Four cutoff pulses for each site gave 75.0 percent mean coverage at an 80 percent target in five SEEG sessions at 20 to 100 ms under a replay that favored sites with high fitting risk

We tested whether four cutoff pulses for each site improve a stimulation response reader when use favors sites with high fitting risk. The public source was OpenNeuro ds004457 version 1.0.2. The offline test used 183 sites, 15 recording pairs, 366 separate test pulses, and five patient sessions. Two pulses per site fit the reader. Four other pulses set the cutoffs. The final two pulses tested it. Replay weights came only from the fit pulses. At an 80 percent target and 20 to 100 ms, mean coverage was 74.97 percent with a cutoff for each site and 78.63 percent with a pooled cutoff. Each patient had equal weight. The paired change was -3.66 percentage points. Its descriptive 95 percent patient bootstrap interval was -12.28 to 2.77 points. The study does not show a clear coverage gain. Sets retained 31.38 and 29.85 percent of candidate sites, respectively. At a 90 percent target, the tested separate rule returned all sites because four cutoff scores require rank five. This is a follow-up in the same sessions used by prior work. Most test pulses overlap fitting at other sites in global time. Exchangeability was not established. The weighted replay does not change stimulation or physiology. A fresh raw rebuild matched every input array. Independent calculations and a second process matched the results. The package preserves three figures, code, input data, model arrays, pulse decisions, source hashes, and reproduction instructions.

SEEG · stimulation response · prediction sets · coverage · site classification · conformal prediction