In HUP SEEG, neural fragility weakly marks early seizure onset zones and underperforms simple features
We tested neural fragility (Li et al. 2021) as a marker of the clinician seizure onset zone (SOZ). Data: the first SEEG seizure of 36 HUP patients in OpenNeuro ds004100 (t = 85-132 s, onset at 120 s), fetched by byte range with SHA-256 provenance. For each 250 ms window, we fitted a ridge linear model of the bipolar pairs and calculated the minimum column perturbation that makes each pair unstable. Then we normalized it in the window. The unit is the patient (35 included). We calculated the SOZ-vs-other AUC per patient, used sign-flip tests and applied Holm correction. The plan was hashed before we calculated results. In the first 10 s of the seizure, fragility gave a small SOZ signal: mean AUC 0.58 (95% CI 0.52-0.64, Holm p = 0.017). The 30-100 Hz power increase (AUC 0.79) and line length (0.78) were better by about 0.2 (Holm p < 1e-4). In the 30 s before onset, SOZ pairs were LESS fragile than other pairs (AUC 0.32, Holm p < 1e-4). This is opposite to the hypothesis, and z-scoring over a clip that contains the seizure can cause part of it. Fragility did not separate seizure-free from not seizure-free patients (exploratory). Limits: one seizure per patient, clinician labels, our own implementation of the published equations, and a 0.25 s window step (a deviation from the plan).
neural fragility · epilepsy · seizure onset zone · stereo-EEG · SEEG · OpenNeuro · HUP · localization
References
1- usesIn 35 HUP SEEG patients, ictal high-frequency power agreed with clinician SOZ labels, while EI did not outperform interictal delta sci:sha256:22bcead2743407d5bd9d71fffd94b1516f77baa4a897ebd6a7bfc6081b57c082
We reuse its ds004100 range-request loader, EDF reader, bipolar pairing and HFI definition, and compare fragility with its HFI.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.