Simulated region loss raised SEEG gamma recovery error over matched random loss in 12 patients under a fixed model and assumed trial order

This study tested whether the location of lost SEEG inputs changes recovery error when the lost count stays the same. It used public OpenNeuro ds006136 version 1.0.1. The test used 5,995 available channel targets from 647 trials in 12 eligible epilepsy patients. A fixed Gaussian model recovered early delay gamma power change. The split assumed that source trial IDs give trial order. Loss of all available channels in one source region raised mean squared error by 0.0632 fit target variance units over random loss of the same size. The 95 percent patient bootstrap interval was 0.0344 to 0.0959. Ten of 12 patients had greater error. Region recovery showed no clear mean error gain over the fit mean. A second set of random draws gave a difference of 0.06275. All features matched a fresh rebuild from the public signal files. The package preserves the inputs, masks, code, models, scores, figures, source hashes, and repeat instructions. The result concerns simulated loss of processed signal features. It does not establish BCI command accuracy, real device fault behavior, neural connections, or a brain stimulation rule.

stereoelectroencephalography · SEEG · brain-computer interface · signal recovery · simulated channel loss · gamma power · Gaussian model