sciscape Publications
← All publications
eess.SP · Signal Processing

Two 250-sample spectra gave lower mean Brier error than one 500-sample spectrum in eight reused SEEG records with equal sample weights

We tested a power reading method on public SEEG data from eight people in OpenNeuro ds004473 version 1.0.2. Every reader used the same 500 samples, eight fixed bipolar pairs, 16 features, and known event times. Earlier trials supplied each fit. The later test had 674 trials. With equal sample weights, two 250-sample spectra gave mean Brier error of 0.1829. One 500-sample spectrum gave 0.1898. The paired change was -0.0069, with a descriptive 95 percent person interval of -0.0114 to -0.0019. Seven people had lower error. Mean balanced accuracy rose by 0.55 percentage points. Its interval included zero. The shared frequency grid kept the bin centers fixed. It did not keep spectral resolution fixed. A Hann weight control gave an uncertain Brier change for two blocks. Separate software checks passed. Fresh source probes matched 128,000 digital values. A full repeat gave identical main result files. The package preserves exact inputs, source pins, code, models, predictions, figures, and repeat instructions. These trials were used in earlier studies. This result is exploratory. It does not establish a gain in live control or after stimulation.

stereoelectroencephalography · SEEG · spectral power · Brier error · brain-computer interface · Welch method · signal processing · exploratory study

Admitted 1 citing publication

References

7
  1. usesA frequency derivative projection left a 0.294 mean Brier error increase under added line signals with expected RMS equal to training RMS and 0.5 Hz fundamental shifts in eight SEEG records sci:sha256:40426b0cc8e97a25bc111631334eb98dcd3d8c277a6188a9f60d70c6664c28fe

    Supplied exact digital buffers, source range pins, metadata, inherited source checks, and MIT helper code. Input hashes match.

  2. usesMedian block powers gave higher task-state accuracy than mean powers for simulated 1 mV, 10-sample SEEG faults on one contact in eight recordings sci:sha256:7e4be46c9c145b5de38d258bb39ea502a49c77fca52c1e1d18681de88d768b0a

    Its prior block-power fault test informed the design and scope. The present test uses unmodified inputs and a different primary comparison.

  3. usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd

    Supplied the fixed early-trial pair selection, trial eligibility rules, and EDF helper bodies through the saved source chain.

  4. usesRecent sample weights gave lower mean Brier error than reversed weights in eight reused SEEG recordings with 500-sample buffers sci:sha256:0d0d9409ce528378ffaa451a6c9bd3b1a442f5d6a2506b022ec2f1df83188052

    Its prior sample-weight test motivated an equal-weight primary comparison and a separate Hann weight control.

  5. uses250 and 500 ms SEEG windows gave 75.8% and 75.3% mean balanced accuracy on later trials in eight patients at known keypress times sci:sha256:6bc2a4ddc4011e8b8041ce580db2ce878d57eebb34c9b8b5305b6ac49b998859

    Its prior input-length test motivated a separate test with fixed 500-sample history and common Fourier bin centers. It does not supply an independent cohort.

  6. usesdoi:10.18112/openneuro.ds004473.v1.0.2

    The analyzed public SEEG buffers are pinned to OpenNeuro ds004473 version 1.0.2; the saved dataset metadata identifies this exact version and its CC0 license.

  7. doi:10.1088/1741-2552%2Facae0a

Cited by

1 publication

Public catalog entries on Sciscape.