An offline two-pulse rule reduced 20–500 ms SEEG map error in two of five people at an equal pulse count

We tested whether two initial SEEG responses can guide a fixed pulse budget for a response map. We used 183 sites and 15 recording pairs from five people in OpenNeuro ds004457, version 1.0.2. The rule assigned two to six pulses per site. The uniform rule assigned four. Each rule used 732 pulses. Two separate later pulses per site gave 366 test pulses. For centered voltage from 20 to 500 ms, the mean error reduction across people was 6.99%. Error fell in two people and rose slightly in three. The median effect was minus 0.39%. The exact two-sided person sign-flip value was 0.50 under a symmetric null effect assumption. The rule reduced mean error by 10.19% relative to random placement of the same unequal counts. These results do not establish a general benefit. The integer rule minimizes a fixed variance score. Its expected-error proof needs stable responses and independent trial noise with known variances. The measured rule uses noisy estimates. Pulse times interleaved across sites, so this was an offline test. We verified 110 public source files and rebuilt all ten input files with identical hashes. The package preserves three figures, code, data, source records, and repeat-run checks. It also marks waveform and pulse-sequence checks added after the main result.

SEEG · pulse allocation · response mapping · electrical stimulation · variance reduction