Slow power level increased mean onset-label AUC in 19 external SEEG patients with the stored reference after training on 36 HUP patients
We tested whether slow SEEG power level adds rank information after a slow power ratio and contact position. We trained fixed models on 36 HUP patients. We tested 19 NIH and Johns Hopkins patients with the stored reference. Mean onset-label AUC increased from 0.694 to 0.762. The mean gain was 0.069. Its 95 percent patient bootstrap interval was 0.004 to 0.133 when both cohorts were resampled and models were fitted again. The exact sign-flip p value was 0.039 under independent, symmetric patient differences. Within the same contact name group, gain was -0.005 and its fixed-fit interval included zero. Uniform recording gain cancels from the centered features. Contact-specific gain and anatomical sampling can remain. These patients were used in earlier publications. This result is a new analysis of reused data. It does not establish clinical use. The package contains all compact inputs, code, source versions, predictions, bootstrap draws, and public retrieval recipes.
stereoelectroencephalography · seizure onset zone · power spectral density · AUC · patient bootstrap · intracranial EEG
References
6- usesThe patient bootstrap interval for AUC gain from low frequency SEEG power above contact order included zero in 19 external patients after HUP training sci:sha256:dfccf5a17b4d72d5934b20adbeb9c4209df42a72110ea3a4a5dac192a6dea83a
Supplied the prior comparison of a power ratio rank with contact order in these reused cohorts. Its result motivated the new ratio-versus-level contrast. It did not supply the new fitted predictions.
- usesA fixed HUP rule gave 58.2 percent mean balanced accuracy against onset marks in 19 external SEEG patients sci:sha256:e3f4c2691bc3b2c5930552ed90015dcc79a21067be11f7b8ecee7aff9861ea8a
Supplied the 19 external SEEG power files, source extraction code, clinical channel labels, and pinned public retrieval records.
- usesA low frequency rank threshold gave 61.5 percent balanced accuracy in 36 SEEG patients when each test patient was excluded from fitting sci:sha256:0d58492a9d57146df8f13fb78f03b5e8c943a65d8fddfb91acdb27a52d756a83
Supplied the 72 HUP power files and source provenance used to train the two fixed models in 36 patients.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
Version 1.1.3 supplied the HUP training cohort used in this analysis.
- usesdoi:10.18112/openneuro.ds003876.v1.0.2
Version 1.0.2 supplied the external NIH and Johns Hopkins test cohort used in this analysis.
- supportsdoi:10.1038/s41598-019-42347-y
This paper supplies the background evidence that motivated the slow and delta band comparison; this analysis uses different band limits and does not reproduce that experiment.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.