The SEEG aperiodic exponent does not change significantly from fixation to encoding or early delay in 9 ds006136 patients (MFG/MTG bipolar channels, 800 ms windows, 30-48 Hz fit)

Question: does the aperiodic (1/f) exponent of SEEG change from fixation to object encoding or to working-memory delay? We used OpenNeuro ds006136 v1.0.1 (load-3 object working memory, bipolar MFG and MTG channels). We analysed 9 of the 13 patients (74 channels). Slow S3 download speed, not the data, stopped the other 4 patients. We wrote the plan before we computed exponents. We fit a line to the log-log Welch spectrum of the trial mean in 800 ms windows. The primary range was 30-48 Hz. We used an exact sign-flip test over patients with Holm correction. Result: the exponent change was not significant. ENC minus FIX was -0.009 (95% CI -0.12 to +0.11, p = 0.91). DEL1 minus FIX was -0.064 (95% CI -0.17 to +0.04, Holm p = 0.62). The mean FIX exponent was 2.87. Secondary results (descriptive): a broad 20-190 Hz fit showed a small flattening trend (ENC -0.023, p = 0.09; DEL1 -0.030, p = 0.06). The 30-48 Hz power decreased slightly during encoding (-0.018 log10 units, 8 of 9 patients, p = 0.008, no correction). When we subtracted the ERP, the values did not change. Limits: 9 patients only; curated trials; 2.5 Hz resolution; no knee or peak model; region results depend on single outlier channels. The package contains numpy-only code, the plan, per-channel values, SHA-256 values of the EDF files and the reproduction steps.

SEEG · aperiodic exponent · working memory · object encoding · spectral analysis · sign-flip test · OpenNeuro · ds006136