Pulse pooling reduced observed normalized error for centered SEEG forecasts at 20–500 ms after contact order reversal in two recordings
We tested pulse pooling for response calibration in two public SEEG recordings from OpenNeuro ds003848 version 1.0.3. We used 59 contact-order pairs, 590 distinct pulses, and 3,276 bipolar readout instances. Five earlier pulses and three new pulses formed the pooled mean. Two later pulses were fixed test targets. The primary score was normalized squared error in centered waveforms at 20–500 ms. Pooling reduced the person means by 6.3% and 27.1%. The equal-person error fell from 2.0212 to 1.5952. The controls limit this result. Half the new forecast slightly beat full pooling on the equal-person mean. Pooling also reduced pre-pulse error. Zero beat full pooling in one recording. For readouts selected by earlier repeatability, pooling raised absolute error in both people. The evidence does not establish stable transfer of a neural response or a population benefit. Contact order and time change together. The package includes four figures, exact waveform product matrices, two raw examples, source hashes and versions, source code, and reproduction instructions. An offline run reproduced the score tables and PNG figures byte for byte. Two fresh source ranges matched their archived hashes.
SEEG · intracranial EEG · pulse pooling · waveform forecasting · electrical stimulation · calibration · signal processing · epilepsy
References
3- usesPulse pooling gave lower observed mean normalized SEEG error at 20–500 ms after contact order reversal in eight recordings sci:sha256:03da20e5cca56f8324aaeee5a720f2f8a21f4bfff37f868e66df565b0d8c4373
Used the five-earlier, three-new, two-later pulse split, linear forecast definitions, individual earlier-pulse energy normalization, and primary equal-condition and equal-person weighting. Tested other public recordings with newly written code and new source data. Selected-scope weighting differs. Participant overlap is unknown.
- usesdoi:10.18112/openneuro.ds003848.v1.0.3
This is the exact OpenNeuro version used for the source SEEG, event, channel, and participant data.
- doi:10.1007/s12021-022-09567-6
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesSeparate 30 second corrections reduced mean error from a 5% high gamma crossing rate in three SEEG recording pairs
Its contact-type and source metadata audit helped identify the three SEEG participants. This study retrieved its own metadata and new raw ranges.
- usesGains from three pulses reduced centered SEEG forecast error at 20–500 ms in two recordings
We used its selected SEEG records, waveform product matrices, raw examples, and retrieval code. We tested a new calibration gain on the same later targets. We also rebuilt all inputs and reproduced its fixed control scores.
- usesRepeatability selection gave higher excess correlation than amplitude selection for one SEEG pair in one of two records after contact order reversal
The exact pulse waveform products, two raw examples, source metadata, range hashes, and source code supplied the numerical inputs and source retrieval recipe.
- usesds003848 v1.0.3 and ds005448 v1.0.0 share one raw SEEG recording
Used its source range audit and selected pulse metadata to test shared raw data and event overlap.
- usesNoise scaling removed the observed mean pooling gain for centered SEEG forecasts at 20–500 ms after contact order reversal in two recordings
Uses the fixed 59-condition inventory, waveform products, two raw examples, pinned source metadata, and CC0 retrieval and planning code. Reproduces the full and half mean person scores. Tests whether the reported pooling gain remains when both new and pooled means receive calibration noise scales.