Quadratic bounds reduced 70 to 110 Hz SEEG power range widths by 11.5% at 100 ms after simulated 25 ms gaps in 21 patients under equal voltage error norm limits

We tested power ranges after short gaps in public SEEG voltage. We used 35 records from 21 patients in OpenNeuro ds006519 version 1.0.0. The feature used a causal 70 to 110 Hz filter and a trailing 100 ms mean of squared output. Each simulated gap had a nominal length of 25 ms. We held the last voltage sample during the gap. We compared two bounds with the same limit on the missing voltage error norm. The quadratic bound used the direction of the filtered signal. The triangle bound used the largest filter gain. At 100 ms after the final missing sample, the quadratic range was 11.5% narrower in the equal patient mean. The descriptive 95% patient resample interval was 10.3% to 12.7%. This result used the source post windows and the early calibration limit R. The quadratic bound first met a planned error limit 23.7 ms sooner on average. This limit was 10% of the earlier mean power. The voltage assumption failed in 33 of 1100 later test cases. One case also exceeded 4R. Thus early calibration did not give an unconditional power guarantee. The package gives a proof, tested code, exact selected voltage, source hashes, result rows, and three figures.

SEEG · signal processing · power bounds · missing input · quadratic optimization · causal filtering · brain-computer interfaces