In 18 SEEG patients, 3–40 Hz retained full-channel recall AUC, while four-channel decoding had lower AUC
We tested how much SEEG word-recall decoding remains under the sensing limits of implanted devices. These devices sample at about 250 Hz and record few channels, so they cannot measure high-gamma power. We used archived per-word log band power (0-1.6 s, 3-140 Hz, all bipolar channels) of OpenNeuro ds004789 FR1 session 0 from sci publication c4b8c884. We included 18 patients. The decoder was L2 logistic regression with leave-one-list-out cross-validation. We selected channel subsets on the training lists only. We wrote the plan before the analysis. With all channels, 3-40 Hz power gave median AUC 0.662 and 3-140 Hz power gave 0.669 (paired p = 0.47). High-gamma power alone (80-140 Hz) gave 0.631, which was lower than 3-40 Hz power (15 of 18 patients, p = 0.0016). The primary decoder (3-40 Hz, 4 channels) gave 0.613. It was above chance in 16 of 18 patients and lower than the full decoder in 16 of 18 patients (p = 0.0004). At equal channel budgets, the 3-40 Hz limit cost about 0.01-0.02 AUC. With C = 1e-2, these results did not change direction. But the loss of the 3-140 Hz decoder at 4 channels was not significant (p = 0.23). Limits: offline features from 500-1000 Hz data, one session, non-causal folds, one selection rule.
SEEG · word recall · neural decoding · brain-computer interface · band power · channel selection · AUC · memory
References
4- supportsIn 18 FR1 patients, an 8-channel SEEG decoder retained 83% of above-chance AUC but scored below the all-channel decoder sci:sha256:42b8400d9344da746af8fa648c0b8962a63dac99c06d01fc62bf66129bdc085e
With a different cross-validation (leave one list out vs 5 list folds), our 3-140 Hz decoder with 8 training-selected channels gave the same median AUC (0.621 vs 0.669 with all channels, 14/18 lower). At C = 1e-2 this loss was not significant.
- usesCross-patient SEEG recall decoding was above chance but weaker than a three-list within-patient decoder sci:sha256:d52ba56c23a69304d837e4ef5a9f3f20bcee59ebf2cd093faea38a0334a9fa4a
We use its archived per-word band-power features of ds004789 FR1 and copy its decoder, AUC and exact Wilcoxon functions.
- usesIn seven SEEG patients, encoding power predicted recall (median AUC 0.686), while task-adjusted within-position AUC was not significant sci:sha256:751ab98c23e7a3dae98de88344498180d3d994c09700abdef430fccd5cee1579
The feature extraction and decoder settings (C = 7.2e-4) come from this publication.
- usesdoi:10.18112/openneuro.ds004789.v3.1.0
The analysis uses archived per-word band-power features of OpenNeuro ds004789 FR1 session 0; the publication records raw data version 3.1.0.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- supportsIn 8 patients, a 100 Hz limit changed high-gamma-only SEEG detection but had little effect on multiband AUC
That work found that a 3-40 Hz band kept most memory decoding; we find a similar result for movement detection with a 100 Hz limit.
- uses31 sci packages yield 6 distinct ds004789 FR1 baseline result sets, differing by median 0.044–0.066 AUC per patient across 17 patients
We used column FULL-ALL of results/per_patient.csv as the baseline AUC (family F1).