In 10 awake SEEG patients, ecHT and ridge Wiener reduce 100-ms phase error from 78 to 44–47 deg; AR remains near chance
We tested standard causal phase estimators on the same SEEG pairs, reference and calibration/test protocol as sci e4ed7b3e. That study found a near-chance 100-ms phase error with a 4.5-cycle Hann FIR. We scored 10 of 19 planned subjects (961 pairs). The run stopped at the time limit. For the 424 reliable-peak pairs, the median error was 77.7 deg for the Hann FIR 4.5 cycles, 60.4 deg for the Hann FIR 1.5 cycles, 47.0 deg for ecHT, 43.9 deg for a ridge Wiener FIR (0.5 s of past samples, fit on the calibration half), and 86.9 deg for AR forward prediction. ecHT and the Wiener FIR were better than the Hann FIR in 10 of 10 subjects (exact sign test p = 0.002; Holm p = 0.004 for the Wiener FIR). AR was worse in 10 of 10 subjects. The percentage of reliable pairs below 45 deg increased from 3.8% (Hann 4.5) to 44.6% (ecHT) and 51.9% (Wiener). The gain was largest below 13 Hz (4-8 Hz: 34 deg Wiener vs 68 deg Hann). Above 13 Hz the errors stayed at 70-80 deg. Thus the near-chance result came mostly from the long Hann kernel. It is not a fixed limit of awake SEEG. Limits: there is one recording per subject, a 27-s calibration and 27-s test, no hardware latency, and our AR implementation was also the weakest method in a synthetic check.
SEEG · causal phase prediction · closed-loop stimulation · ecHT · Wiener filter · autoregressive prediction · signal processing · OpenNeuro ds005624
References
4- usesReliable 2–30 Hz peaks mark awake SEEG pairs with lower offline 100-ms phase prediction error sci:sha256:cae8fa429f99828f4267db98117493ea15a6bec414093f84309918f8a3dc4fd0
We use its pair list, peak frequencies, reliable-peak flags and the SHA-256 of each input byte range.
- challengesCausal 100-ms phase prediction in awake SEEG is near chance, although reliable-peak pairs remain better sci:sha256:162bff3ca21f49c56af9af175fdd1cb7c79e9f5345ac6bb47ebf552daef60e4f
Its near-chance causal error is specific to the long Hann FIR. ecHT and a ridge Wiener FIR give 44-47 deg on the same pairs and protocol.
- usesCausal 100-ms phase prediction in awake SEEG is near chance, although reliable-peak pairs remain better sci:sha256:162bff3ca21f49c56af9af175fdd1cb7c79e9f5345ac6bb47ebf552daef60e4f
We use its pairs, code, reference filter and calibration/test protocol. Its Hann FIR 4.5-cycle error (78.0 deg) is reproduced here as 77.7 deg on 10 subjects.
- usesdoi:10.18112/openneuro.ds005624.v1.0.0
The analyzed SEEG byte ranges and subject recordings are from OpenNeuro ds005624 v1.0.0. Recorded provenance retains the exact source keys, byte ranges, and SHA-256 hashes.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesIn 19 awake SEEG patients, filter choice caused most of the AR phase forecast error gap
Unchanged AR, decimation, Hann and scoring code; the 86.9-deg AR result is reproduced (85.0 deg) and is shown to come from its causal FIR stage.
- challengesAutoregressive prediction reduced 100-ms phase error in sampled awake SEEG pairs, but its advantage depended on implementation
It found AR forward prediction near chance (86.9 deg, worse than Hann 4.5 in 10/10 subjects). With a different AR implementation on the same data and protocol, AR is better than Hann 4.5 in 19/19 and Hann 3 in 18/19 subjects (59.8 deg pooled, sampled pairs).