Maps fitted to one SEEG record retained 91.9 percent mean gamma energy in another record after common mean removal in 33 patients

We tested spatial compression maps in public SEEG data from 33 patients with epilepsy. Each shaft supplied eight consecutive contacts. Each patient supplied two 60-second interictal records. We removed the common contact mean. Principal component analysis fitted a map to retain at least 90 percent gamma energy in record A. The map used a mean of 4.27 of seven output signals. It retained 91.95 percent mean energy in record B. The descriptive 95 percent patient bootstrap interval was 91.33 to 92.55 percent. Fixed cosine maps retained 78.13 percent at the same output count. The paired mean gain was 13.82 percentage points. However, 60 of 354 shaft blocks missed the 90 percent target in B. The worst block retained 38.91 percent. Reverse-record and frequency checks gave similar mean results. The file order does not establish time order or a known time gap. The result concerns band energy after common mean removal. It does not establish task information, long-term transfer, hardware savings, or stimulation benefit. The package contains inputs, maps, results, three figures, source checks, code, and repeat instructions.

SEEG · stereoelectroencephalography · spatial compression · principal component analysis · gamma band · signal processing · energy retention · brain-computer interfaces