SEEG power changes gave no clear gain in five-number transfer between images and sounds on later trials in ten patients with a ridge penalty of 10

We tested five-number reading across images and sounds in public SEEG data. The source was OpenNeuro ds005415 version 1.0.0. Ten selected patients supplied 649 early training trials and 331 later test trials. A fixed ridge reader used power changes from all eligible pairs in three bands. Transfer balanced accuracy was 20.27%. A separate before-onset control gave 20.02%. The paired gain was +0.25 percentage points, with a 95% patient-bootstrap interval from -6.15 to +7.24 points. The test did not establish a gain. The interval permits gains and losses of practical size. Training accuracy was 100%, but later same-group accuracy was 17.45%. Thus, the poor later score cannot be assigned only to sensory transfer. Other fixed penalties and an eight-pair test also gave uncertain gains. This is exploratory reuse in patients with epilepsy and two fixed task schedules. It does not establish a lack of number information, freely chosen intent reading, or a stimulation result. The package preserves source ranges, all fits and predictions, three figures, and reproduction code. All 1,044 retained source ranges matched fresh public downloads. An independent analysis run matched all 190 numeric output files.

stereotactic electroencephalography · SEEG · number reading · cross-sensory transfer · brain-computer interface · ridge regression · neural decoding