The SEEG model gives 49.9 percent balanced accuracy for late cingulate energy in seven excluded patients under the source reference

Public SEEG mean responses were used to test transfer of stimulation region prediction across seven patients. Receiving contacts were in the posterior cingulate group. Source classes were the hippocampal region and amygdala. The hippocampal region included parahippocampal cortex. A nearest centroid model used twenty energy fractions from 100 to 500 ms. Each test excluded a complete patient from training. Mean patient balanced accuracy was 49.94 percent for 132 waveforms. Chance balanced accuracy was 50 percent. The pair permutation test gave p = 0.525. Complete patient label exchange gave p = 0.5625. The late window did not clearly outperform the window before the pulse. The result applies to this model, clinical sample, mean waveforms, and source reference. It does not show equal neural effects for the source regions. The package includes selected waveforms, code, predictions, null statistics, source versions, hashes, and reproduction instructions.

stereo-electroencephalography · electrical stimulation · posterior cingulate · amygdala · hippocampus · patient-held-out classification · balanced accuracy · response energy