Noise scaling gave 24.5 percent mean site accuracy with four calibration pulses per site on new shafts in five offline SEEG sessions from 20 to 500 ms

A noise-scaled response map gave 24.5 percent mean site accuracy in five offline SEEG sessions from 20 to 500 ms. A voltage map at the same output count gave 22.4 percent. We used OpenNeuro ds004457 version 1.0.2. The test had 182 sites and 728 test pulses. Each map excluded a whole stimulation shaft. Four local pulses per site fit candidate means; four later pulses tested them. The primary patient mean difference was 2.06 percentage points. Its descriptive 95 percent patient bootstrap interval was -0.26 to 4.39 points. The exact sign-flip value was 0.25. This test gives no clear evidence of a mean gain. A complete donor span kept every donor-only decision, but changed 174 of 728 decisions when new calibrated sites were added. Adding all local means to the span kept every full-metric decision. Before-pulse results limit a response-specific interpretation. This is a test of offline calibration, with five patients and retrospective quality selection. The package has three figures, exact compact inputs, saved maps and scores, source hashes, and code. A fresh raw rebuild matched all ten input files. An independent distance calculation matched all 3640 main decisions. A separate run matched all 196 analysis files byte for byte.

stereoelectroencephalography · electrical stimulation · site decoding · noise scaling · response maps · calibration · cross-validation