SEEG pair-level memory-effect maps showed weak agreement across two FR1 sessions

Question: is the pair-level encoding subsequent memory effect (SME) map of an SEEG patient the same in two FR1 sessions? Stimulation studies select sites from such maps. Data: ds004789 FR1. Session 0 uses the archived features of d3aeb47e. Session 1 uses the same code with only the session changed, for all 26 d3aeb47e subjects that have a session-1 EDF. 25 patients and 2511 matched bipolar pairs passed inclusion. The pre-registered test (PLAN.md) used the per-patient Spearman rho of the Welch t maps (recalled vs not recalled), sign-flip tests, and Holm correction. Result: the maps agreed between sessions, but weakly. Mean rho was 0.13 for HFA 70-150 Hz (95% CI 0.06-0.22, Holm p = 0.002), 0.22 for theta 3-8 Hz (0.14-0.30, p = 0.0003), and 0.23 for HFA minus theta (0.14-0.31, p = 0.0003). In 5-7 of 25 patients, rho was 0 or less. The positive control (serial-position slope maps) passed (rho 0.22 and 0.29). Between-session rho was equal to or larger than the within-session odd/even-list split-half rho. Exploratory: with serial-position and list effects removed, rho stayed positive (0.10-0.19, p <= 0.001). Disattenuation ratios for theta and tilt were 0.91-0.95 after this control and 0.69-0.71 without it. Thus the low agreement comes mostly from noise in one session of about 300 words, not from a large change between sessions. Limits: 25 non-random patients, session dates not checked, pairs matched by name, one time window, and no anatomy. A stable non-neural source could also cause the HFA agreement. The package contains the code, features of both sessions, source hashes, results, and a run script.

stereo-EEG · subsequent memory effect · free recall · test-retest reliability · electrode pairs · theta · high-frequency activity