One right lingual-gyrus SEEG pair decoded seven flicker frequencies at 98% from 1-second windows in one patient

In one patient with clinical photic stimulation, we assessed whether short SEEG windows could identify flicker frequency without a trained classifier. We analyzed 154 adjacent bipolar pairs from OpenNeuro ds006392 v1.0.1 and selected pairs using response scores from the other eye condition. The strongest pair, ROC1–ROC2 in the right lingual gyrus, had peak response z scores of 44.7 at 8 Hz with eyes open and 66.4 at 3 Hz and 59.9 at 10 Hz with eyes closed. For seven classes (3–30 Hz), this pair classified 48 of 49 one-second windows correctly in each eye condition (98%; chance 14%). With 0.5-second windows, accuracy was 91% with eyes open and 81% with eyes closed. All 154 pairs, random four-pair sets, and nonvisual gray-matter pairs performed worse. The 30-Hz fundamental response was not clear, although the decoder still classified that class. These results show cross-condition frequency read-out from a focal SEEG pair in this recording; they do not establish performance in a multi-target BCI. There was one patient and one train per frequency and condition, windows within a train were not independent, and binomial p values are therefore optimistic. Feature z-scoring used all test windows in each block, and photic artifact cannot be fully excluded. ITR values are nominal.

SEEG · visual cortex · flicker frequency · brain-computer interface · frequency decoding · steady-state visual evoked potential · cross-condition selection