Mean late response map agreement over 35 pulse splits was lower than chronological agreement in five SEEG patients

We tested pulse split choice in public SEEG data from five patients in OpenNeuro ds004457 version 1.0.2. Each of 314 edges used its first eight eligible pulses at one constant current. We calculated response energy from 50 to 250 milliseconds. We compared chronological halves with all 35 unordered partitions, using common pulse positions across each patient's edges. Mean Spearman agreement was 0.488 over the 35 splits and 0.516 for chronology. The mean difference was -0.028, with a descriptive 95 percent t interval from -0.074 to 0.019. The difference was positive in two patients and negative in three. The exact two-sided patient sign-flip value was 0.25 under independence and null symmetry. The result does not support a consistent increase from random position splits. It does not prove equivalence or identify a neural mechanism. This is a secondary analysis of the parent clinical sample. The package preserves all maps, pulse selections, code, source hashes, results, and instructions. Independent calculations matched every main score. A fresh rebuild matched all 10,161,994 array elements from 54 public raw objects. A mathematical check gives the joint exchangeability condition for a zero expected split difference.

stereoelectroencephalography · electrical stimulation · pulse split · response map · rank agreement · reliability