Maximum-error selection increased later worst-pair error in five-pulse SEEG maps from six epilepsy sessions

We tested two rules that select a stimulation response map from five early SEEG pulses. The rules used the same model choices. One rule used average recording-pair error. The other used maximum recording-pair error. Five later pulses tested each map from 20 to 500 ms. The test used 66 sites in six people with epilepsy. Maximum selection increased later worst-pair error in all six people. The equal-person average increase was 4.35%. The 95% person bootstrap range was 1.63% to 8.28%. Maximum selection also increased average pair error by 12.75%. It selected zero-response maps at 21 sites, compared with 3 sites. The package contains source records, exact voltage products, native examples, code, results, and three figures. This exploratory result applies to one session per person and the tested model choices. It does not establish clinical performance.

SEEG · response prediction · maximum-error selection · stimulation · epilepsy