A fixed noise correction reduced later 20 to 500 ms waveform error at SEEG pairs selected from metadata in eight recordings
A fixed correction to the mean of five calibration pulses was tested with public SEEG data from eight people with epilepsy. Pairs were selected from metadata in OpenNeuro ds004696 version 1.0.1. Across 317 conditions, 1,183 condition–readout cases, and 1,933 later pulses, the correction reduced later 20 to 500 ms centered waveform error by 8.99% compared with the plain mean, with lower error in all eight recordings. Corrected response skill against zero output was 8.75%, while mean pre-pulse skill was -0.10%. Only 44.97% of cases had positive corrected skill with equal person weights, and one recording had greater pre-pulse skill than response skill. This is an offline comparison on a reused cohort; it does not show improved online decoding, stimulation control, or clinical outcomes.
stereoelectroencephalography · electrical stimulation · waveform prediction · noise correction · calibration · signal processing
References
4- usesThree SEEG recording pairs share channel hashes and differ in event labels in ds004457 v1.0.2 and ds004696 v1.0.1 sci:sha256:9f29610431edacf4fe7a6597dfc16e60f353657563c522ba96e16fcda1cb6c9b
Used the recording hash audit to identify reused sources and to limit claims about independent evidence.
- usesAt selected pairs in eight SEEG recordings, scoring all later pulses changed normalized 20 to 500 ms error of a fixed five-pulse forecast by less than 0.7% per person sci:sha256:8a1770ece48aa9c3ddf047742b6b8c8eb8cdb236f69d08b90236467e8a4c4206
Used the metadata selection rule, retrieval code, source hashes, and saved response products. Decoded the selected raw waveforms again. Added a fixed noise correction, a zero-output comparison, and a matched pre-pulse control.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analyzed SEEG recordings and metadata were from OpenNeuro ds004696 version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesGains from other stimulation sites increased normalized 20 to 500 ms SEEG forecast error relative to local gains in eight recordings
We used its metadata-selected SEEG cohort, five-pulse moments, later-pulse products, and raw retrieval code. A fresh rebuild from 192 hashed raw objects reproduced its five compact input files byte for byte. We then tested gains learned at other stimulation sites.
- usesMean rates of shared large errors in fixed SEEG waveform forecasts exceeded an independent pulse reference before and after pulses in eight people
Used the metadata selection, fixed noise correction, extraction and retrieval code, waveform products, and voltage examples. A fresh public-source download and decode reproduced all five data files byte for byte. The new endpoint measures joint large waveform errors.
- usesA corrected mean increased average normalized 100–500 ms forecast error in five SEEG recordings with four calibration pulses per site
Its noise correction supplied the fixed multiplier. This study applied that rule to baseline-referenced mean voltage. The tests used a zero-offset forecast and a matched control before the pulse.
- usesFive-pulse SEEG selection with a before-pulse score gave no clear 20 to 500 ms test gain in eight people
This work supplied the exact metadata cohort, readout pool, source manifests, retrieval code, and numerical comparison data. We reused the cohort and rebuilt the raw pulse products. The new question tests a before-pulse term during readout selection, with a different test endpoint.
- usesLinear monitors on other shafts gave no clear SEEG pulse matching gain at 20 to 500 ms with five fit pulses and penalty 1 in eight people
This work used its original metadata selection rules, checked raw retrieval code, and noise correction for the fixed baseline.
- usesTraining energy normalization gave a positive mean five-pulse error gap for 20 to 500 ms responses under uniform partitions of eight SEEG recordings
We used the metadata selection, exclusions, raw-file manifest, and retrieval code. We used its five-pulse matrices and later scalar products to check our new all-pulse extraction.