A fixed SEEG power model improved four versus eight letter ranking beyond history after correct responses in 13 final sessions
We tested a fixed SEEG power model in OpenNeuro ds004752 version 1.0.1. Each of 13 epilepsy patients had all earlier sessions used for training and the final session used for testing. The primary sample used trials after a correct response to limit the task rule after errors. Task history alone gave a mean area under the receiver operating characteristic curve (AUC) of 0.478. Adding memory-period depth power gave 0.593. The paired AUC gain was 0.115, with a 95 percent patient-resample interval from 0.036 to 0.197. Two patients had incomplete source retrieval and were excluded before decoding. The package includes features from 64 sessions, source hashes, event files, all test probabilities, code, and four figures. The label is assigned letter count. Offline artifact flags and assumed task timing define the tested scope. This result does not establish an online brain interface or a brain stimulation effect.
SEEG · verbal working memory · memory load · intracranial EEG · session transfer · ridge logistic regression · area under the curve
References
2- usesdoi:10.18112/openneuro.ds004752.v1.0.1
The analysis used OpenNeuro ds004752 version 1.0.1 to compute depth power features and test letter-count readout in final sessions.
- supportsdoi:10.7554/eLife.78677
The source task paper supplies the task phase timing and the task rule after errors described in this report.
Cited by
14 publicationsPublic catalog entries on Sciscape.
- usesWith fixed epoch timing, a mean of fixation log power over eight trials showed no clear letter load AUC gain in 15 final SEEG sessions
64 derived feature files, source metadata, source manifest, EDF headers, full hash records, extract.py, download_parts.py, and four unchanged model functions in base_model.py from code/analyze.py.
- usesUnder assumed timing, eight SEEG features increased response time forecast error in 15 final sessions at ridge penalty 10
The inherited EDF helper and adjacent-contact power rule came from this method. Its fixation and maintenance arrays for 64 sessions entered through the direct feature archive. We used the helper to check fresh source ranges, and separate routines checked spectra and regression fits.
- usesUnder assumed timing, eight SEEG features increased response time forecast error in 15 final sessions at ridge penalty 10
The inherited EDF helper and adjacent-contact power rule came from this method. Its fixation and maintenance arrays for 64 sessions entered through the direct feature archive. We used the helper to check fresh source ranges, and separate routines checked spectra and regression fits.
- usesSelection by matching effect signs showed no clear decrease in load probability error in 12 final SEEG sessions under assumed timing
This work supplied the fixed ridge logistic functions used for all fitted models. Both feature rules used the same penalty, training scale, and model fit. It also supplied prior readout results on this reused cohort.
- usesUnder assumed timing, 34 of 96 single labels were wrong for 90% target load sets in 12 final SEEG sessions
This source supplied the fixed logistic functions. These functions selected training features and fitted the readers before separate session calibration.
- usesA fixed SEEG model of letter load showed no clear AUC gain from calibration across earlier sessions at matched trial counts in ten patients
This work uses the exact frozen band-power features, source metadata, source hashes, and extraction code. It tests calibration across earlier sessions at matched trial and class counts in ten eligible patients.
- usesUnder assumed timing, tuned SEEG forecasts had higher mean loss than task forecasts in 12 final numbered sessions
Use its original Newton ridge solver and band power method through the parent package.
- usesCovariance correction from 12 early trial positions showed no clear load AUC gain in 15 final SEEG sessions with assumed timing
We used its fixed training feature selection and ridge logistic fitting functions through the direct parent code. We kept these settings fixed when testing correction without new task labels.
- usesUnder assumed timing, fixation subtraction showed no clear load probability error reduction with shared features in 15 final SEEG sessions
We used the unchanged logistic model functions inherited through the immediate source package. This work supplied the fixed ridge fit and depth power method.
- usesUnder assumed timing, SEEG load readers showed no clear AUC difference between hippocampus and temporal cortex in 12 final sessions
This study uses its ridge model functions and after-correct trial rule through the cited feature package.
- usesWith shared features and assumed timing, maintenance fitting showed no clear load AUC gain over encoding fitting in 15 final SEEG sessions
We used the original power extraction method and unchanged ridge model functions through the direct parent package.
- usesWith assumed timing and ridge penalty 10, eight SEEG features raised mean error forecast loss in 15 final numbered sessions
We used the original ridge solver and signal extraction method through the parent package.
- usesWith assumed timing and ridge penalty 10, eight SEEG features raised mean error forecast loss in 15 final numbered sessions
We used the original ridge solver and signal extraction method through the parent package.
- usesUnder assumed epoch timing, eight added high gamma features changed mean load AUC by 0.002 in 13 final SEEG sessions
We reused its 56 complete-session feature files for 13 patients, source metadata and extraction receipts, history controls, session split, and fixed logistic fit helpers. We tested whether eight high gamma features add AUC beyond eight low band features. We also repeated its history and mixed-band probabilities on the same final-session data.