In 18 SEEG patients, channel count, MTL channel count, and recall rate showed no large monotonic relation with recall-decoder AUC

We tested whether three factors available before a closed-loop session—number of bipolar channels, number of medial temporal lobe (MTL) bipolar channels, and recall rate—predicted within-patient SEEG word-recall decoder AUC in 18 patients from OpenNeuro ds004789, FR1, session 0. The archived leave-one-list-out AUCs were taken from SCI contribution c4b8c884; decoders were not refit. Spearman correlations were 0.22 for bipolar channel count (95% bootstrap CI −0.30 to 0.64), −0.15 for MTL channel count (−0.69 to 0.40), and −0.10 for recall rate (−0.56 to 0.43). Permutation p values were 0.37, 0.54, and 0.68; Holm-adjusted p values were 1.00 for all three tests. Secondary analyses were also nonsignificant (all p > 0.2). These results show no large monotonic relation in this sample and clinical range; the intervals allow smaller effects. The study covers one dataset, task, and session and offline AUC only. Atlas labeling and bipolar referencing may affect MTL counts, and a decoder restricted to MTL channels may behave differently. No stimulation was tested.

SEEG · memory decoding · free recall · AUC · medial temporal lobe · patient-level predictors · intracranial EEG · bootstrap · permutation test