Within-session SEEG decoding showed spectral shape carried working-memory load, but non-inferiority to full band power was not established
This study tested whether spectral-shape features of stereoelectroencephalography (SEEG) band power retain verbal working-memory-load information after removing each channel's broadband level. Archived features from 15 patients in OpenNeuro ds004752 (session 1) were evaluated for set size 8 versus 4 with Ledoit-Wolf shrinkage linear discriminant analysis and repeated stratified five-fold cross-validation. Spectral shape decoded load above chance (mean AUC 0.688; p = 0.0002; 14/15 patients above 0.5), compared with 0.705 for full band power and 0.650 for broadband level alone. The mean shape-minus-full difference was -0.018 (95% CI -0.071 to 0.030). Non-inferiority at a 0.05 AUC margin was not established (p = 0.126); the data do not exclude a loss greater than 0.05. Shape-minus-level was not significant. In an algebraic gain-offset illustration, full-feature AUC fell as simulated channel offsets increased, while shape AUC was unchanged by construction. This was a within-session analysis and did not test real cross-day drift. The small sample and trial counts produced wide uncertainty.
stereoelectroencephalography · working memory · memory load · band power · spectral shape · gain invariance · linear discriminant analysis · brain-computer interface
References
4- usesMemory load is decodable across SEEG regions, with no detected medial temporal advantage in 15 patients sci:sha256:54d6bf7a4177c2592af970407f757ef23a8ddeb3790b39ea042ad858c1d09fbc
Input band-power features (key bp) for 15 patients of ds004752, copied with SHA-256.
- supportsIn two SEEG patients, band-power features shifted more between days than within a day, and working-memory-load decoding transferred less well across days sci:sha256:4d02e639292d3d2ecc9da767e072ccd99c8fe28943c38f3b2ce84c4e1a392744
That study found large cross-day shifts of band power. This study shows that gain-invariant spectral-shape features keep most of the within-session load information, which motivates a cross-day test of these features.
- usesIn session 1 of 15 patients, IEDs did not decode memory load; masking raised SEEG AUC from 0.70 to 0.72 sci:sha256:632c406c34e1530eea3bb23516da5054eb6c64dbaa63783d6d4628a34fe21bbb
Original computation of the band-power features from the raw EDF files.
- doi:10.7554/eLife.78677
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesArchived SEEG pipelines agree on patient decodability rankings, but shared trial noise explains about 38% of patient variance
Per-patient load 8 vs 4 AUC table (column AUC_FULL) is one pipeline in the cross-pipeline matrix.
- supportsIn ds004789 FR1 data, spectral-shape features retained word-recall decoding accuracy under simulated per-channel gain drift
Shape features without the broadband level keep most decoding information; here shown non-inferior to full log power at a 0.02 AUC margin for FR1 recall decoding (ds004789).
- challengesSpectral SHAPE features do not improve cross-day SEEG working-memory load decoding
This package does the real cross-day test that was stated as a limit there. The within-session equivalence of SHAPE and FULL replicated, but SHAPE gave no reliable cross-day benefit over FULL.