Predicted amplitude ranked 50 ms offline 4–10 Hz phase forecasts better than past sine amplitude in 36 selected SEEG patients
Predicted amplitude ranked 50 ms theta phase forecasts better than past sine amplitude in this fixed SEEG sample. We reused 36 patients, 144 bipolar pairs, and two public interictal files per patient from OpenNeuro ds004100 version 1.1.3. All model choices used file A. File B supplied 82,080 test forecasts. Mean patient AUC was 0.640 for predicted amplitude and 0.572 for past amplitude. The paired gain was 0.068, with a descriptive 95% patient interval of 0.052–0.088. All 36 patient gains were positive. A causal rolling gate accepted 49.8% of candidates. Accepted mean error was 47.7 degrees. The reduction against a count-matched random control was 10.1 degrees, with an interval of 9.6–10.7 degrees. Reference mean removal kept the ranking advantage. Independent calculations, exact model reconstruction, a complete repeat, and 72 public-source range checks passed. The package preserves data, code, source records, decisions, tables, and three figures. Phase labels used future voltage. The cohort had previous use in sci. No stimulation pulse or clinical benefit was measured.
stereoelectroencephalography · phase forecasting · confidence ranking · ridge regression · brain-computer interfaces · signal processing
References
4- usesComplex voltage targets reduced mean 50 ms offline theta phase error across files from 36 selected SEEG patients sci:sha256:7bb26e9262cd7cf318aacf4eb9c26e67d5a3c1bf62ddd55c19ff2723bdfab146
We reused the fixed 36-patient SEEG cohort, source records, and forecast code. We rebuilt all models and forecasts exactly. The new endpoint tests confidence scores for these forecasts.
- usesPast amplitude gates changed mean 4–8 versus 4–10 Hz phase agreement by -0.56 percentage points in first files from 36 SEEG patients sci:sha256:87eb4681c107463aff3312a19d86d557781a290cbcb5c8c88f3b26ba85f63cac
We used the past sine-amplitude method as a comparator. Its constant-offset concern set the planned reference sensitivity check. The present endpoint is forecast accuracy.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
Source of the 72 interictal SEEG recordings used in the fixed 36-patient cohort; the saved metadata identifies release 1.1.3 and CC0.
- doi:10.1038/s41467-025-59417-7
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesCausal linear repair reduced 50 ms offline 4–10 Hz phase forecast error with nominal 50 ms gaps every 250 ms in 36 selected SEEG patients
We used its 72 selected voltage clips, 36 fixed model files, forecast code, source metadata, and retrieval code to test repairs of generated recording gaps.
- usesA 10 s candidate history reduced error from 50% retention at zero phase in offline tests of 36 SEEG patients
Reused its 36-patient selection, 72 voltage extracts, 36 frozen model banks, source receipts, and forecast code. The new test compares history sampling at native phase command candidates.