Contact normalization kept SEEG site decisions fixed under simulated five percent contact gain changes but reduced accuracy in five offline sessions
Normalizing each contact before bipolar subtraction removed sensitivity to simulated contact gain changes. It also reduced site accuracy in five public SEEG sessions. We compared two fixed cosine readers on OpenNeuro ds004457 version 1.0.2. The test used 183 stimulation sites, 732 calibration pulses, and 732 separate query pulses. With the source reference, mean accuracy fell from 35.8% to 23.4%. These means give each patient equal weight. All normalized decisions stayed fixed under every joint five percent gain corner. The raw reader had 30.4% mean minimum accuracy over fixed session gain vectors. A shared reference change reduced normalized accuracy to 17.7% and changed 427 of 732 decisions. Raw decisions stayed fixed under that reference change. We provide an exact positive gain proof and a scoped two-contact limit. We also provide independent numerical checks, three figures, contact voltage bins, fitted readers, code, source hashes, and retrieval instructions. Fresh extraction reproduced all five archived input arrays. The split was within sites and shared source blocks. The analysis did not isolate neural signals or test a live brain interface. The result supports separate gain and reference tests when selecting preprocessing.
stereoelectroencephalography · contact normalization · bipolar signals · gain invariance · reference sensitivity · site classification
References
8- usesPulse time alone gave 74.8% mean site accuracy in five offline SEEG tests with shared calibration and test blocks sci:sha256:07f3c9173d8e02cafb724475eb4d7d43bb6e6bf6e042867f73744b4ea8895b75
Supplied the prior block and pulse-time control that limits interpretation of site classification in this cohort.
- usesSeparate pair normalization kept more SEEG site decisions correct under independent constant positive output gains within ±6 dB in five sessions sci:sha256:049b6cc7ffa68c48b5d95649d4fb2722cd461d48c65f9cb1d8cd239f2c68d050
Supplied the separate-pair gain normalization rule and the per-pair covariance rule through the input code. Its gain acted after bipolar subtraction. The present work tests contact gains before subtraction.
- usesSimulated joint contact gain errors within five percent changed 172 SEEG decisions that passed separate contact tests in five offline sessions sci:sha256:a275c0fdc487a0b6b3de7d9e6e2bfa37e31bb06e78fe40b79ac5691195df831b
Showed the need to test joint contact gains before subtraction. Its affine reader and exact margin bounds differ from the cosine readers in the present comparison.
- usesNoise correction did not increase mean eight-site SEEG accuracy at 20–100 ms with four calibration pulses per site in five sessions sci:sha256:eef87e94853fad626caddee9fff3146fa92c84241f432ff4a03376aaf5f27399
Supplied the original contact selection and source bin workflow through the input publication. The present work keeps the inherited patient cohort and pulse allocation.
- usesA common time split supported at most four stimulation sites in five SEEG sessions with four calibration and four test pulses per site sci:sha256:095aa8e1bce3412fc1950ab5f561b5aebe376e44dc239f92addc13a3b59630b7
Supplied the prior session timing audit. The present allocation check also finds 696 queries before the final calibration pulse.
- usesAbsolute pair scores kept SEEG site decisions fixed under simulated output sign reversals in five sessions sci:sha256:8a26582c314b8be03ef05d9206ae910fa50daac8c1ccee11c5df2d7794d6821e
Supplied contact bins, source metadata, retrieval and extraction code, cohort selection, covariance cosine code, and nominal baseline predictions. The new raw reader reproduced all 732 nominal predictions.
- usesdoi:10.18112/openneuro.ds004457.v1.0.2
The analyzed contact voltage data and event records came from OpenNeuro ds004457 version 1.0.2.
- doi:10.1523/JNEUROSCI.1325-22.2023
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.