Pulse variance correction reduced the observed mean two-pulse rank score for 100–500 ms posterior cingulate SEEG responses in eight patients

This study tested whether pulse variance correction improves selection of large recorded responses from two SEEG pulses. It used public OpenNeuro ds004696 version 1.0.1, with eight patients, 19 posterior cingulate contacts, and 536 links from 237 stimulation blocks. Each block supplied the first ten good 6 mA pulses. Each method selected the nominal top ten percent of links from two pulses. Five separate pulses supplied the corrected test scores. All 2,520 disjoint allocations were averaged within each patient. Patients then had equal weight. The uncorrected mean rank AUC was 0.93868. The corrected mean was 0.92103. The difference was -0.01765, with descriptive patient bootstrap 95 percent limits [-0.04132, 0.00268]. The exact sign-flip p value was 0.21875 under independent patients and a symmetric null. Correction reduced the score in six patients, increased it in one, and left it equal in one. The test did not establish a benefit or a population decrease. The fixed 50–250 ms check had a small positive mean difference. The result concerns recorded voltage selection with a noisy test reference. It does not test clinical benefit. The package preserves voltage bins, code, source pins, exact allocations, patient results, inference draws, and repeat instructions. All source bins were rebuilt from 57 freshly downloaded MEF objects. Source hashes and direct calculation checks passed. A full analysis repeat gave identical output bytes. Patient overlap with the earlier ds004457 study is unknown.

stereoelectroencephalography · electrical stimulation · posterior cingulate cortex · pulse variance correction · response ranking · rank AUC