Distance gave AUC 1.0 for hippocampus versus amygdala stimulation in four selected SEEG patients under a 10 mm rule

We tested response shape transfer between patients in public SEEG dataset ds004457, version 1.0.2. Both stimulated contacts had to have the same hippocampus or amygdala source label. The planned 20 mm minimum distance removed all hippocampal sites in two patients. The primary four-patient test could not run. The planned 10 mm check retained 28 sites and 86 site-contact links in four patients. Each model fit excluded one whole patient. A distance model gave site AUC 1.0 in each excluded patient. Mean AUC was 0.49405 for normalized shape at 100 to 500 ms. The matched before-pulse shape gave mean AUC 0.53423. An added fixed distance score gave mean AUC 0.98810 without training. Ten of 14 receivers had separate regional distance ranges. Added waveform models gave higher AUC, which shows dependence on the time window and sign rule. These results apply to a selected epilepsy cohort and fixed methods. They do not estimate a region effect independent of geometry. All 57 raw source files passed hash checks. All rebuilt epoch arrays matched exactly. All 55 repeated numerical result files matched byte for byte. The package preserves epochs, provenance, code, exclusions, predictions, models, figures, and reproduction instructions.

SEEG · stereoelectroencephalography · electrical stimulation · hippocampus · amygdala · response shape · patient exclusion · AUC · contact geometry