022 DF 250 runs1to5 500 Hz AD remove CI

elect.set

The above file after ICA was run: 37F CI022 DF 250 runs1to5 500 Hz sampling ICA.set

Electrode Names:

The high-lighted channels were directly over the CI, and gel was not added to these.

The high-lighted represent the electrode NUMBER (not label) … e2 is channel #1 … therefore e49 is

channel #49.

Lets remove those channels and save

Run ICA

IC 1 – eye blink

IC 2 – horiz

IC 7 and IC 10 - EKG

IC 11 – weird looking pulse?

IC 12 – weird looking pulse?

REMOVE: 1 2 7 10 11 12 ICs

Save as “37F CI022 DF 250 removed regular artifacts.set”

Lets extract the epochs, removed noisy epochs, average reference

58 trials were rejected (exceeded 75 uV)

Create an averaged reference and save “37F CI022 DF 250 epoched cleaned AR.set”

ERP looks good in this subject however, the artifact is a pulse (onset and offset of frequency change) …

these artifacts are particularly big near the CI

Lets now re-run ICA (sometimes this is not necessary, can use the IC’s from the previous .. usually

doesn’t make too big of a difference) but lets do it anyways

Save as: “37F CI022 DF 250 CI ICA.set”

Now lets identify CI related artifacts

Based on:

1. topography

2. time waveform

Plot the ICs … looks like IC 19 is the only CI-related artifact

Most of the artifact is removed

Save as “37F CI022 DF 250 epoched 1 IC CI removed.set”

Now we must add “blank channels” for the ones we removed in the beginning (where the CI was)

missing_chns{1}='e46'; missing_chns{2}='e49'; missing_chns{3}='e57'; missing_chns{4}='e58'; missing_chns{5}='e63';

EEG.data(end+i,:) = 0; for i=1:length(missing_chns)

EEG.nbchan = size(EEG.data,1);

if ~isempty(EEG.chanlocs) EEG.chanlocs(end+i).label = missing_chns{i}; end;

end

[ALLEEG EEG CURRENTSET] = eeg_store(ALLEEG, EEG, CURRENTSET); eeglab redraw

we must also add these locations (at the end) in the SFP file

(“37F_CI022_visit5__removedCI_addedchs.sfp”)

lets plot and make sure these are zeroed

Lets now interpolate those added channels

Save as “37F CI022 DF 250 epoched 1 IC CI removed interpolated.set”

Plot:

Looks good!

Just FYI …. The order of the channels are different now (and will not be consistent across subjects)