State correction reduced 70 to 110 Hz error after simulated 13-sample gaps in baseline SEEG from 20 patients with clean endpoints

We tested filter state correction at the first sample after a known gap. The data were public baseline SEEG from 20 patients and 1,671 bipolar channels. Each main test gap removed 13 samples at 4096 Hz, or 3.1738 ms. With unchanged recorded endpoints, mean patient median 70 to 110 Hz error at observed samples fell from 51.06% of reference RMS with hold to 25.40% with correction. The paired decrease was 25.66 percentage points, with a descriptive 95% patient bootstrap interval of 22.89 to 29.26 points. All 20 patients had lower error. The correction needs only the first endpoint and keeps prior emitted outputs fixed. A proof and numerical checks showed that it matches retrospective line filtering at observed samples. An added endpoint error of one early voltage RMS made correction worse than hold in every patient. All 20 original source hashes and 10,096,735 saved voltage values matched. An independent filter recurrence and a fresh repeat passed. The package preserves the streaming tool, inputs, provenance, tables, four figures, and reproduction instructions. The patients and data are reused. The gaps were simulated in baseline without delivered stimulation. The test does not establish decoding, hardware, or clinical performance.

stereoelectroencephalography · signal processing · filter state correction · missing samples · neural recording · endpoint error