Thalamic recruitment was repeatable within patients but not more repeatable than white matter or cortex

We tested if ipsilateral thalamic recruitment is specific to the patient. We used the pair-level recruitment latencies of AutoSci package 5eed1e5b (OpenNeuro ds007445, pediatric thalamic SEEG). Onset is 1-s line length at z >= 3 above baseline for 3 s. Latencies are censored at min(58 s, seizure duration). For each seizure, we calculated D, the median thalamic latency minus the median SOZ latency. We also calculated F, the fraction of thalamic pairs that were recruited. We used a one-way ICC(1), a 20000-permutation one-sided p, a patient-cluster bootstrap CI and Holm correction. The plan was written before the analysis. H1: ICC(D) = 0.38 (95% CI 0.12 to 0.61, Holm p = 0.008, 15 patients, 44 seizures). H2: ICC(F) = 0.53 (CI 0.15 to 0.76, Holm p = 0.0008, 17 patients, 49 seizures). Both results stay when the latencies are capped at 20 s and when sub-005 is removed. But the median-latency ICC was 0.70 in the thalamus, 0.87 in white matter and 0.85 in non-SOZ cortex. Thalamus minus white matter = -0.20 (CI -0.52 to 0.10). Thalamus minus cortex = -0.19 (CI -0.44 to 0.01). Thus the repeatability is not specific to the thalamus. It agrees with a patient-level spread pattern. Seizure duration is also repeatable (ICC 0.63), and the censoring partly follows it. Limits: there are only 2 or 3 seizures per patient, we used one onset detector, and the thalamic nuclei are pooled.

thalamic SEEG · seizure recruitment · intraclass correlation · repeatability · pediatric epilepsy