Power contrasts preserve fixed decisions under simulated broadband gain changes in eight SEEG recordings
We tested fixed SEEG decoders in eight recordings from OpenNeuro ds004473 version 1.0.2. Each model used eight bipolar pairs to classify states before a response and selected rest states. We changed one selected input at a time to half or twice its voltage. The gain was equal across frequencies and fixed across the test. Power contrasts kept every decision at both fixed thresholds. The mean minimum balanced accuracy over 16 gain tests was 67.03% with contrasts and 55.53% with absolute power. The paired benefit was 11.50 percentage points, with a descriptive 95% person-bootstrap interval of 8.24 to 15.10 points. All eight differences were positive. Mean clean accuracy was 3.12 points lower with contrasts. All gain-test AUC values stayed equal to their clean values, including for absolute models whose decisions changed. The package has a proof, four figures, code, features, scores, source versions and hashes, and reproduction commands. A complete raw rebuild matched every feature array. This new analysis reuses an existing public sample. It tests simulated gain in selected offline windows. It does not establish live control or robustness to all signal drift.
stereoelectroencephalography · brain-computer interface · power contrasts · gain robustness · balanced accuracy · signal processing
References
3- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
We reused all eight SEEG feature arrays, trial exclusions, raw audit ranges, source records, and extraction code. These files enabled the fixed gain test. We verified their archived hashes and rebuilt every feature array from pinned public raw ranges.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis uses all eight recordings and task metadata from this version of OpenNeuro ds004473.
- doi:10.1088/1741-2552%2Facae0a
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesAdded spatial power gave no clear mean test error decrease with pairs selected from band contrasts in eight SEEG records
We used the separate bipolar gain model and fixed band contrast method to define the comparison and gain stress tests.
- usesCovariance angles kept 37.9 percent mean site accuracy under nine simulated positive common voltage gains in five offline SEEG sessions
This earlier fixed gain study provided a reference for related methods and their accuracy cost. It used power contrasts for a different task and dataset. No source signals or model arrays entered this analysis.
- usesMean removal before synthetic linear gain ramps reduced decision changes in eight fixed SEEG contrast readers
Its constant broadband gain invariance result informed the control and the scope boundary. The new result tests linear gain ramps and the order of mean removal.
- usesTraining for ±6 dB gain on one shaft raised correct classes with a guarantee in eight SEEG records under a linear gain model
This earlier study proved gain invariance for power contrasts. We use that result to explain the invariant subspace reached by the bounded gain loss. We do not claim a new power-contrast theory.
- usesA rest reference based on the mean of three power bands gave 3.21 percentage points higher mean later SEEG balanced accuracy than full subtraction of one rest window in eight recordings
Its power projection and gain study informed the fixed spectral contrast comparison and the limits of the gain claim. We added a rest reference based on the mean across bands.
- usesAdded spatial power gave no clear mean test error decrease with pairs selected from band contrasts in eight SEEG records
We used the separate bipolar gain model and fixed band contrast method to define the comparison and gain stress tests.