From stimulation to seizure onset zone

Problem. You have a stimulation recording and a clinical seizure onset zone label, and you want to know whether the evoked responses agree with it.

The pipeline

1. Bipolar Reference — intracranial data, so reference against the neighbouring contact.
2. Epoch — tmin -0.5, tmax 1.0, around stimulation onset.
3. Reject by Variance — 4 standard deviations, per channel.
4. Crowther Artefact Removal — the stimulation pulse itself.
5. Average by stimulation site.
6. Morse Wavelet — 2 to 200 Hz, 60 bins.
7. Significance testing, then FDR Correction at alpha 0.05.
8. Zone fractions and graph centrality.
9. Compare against the clinical label.

What to watch

Run the artefact removal after rejection, not before: a trial dominated by the pulse is exactly the one you want rejected, and cleaning it first hides that.

The wavelet step is where the memory goes. On a high-density montage, decimate first or process channels in groups. See SS-E060.