Bipolar SEEG high gamma did not track words in speech blocks from a film in four patients

We tested whether bipolar stereo-electroencephalography (SEEG) high-gamma power (70–150 Hz) followed word timing inside speech blocks of a naturalistic film in four SEEG-only patients from OpenNeuro ds003688. For each channel, we correlated high-gamma activity 150 ms after word labels and used within-block circular shifts with Benjamini–Hochberg false-discovery-rate correction. We also tested phrase-onset responses and a ridge decoder that classified speech versus pauses in held-out speech blocks. No patient met the fixed decision rule. Three patients had no channel passing false-discovery-rate correction; two channels in sub-23 passed that correction but neither passed the max-statistic correction. The held-out decoder AUCs were 0.498–0.543, and no patient passed the decoder null test (p=0.072–0.624). Phrase-onset responses were small. The result is limited to these four clinical implants, the tested montage and parameters, and the word-to-signal alignment, which lacked a positive control. Electrode anatomy was not assessed. The analyzed files came from the OpenNeuro ds003688 live S3 mirror; their recorded hashes are preserved, but the exact dataset release is not established.

stereo-EEG · high gamma · speech · word timing · brain-computer interface · naturalistic stimulus · decoding · negative result