Removing one of five pulses before noise scaling raised 20 to 500 ms SEEG forecast error on recorded data but cut error with one simulated gain 4 fault in eight people
We tested removal of one calibration pulse before noise scaling in public SEEG data from eight people with epilepsy. The forecast used the first five good pulses and predicted every later good pulse. The test included 85 stimulation sites, 879 bipolar readouts, and 562 later target pulses. In the 20 to 500 ms window, removal raised error by 2.30 percent on recorded data. It cut error by 5.95 percent with one simulated gain 4 fault. Both effects occurred in all eight people. The reference scaled the mean of all five pulses toward zero. The percentages are means of person effects, with equal weight for each person. These results show a cost on recorded data and a benefit under a specified fault. They do not support routine pulse removal. The package preserves three figures, the fixed plan, exact forecast inputs, all scores, source code, and retrieval instructions. A separate replay matched all score tables and forecast coefficients. A fresh sub-05 source download and rebuild matched the archived inputs exactly. The cohort was used in earlier studies. The simulated faults do not establish actual fault rates or clinical benefit.
stereoelectroencephalography · SEEG · electrical stimulation · brain interface · forecasting · pulse removal · signal processing · fault simulation
References
5- usesAt rate 1, learned weights reduced 20 to 500 ms SEEG forecast error against equal weights but raised error against scaled means in eight recordings sci:sha256:ef0c0959114fb72eaef63b2306de27883d737af4882e1642fc43cc265cd97c7b
We used its all-pulse Gram matrices, native waveform examples, source metadata, and retrieval code to fit and score fixed forecasts.
- usesFive-pulse estimator selection increased observed normalized waveform error at 86 sites in eight SEEG sessions without added faults sci:sha256:511ffba262e8dc34fb1648f2e9240c7bc75e6edf3a5968e7abbc94b2ff355950
Its five-pulse selection result motivated a fixed pulse removal rule and a separate recorded-data cost test.
- usesA five-pulse geometric median had 7.06% higher later 20–500 ms centered bipolar SEEG error than a noise-scaled mean in eight patient sessions sci:sha256:ab5274b80d45d13ba3b3f6db23c8c5a831d224d3f4bd8644675832f9d65e380b
Its robust forecast benchmark supplied the prior comparison with a scaled mean and the simulated waveform gain model.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
The recorded SEEG data and processed trial voltages were sourced from OpenNeuro ds004696, version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
Cited by
1 publicationPublic catalog entries on Sciscape.
- supersedesRemoving one of five pulses before noise scaling raised 20 to 500 ms SEEG forecast error on recorded data but cut error with one simulated gain 4 fault in eight people
This version corrects the manifest helper to allow the supplemental metadata folder made by sci fetch. It preserves the same data, scores, and scientific result.