Early fitted partial state correction lowered expected 70 to 110 Hz error after 13-sample gaps in baseline SEEG from 20 patients with independent endpoint errors of one early voltage RMS
We tested whether a partial filter state correction fitted to early SEEG could lower later error after known gaps. The input was public baseline SEEG from OpenNeuro ds005169 release 1.0.0: 20 patients and 1,671 bipolar channels. The test used a forward 70 to 110 Hz filter and 13-sample gaps at 4096 Hz. The first 16 gaps fitted one weight per patient. The last 32 gaps tested it after offline replay of the early record. The main model assumed independent zero mean returned-sample errors with RMS equal to early channel voltage RMS. Mean patient expected normalized squared error was 0.458334 with hold, 0.449473 with fitted correction, and 3.652747 with full correction. The fitted rule reduced error by 1.93% versus hold and 87.69% versus full correction. Ten weights were zero. The other ten reduced later error versus hold. Doubling endpoint error made all ten fixed nonzero weights worse than hold. This is a small benefit under a specified error model. It does not show recovery during delivered stimulation or better decoding. The package preserves exact voltage arrays, a streaming tool, an expected error proof, all scores, three figures, source hashes, and repeat instructions. All supplied voltages matched complete original source files. A second analysis matched all numerical results.
stereoelectroencephalography · SEEG · filter state correction · signal processing · endpoint error · expected error · brain-computer interface · missing samples
References
3- usesA voltage gate reduced 70–110 Hz SEEG repair error but exceeded hold error with simulated endpoint faults equal to voltage RMS in 20 patients sci:sha256:e73fab899791a78e68ec9fd8c6f96201d5a9b35ceb9e6815f3609b48270c2d81
The earlier voltage gate result showed that an endpoint check could reduce full correction error but did not beat hold under its main endpoint fault. This known result helped define the question about partial correction. Its numerical scores were not used to fit weights or to make a direct method comparison.
- usesState correction reduced 70 to 110 Hz error after simulated 13-sample gaps in baseline SEEG from 20 patients with clean endpoints sci:sha256:6e79fa2c0f3fd3bf99e67f05cc78959b6a8c7ec430c7389a9ebd20941a5d553e
We reused all 20 baseline SEEG arrays, 1,671 channel selections, 48 gap times per patient, source pins, metadata, and retrieval code. The same cohort supported the earlier filter state correction result. This study adds early fitted partial weights, an exact expected error calculation, and fixed later tests. It is not an independent cohort replication.
- usesdoi:10.18112/openneuro.ds005169.v1.0.0
The analysis reused baseline SEEG from OpenNeuro ds005169 release 1.0.0; the source metadata and supplied voltage arrays are pinned to this exact version.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesWith clean endpoints, delayed interpolation gave lower 70–110 Hz error than a causal model for simulated 3.17 ms SEEG gaps in 18 patients
The prior study informed the question about filter state error after known short gaps. The present study tests earlier fitted causal prediction in ds006519 with accurate recorded endpoints. It uses different data and has a different endpoint error assumption.
- usesA weight from other patients did not lower observed late SEEG power forecast error in 20 patients
Read after score computation to compare study scope. It tests partial filter state correction on baseline data with simulated gaps and faults. The present test uses observed post-stimulation power and fitting in other patients.