A 20 Hz harmonic mask did not increase speech period accuracy on 42 later trials in one bipolar SEEG record with reported VNS artifact
We tested a 20 Hz harmonic mask for reading speech periods from public SEEG data. We used OpenNeuro ds004993 version 1.1.2, with adjacent contact subtraction and an early-to-later trial split. In W3, the source reports vagus nerve stimulation artifact. The ordinary reader, harmonic mask, and equal-bin-count control each classified all 84 windows from 42 later sentence trials correctly. The harmonic gain was zero percentage points. This is a test ceiling, not proof of equal future performance. In W2, the ordinary reader classified 63 of 68 later windows correctly. The harmonic mask classified 60 and the control classified 59. Fixed-pair checks gave the same accuracy results. The package preserves unmasked window spectra, sample windows for audit, models, all predictions, three figures, a local analysis plan, code, and a pinned retrieval recipe. Full source hashes pass. Independent signal and model checks pass. A full voltage replay matches every derived array. The test uses known sentence times during passive listening in two short records. It does not test direct brain stimulation or performance across patients. Version 2 corrects the package checker to exclude metadata that sci adds after download.
speech period classification · stereoelectroencephalography · spectral masking · vagus nerve stimulation artifact · ridge regression · signal processing
References
2- supersedesA 20 Hz harmonic mask did not increase speech period accuracy on 42 later trials in one bipolar SEEG record with reported VNS artifact sci:sha256:d7fc17c277d0a0d0f448f35ac3f3b843c57214d79da828941f5283a10fc8e94e
This version corrects the extra package checker so that it excludes the metadata folder added by sci fetch. All scientific results, models, and signal arrays are the same.
- usesdoi:10.18112/openneuro.ds004993.v1.1.2
The analyzed W2 and W3 recordings and event metadata came from OpenNeuro ds004993 version 1.1.2.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesCalibrated thresholds removed two silent alarms and missed six sentences in two SEEG replays with a 0.5 s offset allowance
We reused the W2 and W3 source pins, adapted the retrieval code, and used its spectral and ridge reader recipe. Its known-time accuracy results motivated a continuous replay test. Our fit split and alarm analysis differ.
- usesFour time bins gave no gain over repeated mean log power on 40 later sentence trials in one SEEG record
We used fixed source metadata and retrieval versions from this publication. We reused four digital sample windows for an EDF reader audit. Its speech-period result motivated the sentence question. The new test identifies known sentence items in the same W2 record.
- A fixed SEEG reader gave no positive outputs from 258 inputs with no annotated sound and 94 at quiet decision times in one later listening test
[Legacy relation: reproduces] Reproduced the ordinary fixed reader, all selected-window PSD values, model coefficients, and the W2 63/68 and W3 84/84 later window results. Added continuous input labels, separate threshold calibration, and an explicit post hoc decision-time audit.
- usesFive score confirmation reduced observed onset recall in two public SEEG records after separate calibration
Its supplied-onset window classification result motivated the continuous command question. We used its pinned source metadata, S3 versions, and retrieval recipe. We adapted its CC0 retrieval code to bounded curl requests. We did not reuse its features, fitted readers, split, or predictions.