In 53 ds007445 seizures, thalamic line-length rise followed SOZ but not white matter at z > 3

We measured when the ipsilateral thalamus shows a seizure-related line-length increase in OpenNeuro ds007445 v1.0.4. We used 54 EDF files (53 seizures, first 3 per patient, 19 patients) and read only [onset-60 s, onset+63 s] with HTTP Range requests. We used the dataset channel labels for the regions: SOZ, ipsilateral thalamus (AN, CM, MDm, PUL, VPL, LP, Vlp), ipsilateral white matter (WM) and ipsilateral non-SOZ cortex. For each pair, we calculated the 1-80 Hz line length in 1-s bins and the robust z against the [-60, -10) s baseline. The latency is the first time at which the region z is more than 3 for 3 s, and it is censored at 60 s. We wrote the plan and its hash before we calculated features. We used exact sign-flip tests with the patient as the unit and the Holm correction. Results: the median patient latency was 1 s in the SOZ, 5 s in non-SOZ cortex, 14 s in WM and 14 s in the thalamus. The thalamus came later than the SOZ (17 patients; median difference 8 s; 12 of 16 non-zero differences were positive; Holm p = 0.020). The thalamus did not come later than WM (median difference 1 s, p = 0.87). The difference from cortex was not significant (Holm p = 0.31). The thalamic trace passed the threshold within 60 s in 63 % of seizures. At a strict threshold (z > 5), the thalamus came later than WM and cortex. But at this threshold most thalamic values were censored, so the result shows weak thalamic change, not a measured delay. Limits: dataset clinical labels, 1-s resolution, mostly 200 Hz data, and an FFT filter because scipy was not available. The package contains the code, the plan, the per-pair 1-s line-length time series, the byte-range SHA-256 values and the reproduction steps.

seizure onset · SEEG · thalamus · line length · ictal detection · latency · intracranial EEG