With a Hann window, three phase-agreement features did not lower mean later-trial Brier error in eight selected SEEG records
We tested whether three phase-agreement values improved a public SEEG reader beyond band power. The reader used fixed selected contacts and earlier-to-later trials from eight people. Mean later Brier error was 0.181243 with power alone and 0.183465 with a rule that could keep or remove the added values. The paired change was +0.002223 [-0.000364, +0.006630]. The exact two-sided sign-flip p value was 0.50. The added block was removed in three records. A rectangular window gave a small mean benefit; its interval also included zero. A separate check found positive excess agreement in independent white noise after a Hann window. The phase reference does not preserve the relation between nearby Fourier bins. Positive values do not establish a brain connection. The package contains selected samples, source hashes, code, model fits, later predictions, phase controls, four figures, and reproduction instructions. The cohort and test period have prior results. This exploratory study does not establish live interface performance or a stimulation effect.
stereoelectroencephalography · brain-computer interface · phase agreement · band power · Brier score · feature selection
References
6- usesSeparate power and phase penalties reduced mean Brier error against a shared penalty on later trials in eight selected SEEG records sci:sha256:52c83dfcc8e81bd04f93edff10c72e5c45f7632911769c2f4448175b2904737c
This study uses separate power and phase feature penalties and the rule that can remove the added block. These rules control the cost of extra features.
- usesAdded imaginary coherency increased mean Brier error for tuned linear readers in eight SEEG records with fixed 500-sample inputs sci:sha256:2263fe7cecddd0d4616d9f4cfb58a07f322990304de5c63f2a497260c14e6bb5
Its test of many imaginary coherency values motivated the test of a smaller nonlinear phase representation with the same power bands and later probability error.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
This study retains the original contact and trial selection. It uses the same rest and before-response labels in eight records.
- usesTwo 250-sample spectra gave lower mean Brier error than one 500-sample spectrum in eight reused SEEG records with equal sample weights sci:sha256:4e43a4b6020c3f64befdbb992e016a7bc61c3ed8e557a01c7102221b1c1863ae
This study uses the exact selected signal samples, source metadata, source-range recipe, fixed trial splits, and numerical helpers from this publication. The new fits use three phase-agreement features.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The study uses OpenNeuro ds004473 version 1.0.2 as its source dataset.
- doi:10.1088/1741-2552%2Facae0a
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.