Visual-percept current varied across sites within patients and was not predictable from nearby SEEG sites

We assessed whether a visual-percept current at a new stereo-electroencephalography (SEEG) site could be predicted from currents at other sites in the same patient. The analysis used 112 stimulated contact pairs from 20 patients in OpenNeuro ds005169; 18 patients contributed 110 sites to leave-one-site-out predictions. The outcome was log2 of the clinical current that produced a visual effect, not a staircase threshold. Between-patient variance was 30% (ICC(1) = 0.30, permutation p = 0.0001), while 70% was between sites within patients. Patient-weighted mean absolute errors were 0.82 log2 for the mean of other patients, 0.67 for the mean of other sites in the patient, 0.60 for the nearest site, and 0.59 for inverse-distance-squared weighting. No paired comparison was significant after Holm correction across 18 patients. Across 453 within-patient site pairs, current differences did not increase with distance (Spearman = 0.01, p = 0.34). These results indicate substantial site-to-site variation in this sample and do not show that nearby sites or patient-specific averages reliably improve prediction. The data include only sites with visual effects; the clinical ramp protocol could vary between patients, and the sample had low power to detect small gains.

stereo-electroencephalography · direct cortical stimulation · visual percept · current calibration · brain-computer interface · within-patient variation · OpenNeuro