Imagery calibration gave a fixed EEG rule higher mean accuracy on later odd cues in 94 people in ds004362 v1.0.0

We tested whether real hand movement could replace imagined hand movement for personal EEG calibration. A fixed four-feature rule predicted left or right cues in later imagery runs from OpenNeuro ds004362 v1.0.0. The primary score used original odd trial numbers because even cues follow a known label rule. In 94 eligible people, mean balanced accuracy was 56.39% with imagery calibration and 52.23% with movement calibration. The paired difference was 4.16 percentage points. Its 95% person bootstrap interval was 0.63 to 7.67 points. Several planned checks had intervals that included zero. In a matched control with 90 people, movement calibration predicted even cues from before-cue EEG at 54.86%. This finding supports separate cue-parity scores and a before-cue control. The evidence concerns cue prediction with this fixed rule and selection rule. It does not identify a neural cause or establish clinical use. All 436 source EDF files passed hash checks. Independent calculations checked the saved features, models, and predictions. Two offline runs gave identical numeric files. The package preserves lossless local signal windows, source identities, code, all planned results, and repeat instructions.

electroencephalography · motor imagery · brain-computer interface · calibration · cue prediction · balanced accuracy · EEG