Version 1.0.0 of ds005415 has an EDF length mismatch and an event sample rate mismatch
We checked the 13 public SEEG records in OpenNeuro ds005415, version 1.0.0. One source EDF file is shorter than its header requires. Its full hash matches the source repository. One event sample column uses a 1,000 Hz clock, while its EDF signal uses 2,000 Hz. Raw signal probes also support a possible event time origin problem in four records. This later finding is exploratory. The package provides audit code, source metadata, raw signal evidence, hashes, retrieval instructions, and diagnostic transfer results. The decoder results cannot support a neural conclusion until signal alignment is established.
stereoelectroencephalography · SEEG · neural decoding · number cognition · event timing · data audit · OpenNeuro · reproducibility
References
2- usesdoi:10.18112/openneuro.ds005415.v1.0.0
The analyses checked source metadata and pinned signal byte ranges from OpenNeuro Numbers, version 1.0.0. The recorded release and content hashes identify the files used.
- doi:10.1371/journal.pone.0313155
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesRecord time stamps differ by up to 39.5 ms from the continuous header clock in 3 of 13 SEEG files in ds005415 version 1.0.0
The earlier audit supplied initial object versions, expected annex hashes, initial record time stamps, and the known short-file finding. We checked all annex keys at the pinned commit and verified every complete source object hash.
- usesJoint label changes preserved shared task labels and class counts in 944 SEEG trials from 12 patients
We used 30 archived raw timing probes and their source records to check the reported event-to-signal origin concern. A separate parser reproduced all 120 auxiliary pair measurements. This evidence limited the neural interpretation of our decoder scores. We added a joint label method for the shared task design.
- usesConstant event time shifts aligned most light sensor steps in four ds005415 v1.0.0 SEEG recordings with nominal sample timing
The earlier audit supplied source tables, row matches, headers, object versions, four channel pairs, and reference sensor delays. Its 12 calibration events were excluded from the primary test.