High gamma reduced paired window test loss after a voltage limit of three robust standard deviations in eight SEEG recordings

We tested whether high gamma improves paired task window ranking after a fixed voltage limit. We used public OpenNeuro ds004473, version 1.0.1, with eight patients and 2,030 trials. Each bipolar window was limited to its median plus or minus three robust standard deviations. Models used early trials for training and later trials for testing. Adding 70 to 110 Hz power to 13 to 30 Hz power reduced mean test loss from 0.4722 to 0.3939 nats. The equal-patient mean gain was 0.0783 nats per pair. All eight patients had a positive gain. The descriptive 95% patient bootstrap interval was 0.0454 to 0.1145 nats under independent, exchangeable patient sampling. The exact two-sided patient sign test gave p = 0.0078125 under independent patient effects and a symmetric joint zero null. The voltage limit changed 1.24% of sample values with equal patient weights. All fixed sensitivity settings had a positive mean gain. A synthetic beta burst showed that the limit can create high gamma power. The result does not identify the neural source of the gain. This processing study reuses the cohort and test trials from earlier work. It is not an independent cohort replication. The package preserves new limited features, raw fixtures, code, source versions, hashes, models, results, and repeat instructions. Fresh raw range checks, separate numerical checks, and a complete offline repeat passed.

stereoelectroencephalography · SEEG · high gamma · beta power · voltage limiting · paired ranking · signal processing