All 53 lateralized SEEG-evoked percepts were contralateral; elevation and eccentricity tests were inconclusive

We used the public SEEG stimulation data set ds005169 v1.0.0 (112 stimulation events that caused visual percepts, 20 patients). We tested if the visual-field position of each percept agrees with retinotopic rules. The data set gives the clinical percept labels and the contact coordinates. We wrote the tests before the analysis. H1: 53 events had a percept in one hemifield only. All 53 were contralateral to the stimulated hemisphere (19 left-hemisphere sites, 34 right-hemisphere sites), in 16 of 16 patients. The exact patient-level sign-flip test gave p = 3.1e-5 (Holm p = 9.2e-5). H2: upper-field percepts came from lower sites than lower-field percepts in the pooled data (AUC 0.78, p = 0.021). But the coordinates are in native patient space, so the primary test was within patient. Only 2 patients had both classes (AUC 0.85, permutation p = 0.12, Holm p = 0.24). H3: no patient had both central and peripheral percepts, so we could not do the within-patient test. The pooled AUC was 0.65 (p = 0.32). Limits: all events caused a percept, so there are no negative trials. The labels are clinical and coarse. Elevation and eccentricity have low power. For a visual prosthesis, the hemisphere of a contact predicts the side of a lateralized percept in these data. But this data set cannot predict the position in the hemifield. The package contains the code, an S3 manifest with ETags, the event table, statistics and figures.

stereo-EEG · visual cortex · cortical stimulation · visual percepts · retinotopy · visual prosthesis