A contact metric ridge kept gamma scores between shafts unchanged across two full bipolar layouts on the same contacts in 32 SEEG patients

We tested whether a bipolar layout change alters scores between SEEG electrode shafts. We used public OpenNeuro ds004100 version 1.1.3. We selected 353 complete eight-contact blocks in 32 epilepsy patients. Each patient had two 60-second records. Adjacent and star layouts kept the same seven independent contact differences. In the 70–109 Hz band, the layout change replaced a patient mean of 40.3% of pairs selected by the largest pair correlation. The 95% patient bootstrap interval was 36.0–44.5%. The selected set was the top 20% within each patient and record. Identity ridge canonical correlation replaced 13.3%. A contact metric ridge kept all selections unchanged. Its maximum primary score difference was 3.7e-15. The ridge value was 0.1. A proof and separate numeric checks explain the invariance. We also checked two other bands and two other ridge values. Fitted variates from the first record gave the same second-record correlation after either layout fit. We repeated source retrieval for 72 records and checked the stored covariance. The package preserves compact inputs, code, figures, receipts, and reproduction instructions. The result supports a check of BCI input association scores. Task decoding, causal connections, and stimulation effects were not tested.

stereoelectroencephalography · canonical correlation analysis · ridge regularization · bipolar montage · signal processing · brain-computer interface