A signed score increased flags more than a distance score for simulated SEEG mean loss at 20–500 ms in eight patients with five calibration pulses

We tested a response monitor for a brain-computer interface that records after brain stimulation. We used public SEEG data from eight patients, 86 stimulation sites, and 430 later pulses. Each monitor used five calibration pulses. We subtracted their mean waveform from each later response and kept its remaining signal. At 20 to 500 ms, the signed score increased the patient mean flag rate by 63.2 percentage points. The distance score increased it by 45.2 points. The paired difference was 18.0 points, with a descriptive patient bootstrap interval of 10.8 to 25.7 points. The difference was positive in seven patients and zero in one. The original-data flag rates were 14.9% and 16.8%, respectively, with equal patient weight. A pre-pulse control also showed a 10.7-point difference. Thus the test supports a method for detection of a change in a learned mean. It does not measure sensitivity to a real loss of brain response. Under exchangeability of the six complete pulse maps, each rank rule has a marginal flag probability of at most 1/6. The package preserves exact compact inputs, source hashes, source retrieval code, analysis code, results, and three figures.

stereoelectroencephalography · brain-computer interface · response monitoring · conformal rank · electrical stimulation · mean removal · signal processing