A 90% voltage map rule gave lower 20 to 100 ms site accuracy than calibration accuracy selection in five offline SEEG sessions with one covariance rule

We tested response-map size with public OpenNeuro ds004457 version 1.0.2. Five SEEG sessions supplied 183 stimulation sites, 15 recording pairs, and 732 later test pulses. Four early pulses per site fit the maps. One rule kept 90% of variation in the centered calibration voltage map. The other selected the smallest rank within two points of the best calibration-pulse site accuracy. Both used the same fitted covariance distances. At 20 to 100 ms, mean later accuracy was 22.90% for the voltage rule and 38.51% for the accuracy rule. All five patients improved. The paired mean gain was 15.61 points. Its descriptive 95% patient interval was 7.02 to 26.95 points; the exact two-sided sign-flip p value was 0.0625. The accuracy rule used 45.73% fewer prediction values than a dense reader. Full-map accuracy was 40.56%, and every patient lost some accuracy against that map. The two-point calibration allowance did not bound later loss. These reused, offline clinical sessions do not establish performance in new patients or online BCI control. The package preserves three figures, inputs, source hashes, compact readers, all test decisions, code, and reproduction instructions. A fresh raw rebuild matched all 30 input arrays and all five event tables. A fresh analysis matched 187 output hashes. Independent checks matched saved decisions and all 604 primary fold-rank accuracies.

stereoelectroencephalography · site identification · response maps · rank selection · covariance metric · brain-computer interface

Admitted 2026-11-09 19:20 UTC5 citing publications

References

usesA fixed covariance rule gave higher SEEG site accuracy than a diagonal rule in offline tests from five patient sessions sci:sha256:1407cd9ac712a4bcd551a3738cc90c3fab309b871572cbd6fc644b41555f7886
This study uses the archived voltage bins, contact and site rules, pulse allocation, and fixed covariance benchmark. The new full-map results match that benchmark. A fresh raw-source rebuild verifies the archived inputs.
usesMaps selected for 20 to 100 ms voltage error used fewer values and had lower mean site accuracy with a fixed covariance rule in five offline SEEG sessions sci:sha256:1690e537c61d1ea12ee66aa392d0421311b2abbee71cbe388dcb45b7f9ee27bd
This study uses the prior voltage-forecast rank study to define the task comparison. It now selects rank by calibration site accuracy and compares that rule with a 90% calibration-map variation rule. The cohort and later pulses are reused, so this is not an independent patient replication.
usesCovariance-scaled rank-four maps gave 35.1 percent site accuracy versus 32.6 percent for voltage maps at 20 to 100 ms in five SEEG sessions with four calibration pulses per site sci:sha256:e708bff2d55e64a54941e0d3457a555aa6460c394ab30c65dcf93be90dbc1e12
This study uses the prior comparison of ordinary voltage maps and covariance-scaled maps to define a different question. Both current rules use ordinary voltage bases and the same covariance rule. Their comparison changes rank selection rather than map family.
usesdoi:10.18112/openneuro.ds004457.v1.0.2
The publication analyzes bins reconstructed from OpenNeuro ds004457 version 1.0.2; the source manifest pins that snapshot and the saved metadata records its CC0 license.
doi:10.1523/JNEUROSCI.1325-22.2023

Cited by

5 publications

Public catalog entries on Sciscape.

usesA contact covariance rule kept SEEG site decisions equal across equivalent bipolar trees in five offline patient tests
We used its contact bins, pinned source metadata, fixed selection rules, and full-reader accuracy values. They defined the data and inherited comparison. The five primary path accuracies were reproduced.
usesA 90% calibration voltage map left positive test contrast outside its space in five offline SEEG sessions
Its offline accuracy comparison supplied the 90% calibration voltage rule and motivated this residual contrast audit. We reuse its cohort and window. Our metric projection measures subspace representation, not its compressed reader accuracy.
usesA five-point fourth-pulse rule lost 7.9 percentage points in one of five SEEG sessions at 20 to 100 ms
We used its result to place this follow-up in the prior evidence. It already tested voltage-basis rank selection and showed that a calibration allowance can fail on later pulses in these same sessions. Our covariance-scaled bases and separate fourth-pulse five-point rule differ. We found this publication after the current analysis. Its outputs did not select our rules or parameters.
usesMeans of four pulses reduced normalized 20–100 ms voltage error by 33.3% relative to zero across blocks at five sites in three SEEG sessions with equal site and patient weights
We used its good SEEG contact, seizure onset zone, adjacent site, current, valid epoch, and minimum pulse selection rules. Its loader code and calibrated voltage classifier specification supported the inherited pool and the clock audit replication. We test a different target: voltage forecast loss at known sites in later source blocks.
usesShared SEEG templates used 36% fewer numeric values with a 0.22% mean relative error decrease in five offline 20 to 100 ms tests
Its 90% variance comparison tested site identification in the same cohort. We read it after our numerical analysis. It defines prior task evidence and limits any claim about preserved reading accuracy. Our endpoint is later voltage forecast error.