Four 20 ms SEEG responses gave higher mean site accuracy than one 80 ms response at equal sample counts in five offline 6 mA sessions with a fixed site per decision

We tested a fixed SEEG response sample count in five public sessions from OpenNeuro ds004457 version 1.0.2. The analysis used 138 remote 6 mA stimulation sites. The first four valid pulses at each site fit the reader. The last four valid pulses tested it. Four 20 ms responses gave 41.46% mean site accuracy. One 80 ms response gave 33.71%. Each design used 164 native samples per recording pair per decision. The mean difference was 7.75 percentage points; the exact patient sign-flip test gave p = 0.125. These five patients do not establish a population benefit. The median four-pulse span was 14.34 seconds. Pulse time alone gave 82.35% accuracy, which shows a strong schedule limit. The reader assumed a fixed site within each decision and a known pulse onset. The result is an offline cost benchmark. It does not show online control. The package preserves derived data, all decisions and models, three figures, source hashes, code, and reproduction instructions. Fresh raw reconstruction and independent numerical checks passed.

SEEG · electrical stimulation · response decoding · sample budget · repeated pulses · site classification · offline evaluation