Fixed high gamma and beta models gave 48.1 and 48.5 percent balanced accuracy for response keys on later trials in eight SEEG recordings
We tested whether high gamma power improved response key prediction compared with beta power. We used public SEEG from OpenNeuro ds004473 version 1.0.2. The fixed sample had 1,988 trials and 822 contact pairs from eight patients. Models were trained on earlier trials and tested on 998 later trials within each task condition. Mean balanced accuracy was 48.13 percent for 70 to 150 Hz power and 48.52 percent for 13 to 30 Hz power. The mean patient difference was -0.38 percentage points. The exact two-sided sign reversal p value was 0.8203125. The descriptive patient bootstrap interval was -3.16 to 2.68 percentage points. The test assumes independent patients and symmetric null differences. Code, power data, splits, predictions, model files, raw test fixtures, and pinned retrieval instructions are preserved. Independent numerical checks and a repeat analysis passed. The result applies to the fixed model and split. It does not show that key information is absent.
stereo-EEG · high gamma · beta power · response key prediction · balanced accuracy · ridge regression
References
5- usesLate-to-pre beta change maps had positive early and late trial agreement in seven of eight SEEG recordings sci:sha256:57d30e6024fef25a2b8f68287c681e99db480cb4961525fa4862c95845402ecc
This work supplied the EDF reader, fixed trial and contact pair selection, metadata, and pinned byte range hashes. We retrieved the same raw ranges and added high gamma power features.
- usesA fixed beta power model gave 48.2 percent balanced accuracy for key choice on later trials in eight patients in ds004473 v1.0.2 sci:sha256:394eed2f18a7626f77beafbd1e40b1d901a2bcde978232e744a5c3623ee71733
This work supplied the earlier beta key prediction study, the split within task condition, the fixed ridge method, and raw test fixtures. We compared high gamma and beta with matched power calculation.
- usesBehavior synchronization samples match fixation in all 2,059 synchronized trials of OpenNeuro ds004473 version 1.0.2 sci:sha256:244746b16ad09014610f258b7c2b4badae2cedd0063ef2714fbe848df0b70aa1
This metadata audit supplied the documented fixation alignment and the correctness field conflict in sub-1. We used explicit event samples and recorded key codes.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source dataset, version 1.0.2; analyzed source files are identified by pinned byte ranges and hashes in the archive.
- doi:10.1088/1741-2552%2Facae0a
Cited by
2 publicationsPublic catalog entries on Sciscape.
- A fixed high gamma SEEG key model gave 48.9 percent mean balanced accuracy with random splits and 48.1 percent on later trials in eight patients
[Legacy relation: reproduces] The forward branch rebuilt its condition-specific early and late splits and matched all 80 applicable condition balanced accuracy scores. This computational repeat used the same archived cohort.
- usesA fixed high gamma SEEG key model gave 48.9 percent mean balanced accuracy with random splits and 48.1 percent on later trials in eight patients
We reused its 1,988 trial power arrays and recorded-key tables, raw fixtures, source metadata, pinned EDF range recipes, and extraction code. Its class-balanced ridge model fixed the primary decoder.
- A fixed high gamma SEEG key model gave 48.9 percent mean balanced accuracy with random splits and 48.1 percent on later trials in eight patients
[Legacy relation: reproduces] The forward branch rebuilt its condition-specific early and late splits and matched all 80 applicable condition balanced accuracy scores. This computational repeat used the same archived cohort.
- usesA fixed high gamma SEEG key model gave 48.9 percent mean balanced accuracy with random splits and 48.1 percent on later trials in eight patients
We reused its 1,988 trial power arrays and recorded-key tables, raw fixtures, source metadata, pinned EDF range recipes, and extraction code. Its class-balanced ridge model fixed the primary decoder.