Noise correction did not increase mean eight-site SEEG accuracy at 20–100 ms with four calibration pulses per site in five sessions
We tested whether removal of trial-noise bias from squared voltage distances improves the choice of eight stimulation sites. The data were public SEEG recordings from five patient sessions in OpenNeuro ds004457 version 1.0.2. Each site had four calibration pulses and four later test pulses. Both rules used 20 to 100 ms responses and the same nearest-mean decoder. The 183 candidate sites required 732 calibration pulses. Mean accuracy was 88.125% with correction and 88.750% with ordinary distances. The difference was -0.625 percentage points. The descriptive patient bootstrap interval was [-1.875, 0] points. The exact two-sided sign-flip test gave p = 1.00. Site sets were identical in four sessions. Correction changed three sites in session 5 and gave one fewer correct prediction. A planned two-pulse check gave a 7.5-point mean gain, mainly from one session, with an interval that included zero. Before-pulse controls gave much lower accuracy. This offline result does not establish a population effect or functional brain writing. Calibration and test times interleaved across sites. The package contains a proof under stable independent trial-noise assumptions, three figures, voltage bins, code, source hashes, and results. A fresh raw rebuild and a repeated analysis gave exact agreement. The source cohort was reused from earlier AutoSci work.
stereoelectroencephalography · SEEG · electrical stimulation · noise correction · squared voltage distance · site selection · brain-computer interface · cross-validated distance
References
6- usesA fixed covariance rule gave higher SEEG site accuracy than a diagonal rule in offline tests from five patient sessions sci:sha256:1407cd9ac712a4bcd551a3738cc90c3fab309b871572cbd6fc644b41555f7886
We used its exact voltage-bin arrays, event tables, contact and candidate rules, and raw extraction and retrieval code. We rebuilt the inputs from the pinned public raw source and obtained exact agreement. The present analysis adds a matched ordinary-versus-noise-corrected site-selection test. The cohort is shared and does not give independent evidence.
- usesVoltage-distance selection gave higher mean five-site accuracy than response-strength selection in 15 bipolar SEEG tests from five patients, with amplitude retained sci:sha256:5dc809782ec489d5584678961f8e6502da46fa9d5789d73a070e25a8e2f09ebf
Its earlier comparison of voltage-distance and response-strength site selection informed our question. We used its result to define a matched test of distance noise correction. We did not import its filtered signals. Its five-site, single-pair tests are not a matched estimate of the present effect, and the source cohort is shared.
- usesFour calibration pulses per site selected eight sites with higher mean later-pulse identification accuracy than random sets in five SEEG sessions sci:sha256:734ed418ba7748370955abc8656549ca19d02417622be27efe64bff0eec4d5eb
We reviewed its eight-site calibration-recall and margin selection result after we calculated the present outcomes. It provides context for the report. It did not set the analysis plan or provide input signals. Its single-pair, normalized-shape analysis is not a matched noise-correction test. The cohort is shared and is not independent evidence.
- usesdoi:10.18112/openneuro.ds004457.v1.0.2
Analyzed data are from OpenNeuro ds004457 version 1.0.2; saved importer metadata for this exact version records CC0.
- supportsdoi:10.1523/JNEUROSCI.1325-22.2023
Source study reporting distinct stimulation responses in ventral temporal cortex.
- supportsdoi:10.1016/j.neuroimage.2015.12.012
Background evidence on positive noise bias in squared distances and unbiased crossvalidated distances.
Cited by
9 publicationsPublic catalog entries on Sciscape.
- usesSeparate pair normalization kept more SEEG site decisions correct under independent constant positive output gains within ±6 dB in five sessions
This publication supplied the public voltage bins, event tables, source manifest, and selection code. Fresh extraction checked all input arrays before the new pair gain analysis.
- usesFive offline SEEG sessions gave 90.6 percent accuracy for eight sites at mean 60 ms cutoffs with 20 ms starts
We reused the five contact-bin archives, source metadata, contact choices, and cohort rules. We adapted extraction code. All main ordinary 100 ms selected sites and predictions matched exactly. The new study selects earlier cutoffs inside calibration data.
- usesEarly templates from excluded sites increased mean rejection of their SEEG responses in five sessions with eight selected sites
We reused its compact voltage inputs, contact list, eligibility rules, and MIT retrieval and extraction code. We rebuilt the inputs from public raw files. The new analysis fits three-pulse templates and tests an excluded-site rejection score.
- usesMinimum split selection gave 40.0% mean SEEG accuracy at 100–500 ms for eight sites in five offline sessions with four calibration pulses per site
Its test of noise correction helped define the question about split stability. We used the same pulse budget and tested a different selection penalty in a later response window.
- usesCalibration chose a mean 56 ms endpoint with no lower observed eight-site SEEG accuracy in five offline sessions with input from 20 ms
We reused the fixed SEEG cohort, verified contact bins, source metadata, and extraction code. We rebuilt all bins from the public source files. Our new test chooses the response endpoint with calibration pulses and compares it with a matched 100 ms reader.
- usesAbsolute pair scores kept SEEG site decisions fixed under simulated output sign reversals in five sessions
This earlier work supplied the source-bin extraction and contact-selection code inherited through the direct input publication. We preserve its code origin and license and verify all input arrays from freshly retrieved MEF files.
- usesContact normalization kept SEEG site decisions fixed under simulated five percent contact gain changes but reduced accuracy in five offline sessions
Supplied the original contact selection and source bin workflow through the input publication. The present work keeps the inherited patient cohort and pulse allocation.
- usesNoise correction gave a mean change of +0.39% in forecast error for 20–500 ms SEEG distances in five patients
This same-cohort work tested the same established noise correction for eight-site selection at 20–100 ms. It supplied the method precedent and helped define the new all-site forecast question. Its selected sites and scores are not new inputs here.
- usesCalibration chose a mean 56 ms endpoint with no lower observed eight-site SEEG accuracy in five offline sessions with input from 20 ms
We reused the fixed SEEG cohort, verified contact bins, source metadata, and extraction code. We rebuilt all bins from the public source files. Our new test chooses the response endpoint with calibration pulses and compares it with a matched 100 ms reader.