Corrected auditory onsets did not improve SEEG responses or decoding in five OpenNeuro ds005415 subjects

We evaluated whether a corrected trial-onset rule improved stimulus-locked responses and decoding in run 1 of five subjects from OpenNeuro ds005415 v1.0.0. The rule used events.tsv ‘Cue’ for visual trials and ‘Cue Start’ for auditory trials; the comparison used ‘Cue’ for all trials. Across 149–171 bipolar contact pairs per subject, only 0–3 pairs per subject showed high-gamma responses with |t| > 4, for visual and auditory trials. The corrected rule did not improve held-out auditory response t (+0.14, p = 0.50), spoken-versus-beep AUC (difference −0.02, p = 0.88), or auditory-versus-visual AUC (difference −0.05, p = 0.56). Under the corrected rule, format AUCs ranged from 0.46 to 0.57, and number decoding was near chance within formats (mean Spearman ρ 0.00–0.20) and across formats (−0.16 to 0.14). The visual positive control also failed. No group test had p < 0.05; with five subjects, the minimum possible exact sign-flip p was 0.0625. These results show that the tested onsets and features provided no usable decoder signal; they do not show that number information is absent. Only five subjects, one run, one feature set, and one linear decoder were examined, so small effects are not excluded. Poor coverage of responsive cortex or an onset-to-EDF clock offset may explain the weak responses, and these data cannot distinguish those causes.

stereo-electroencephalography · SEEG · high-gamma · brain-computer interface · trial onset · decoding · OpenNeuro · positive control