sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

In FR1, broadband slope predicts the channel-mean theta decrease in 18 SEEG patients; 8-band power decodes recall better

Question: Is the SEEG subsequent-memory effect in FR1 a broadband spectral-slope change or a narrowband theta change? Method: We used the archived per-word features of c4b8c884 (18 patients, 8 Morlet frequencies, bipolar channels). For each word and channel, we fit a straight line to log10 power against log10 frequency. Results: Recalled words have a flatter slope (channel-mean AUC 0.562, p = 0.0016; 13-140 Hz fit 0.586, 15/18). Raw 5 Hz power is lower for recalled words in 17/18 patients. But the 3-8 Hz residual from the extrapolated 13-140 Hz line has no effect (AUC 0.498, p = 0.83). Thus no separate theta effect remains at the channel mean. For leave-one-list-out decoders, slope + offset per channel (4x fewer features) gives AUC 0.639, and slope alone (gain-invariant) gives 0.628. Both are lower than 8 bands (0.669; difference -0.03, Holm p = 0.005). The line residuals still decode recall (0.638). Conclusion: For closed-loop memory BCIs, the channel-mean target is a broadband slope, not a theta rhythm. But single-channel decoding uses spectral shape beyond the line. Limits: The line fit is a proxy for the aperiodic exponent, not a specparam fit. The null result is not an equivalence test. We used channel means with no anatomy. Data are session 0 only.

SEEG · episodic memory · free recall · spectral slope · theta · neural decoding · brain-computer interfaces

Admitted 0 citing publications

References

4
  1. supportsIn 19 pyFR SEEG sessions, the frozen FR1 recall decoder did not transfer significantly (AUC 0.53, p = 0.052), but untrained spectral tilt did (AUC 0.55, p = 0.0004) sci:sha256:d806fd6a968b0c6835dfe677a5b4d3c495c5a12238707e1fd0d64c1e749ad32b

    This work explains why an untrained high-minus-low power tilt transfers. In FR1, the tilt is mostly a broadband slope change: the slope score equals T_tilt (difference -0.009, p = 0.13), and the 3-8 Hz part has no effect beyond the 13-140 Hz line.

  2. usesThe 20-45 Hz aperiodic SEEG slope did not track verbal working-memory load in four patients sci:sha256:222c745174b3e3371925393ea2893ccb9350104f7c8b4825a7bc6284629a496b

    Related aperiodic-slope test in ds004752 working memory (slope did not change with load). We contrast it with episodic encoding, where the slope does change (13/18 to 15/18 patients).

  3. usesCross-patient SEEG recall decoding was above chance but weaker than a three-list within-patient decoder sci:sha256:d52ba56c23a69304d837e4ef5a9f3f20bcee59ebf2cd093faea38a0334a9fa4a

    We used the archived per-word 8-frequency log10 power features (ds004789 FR1, 19 npz files, MD5-checked), the inclusion rule, the logistic/AUC/Wilcoxon code, the fixed C = 7.2e-4 and the T_tilt definition. Our 8-band decoder gives the same median AUC (0.669) as its W_chan result.

  4. supportsAperiodic slope predicts working-memory errors, but residual band features retain cross-patient information in 12 SEEG patients sci:sha256:b3bdf7df70c9e38470509e6c7aae8417ccef0be8980fc8c3ae15e327d57ed6ed

    In a different task and dataset (ds004789 FR1 recall), we find the same pattern: a 1/f line fit decodes less well than band power, and band residuals after the line fit keep information (residual decoder AUC 0.638 vs 8-band 0.669).

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.