Patient offsets from distant same-patient pairs reduced normative SEEG spectral error by 8–34% in 249 RAM patients

We tested whether a patient-specific offset estimated from that patient's own SEEG bipolar pairs at least 40 mm from a target improves a leave-one-patient-out normative spectral map. The analysis used 27,974 pairs from 256 patients in the RAM ds004789 and ds004809 datasets and five spectral features. A normative value was the mean of the 30 nearest pairs from other patients in MNI space, with coverage limited to a 15 mm 30th-neighbour distance. Among 249 patients with at least five target pairs, the offset reduced mean squared error by 16% for relative theta power (95% CI 11–22%), 8% for relative alpha (3–13%), 25% for relative beta (20–30%), 25% for the 2–40 Hz aperiodic exponent (19–31%), and 34% for log broadband power (28–40%); Holm-adjusted p values were 0.0002–0.0011. Only about half the patients improved for theta and alpha. An offset from another patient increased error by 48–76%. Offsets estimated from one or two distant pairs increased error; five pairs improved beta, exponent, and broadband power, but not alpha. This test covered one 60 s word-encoding segment, bipolar montage, and interictal data. It had no seizure-onset or pathology labels and did not test detection of abnormal tissue; observed features may include pathology, which may also bias calibration. The exact OpenNeuro releases used are not recorded in the archived provenance.

stereo-EEG · SEEG · normative maps · patient calibration · spectral features · intracranial electrophysiology · RAM dataset · leave-one-patient-out