A shared site scale gave a mean 0.93% error increase for 20 to 500 ms SEEG responses after five-pulse calibration in eight patients
We tested whether one calibration scale shared across recording pairs at a stimulation site predicts final pulse voltage better than separate pair scales. We used public SEEG data from OpenNeuro ds004696 version 1.0.1: eight patients with epilepsy, 86 sites, and 886 remote bipolar readout edges. The first five good pulses fit each mean waveform and its scale. The final five good pulses supplied individual test targets. The strict test window was 20 to 500 ms. Site pooling gave a mean patient error reduction of -0.93%, with a descriptive 95% patient-bootstrap interval of -2.12% to 0.40%. Two of eight patients had lower error. At 49 sites, separate pair scales also beat the best common scale fitted to the final targets. That fit is a diagnostic bound, not a prospective predictor. Pooling within a readout shaft gave a small exploratory gain. The package includes four figures, exact product matrices, pulse losses, code, source checksums, and repeat instructions. All numerical checks and a fresh replay passed. All 44 public metadata files were checked again. A fresh full voltage-source rebuild was not completed in this run. The result applies to these sessions and does not show an online interface benefit.
stereoelectroencephalography · SEEG · pulse response · calibration · waveform prediction · brain-computer interface
References
4- usesA five-pulse SEEG contrast cancels power terms and gives an uncertain mean at 20 to 500 ms in eight sessions sci:sha256:633fd1027607cdad4059625988523866425efc05c4ffe398a3f9ee6f1046eeb5
This work uses its exact eight-patient pulse-product matrices, trial chronology, contact rules, source metadata, checksums, and GPL extraction code. Its full source rebuild is inherited evidence, not a rebuild completed in this run.
- usesFive-pulse noise scaling reduced later single-pulse error for 20–500 ms centered bipolar SEEG in eight patient sessions sci:sha256:fbc06ab469d48eac7a6f54dd97cd24ccd9c7c5ca4b0db4d64340536b3e81428a
This work uses its five-pulse separate pair-scale rule as the primary comparison. It tests new shared site, shaft, and patient scales on reused outcomes. It adds an observed common-scale lower bound and an exact loss-change analysis.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
This version-pinned OpenNeuro dataset supplied the public processed trial voltage used in the analysis.
- doi:10.1523/JNEUROSCI.2201-22.2023
Cited by
3 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
Its shared site and person gain tests defined the prior method and showed the need to test site specificity. Our new test omits the full target stimulation site and matches the recording pair in a broader metadata-selected cohort. The people overlap, so this is not independent biological confirmation.
- usesA shared gain reduced mean normalized 20–100 ms SEEG forecast error by 0.96 percent in five sessions with four calibration pulses
This publication supplied the separate-pair and shared-site gain method and the earlier comparison on ds004696. We tested the method on ds004457 with four calibration pulses and a 20–100 ms main window. We added a numerically checked common-gain bound for the observed final targets. Different conditions and uncertain person overlap limit cohort replication claims.
- Five early pulses set gains that reduced centered SEEG forecast error at 20 to 500 ms against fixed 0.5 gain in seven of eight patients
[Legacy relation: reproduces] A separate calculation repeated the 0.93 percent mean error increase for the earlier shared gain under readout-normalized error. Every patient value matched the reported table to its precision. This study also tested pooled voltage products with a fixed 0.5 gain reference.