Three-pulse RMS bounds gave mean coverage of 72.2 percent for later SEEG pulses and 38.1 percent for five-pulse sets at 20–500 ms in eight sessions
We tested fixed RMS forecast error bounds in public SEEG data from OpenNeuro ds004696 version 1.0.1. The test used eight patient sessions, 86 stimulation sites, and 886 stimulation-to-readout pairs. At each site, two early pulses set the forecast and scale. The next three set a maximum error bound across all selected bipolar readouts. The final five good source pulses tested the same bound. Voltage was centered within 20 to 500 ms after each pulse. With equal patient weights, coverage of all selected readouts was 72.2 percent for one later pulse (95 percent patient bootstrap interval 64.9 to 77.6) and 38.1 percent for all five later pulses (25.1 to 49.1). Separate readout bounds covered all readouts on only 17.0 percent of later pulses. The joint rule allowed larger RMS error; it did not change the forecast. With distinct exchangeable calibration and future scores conditional on training, the maximum of m calibration scores covers H future scores with probability m/(m+H). This gives references of 75 and 37.5 percent here. The data do not prove exchangeability or online coverage. We rebuilt all eight public processed derivatives with exact array matches and checked 44 public metadata files. The package preserves code, source hashes, exact data products, a fixed analysis record, result tables, three figures, rank checks, and reproduction instructions. The result shows why a brain-computer interface coverage claim must state its readout set and future pulse count.
stereoelectroencephalography · conformal prediction · forecast error · coverage · electrical stimulation · brain-computer interface
References
5- usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
This work uses the original bipolar product extraction method and archived MATLAB source documentation. They define the eligible readouts, source reference operations, and exact fixed-window waveform product representation.
- usesAn offline two-pulse site rank did not show lower 20–500 ms SEEG forecast error at a fixed budget in eight sessions sci:sha256:578806530d9c396756538b53e11352e83c900274801f852ef7211403fa594c3c
This test uses the unchanged 24 patient files, corrected chronological pulse maps, exact Gram products, source metadata, and retrieval code. The new outcome is fixed forecast bound coverage across readouts and future pulses.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analysis uses public processed derivatives and good-channel and good-pulse labels from OpenNeuro ds004696 version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
- doi:10.1080/01621459.2017.1307116
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesSite block pooling increased three-pulse SEEG coverage and mean bound score at 20–500 ms in eight patients
We used its 24 exact numerical input files, source retrieval code, audit code, and corrected chronological pulse maps. We changed the target to the final three good pulses and tested transfer from other patients.
- usesWith fixed forecasts, limits from three recent scores covered 13.5 percent of five-pulse SEEG runs after contact order reversal in eight sessions
This method supplies the fixed two-pulse forecast, fixed scale, joint RMS score, three-score calibration maximum, and five-pulse coverage benchmark. We test a recent-score update rule on a different public SEEG cohort and add a sequence subset proof.
- usesAt 20 to 500 ms, a rule for one pulse accepted all five later SEEG responses at 53.4 percent on average in eight epilepsy patients
Used its earlier distinction between pulse and group coverage and its fixed-score group bound to define this symmetric group analysis.
- usesSEEG bands with fitted shapes had 6.8% lower normalized half-width than constant bands on average for 20–500 ms responses at 86 sites in eight patients
Its RMS limits motivated an explicit coverage endpoint. This study instead tested every voltage sample in a response. Both studies share the same patients.
- usesFive fit pulses and three calibration pulses gave 75.6% mean patient and 70.3% pooled whole response coverage at 86 selected SEEG sites
The earlier RMS rank study supplied related same-cohort evidence and the finite band budget comparison. This study applies the established rank rule to every time sample of one fixed readout, with five fit pulses and three separate calibration pulses.