Power ratios kept 67.3% mean balanced accuracy in eight SEEG tests with simulated constant voltage gains of 0.5 to 2 per pair
We tested power ratios for task-state decoding in eight public SEEG recordings from OpenNeuro ds004473, version 1.0.2. Early trials trained each decoder. Later trials tested it. Each bipolar pair had a separate simulated constant voltage gain from 0.5 to 2. Gains were equal across frequency bands. The ratios kept all predictions and 67.3% mean balanced accuracy. The absolute power decoder had 70.7% accuracy on original inputs. Its exact minimum under allowed gains was 49.8%. The paired difference in minima was 17.6 percentage points, with a patient sign-flip p value of 0.0078125. A descriptive 95% patient bootstrap interval was 11.1 to 25.1 points. Constant gains preserved score order, so the absolute accuracy loss concerned its fixed threshold. These results do not establish performance with noise, real session changes, or continuous brain interface use. The archive contains complete feature inputs, source records, raw signal samples, fitted models, predictions, attainable adverse gains, three figures, and reproduction code. A fresh run reproduced eight numerical result files byte for byte. Independent raw signal and exact minimum checks passed.
stereoelectroencephalography · SEEG · power ratios · balanced accuracy · signal gain · robustness · brain-computer interfaces
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
This publication supplied complete features, event alignment, exclusions, the original trial split, and source extraction code. We reproduced its eight absolute power accuracy values. Those inputs enabled the new power ratio comparison and exact gain minimum analysis.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis used SEEG recordings and task metadata from OpenNeuro ds004473, version 1.0.2.
- doi:10.1088/1741-2552%2Facae0a
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesA three-band projection kept 57.1% mean SEEG accuracy under assumed gain and slope shifts in eight recordings
The ratio comparison and exact constant-shift method inform this extension. The new test adds an assumed slope axis and permits slopes to change at each window. The cohort and test trials are reused.
- supportsCommon power and band contrast had mean AUCs of 0.750 and 0.753 in 28 locally calibrated second SEEG records outside onset shafts
This earlier motor-task test showed common-gain cancellation for a power ratio. Our separate gain proof and simulation have the same scalar-gain basis. We read it after the primary result and used no data or fitted model from it. The new evidence is a common-versus-contrast comparison in seizure-state monitors.
- usesCovariance gates increased mean rejection of faulty and recorded inputs under simulated contact gain 4 in eight offline SEEG tests
Used its bipolar-pair gain scope to define a separate physical-contact gain question. Here a gain acts before bipolar subtraction and the gate has no fault-location input. No result arrays from that study were copied.