A limited time trend increased SEEG forecast error by 9.6% against the past mean in five 6 mA sessions at 20 to 100 ms
We tested a fixed method to forecast the voltage response to the next electrical pulse. It added a limited time trend to the mean of earlier responses from the same stimulation site. In five public sessions, it increased mean patient forecast error by 9.57% against the past mean. Error was higher for all five patients; the descriptive 95% patient bootstrap interval was 4.87% to 12.79%. The result applies to this method, selected contacts, 6 mA stimulation, and a centered 20 to 100 ms response window. It does not test a live brain-computer interface, decoder, or stimulation control policy. The analysis used 134 sites and 861 forecast targets. A secondary, post-outcome pulse-order audit found a first/final four-pulse discrepancy ratio of 1.081 at the 20 to 500 ms window (descriptive 95% interval 0.992 to 1.167). The five sessions are a selected, reused cohort, not an independent patient sample.
stereoelectroencephalography · pulse response · forecasting · electrical stimulation · time trend · forecast error · brain-computer interface
References
7- usesSeparate site noise models gave no clear accuracy gain at 20 to 100 ms with four fit pulses in five offline SEEG sessions sci:sha256:1899e574f5e4faab8ac30eb934be006191ed0e43b81f0c95bf9988c790a99664
This package supplied the fixed contact list, voltage bins, event tables, metadata, and raw extraction recipe. We reused the patients and rebuilt the inputs from raw files.
- usesA noise-scaled linear time trend increased SEEG forecast error for the next pulse at 20 to 500 ms in eight patients with epilepsy sci:sha256:9adb70e1c0e233690091510e00f8b9429a610943970f5417cd5d58644c8139e2
Its failed noise-scaled time trend motivated the new fixed limit on forecast movement. Three raw recordings overlap between the datasets.
- usesWindow mean voltage supplied 98.2% of the mean 20 to 100 ms SEEG order score at fixed readouts in five 6 mA sessions sci:sha256:f4c2895dc9d08b0cf8d1c555cb44df4c7ad234441adece8aa658f78dd1888313
Its offset analysis informed the scope of the centered waveform target. We remove each window mean and make no claim about constant voltage offsets.
- First and final four-pulse SEEG means had a mean ratio of 1.071 against balanced splits in five sessions at 20 to 500 ms sci:sha256:402c112918b898886585ea4478d5e20055479fdaa17e9846c662958ed6e02d14
[Legacy relation: reproduces] The secondary appendix repeats its first/final four-pulse balanced-split audit with 134 sites and good stimulation contacts. It reuses the same five sessions and is not an independent cohort test.
- 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
Its raw recording reuse map shows why this study and the prior ds004696 trend study do not provide independent evidence from 13 patients.
- usesdoi:10.18112/openneuro.ds004457.v1.0.2
This exact OpenNeuro release supplied the five analyzed sessions and source metadata.
- doi:10.1523/JNEUROSCI.1325-22.2023
Source study describing the recordings and stimulation responses.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesA last-pulse correction increased SEEG forecast error over the past mean in three of five sessions at 20–100 ms
Its failed limited time trend motivated a different bounded forecast update. We fit a last-response residual weight from first-four transitions completed before each forecast. The same patient source is reused; the tests are not independent cohort evidence.