The earliest peak rule gave a mean error of -2.32 ms after six-bit encoding of 140 five-pulse SEEG means from eight patients

We tested scalar encoding of public stereoelectroencephalography (SEEG) response means from eight patients with epilepsy. An early check selected 140 bipolar readout pairs. Each later waveform was a mean of five separate pulses in the 20 to 500 ms interval. The encoding range was 1.25 times the early mean absolute peak. At six bits, the earliest tied peak had a mean signed time error of -2.32 ms relative to the full-precision later waveform. The center of the tied time range reduced mean absolute error from 2.32 to 1.36 ms. Six patients had smaller errors and two had larger errors. The planned two-sided sign-flip test gave p = 0.078 under independent patient and symmetry assumptions. The tied range contained the full-precision peak in all 6,202 tested waveform and encoding conditions. A monotone magnitude code explains this result. The range center gives the smallest worst-case absolute time error over the tied set. A range width of at most 20 ms bounds this encoding error by 10 ms. This bound held for 134 of 140 selected means at six bits. The package contains figures, exact mean arrays, source hashes, code, a proof, and retrieval instructions. The result applies to encoding after pulse averaging and time centering. It does not measure neural transmission time.

stereoelectroencephalography · SEEG · scalar encoding · peak timing · quantization · intracranial EEG