Patient calibration covered all 17 test SEEG slope maps at 30 to 80 Hz with nominal 90 percent bands 7.56 times wider than pointwise bands

We tested nominal 90 percent prediction bands for all selected slopes in a patient map. Public SEEG data from OpenNeuro ds004100 version 1.1.3 supplied 36 patients with two files each. A fixed split used 19 patients for calibration and 17 for testing. At 30 to 80 Hz, calibration of the largest standardized error covered every target in all 17 test maps. Pointwise Student t bands covered no complete map. Bonferroni bands covered two maps. The mean full width was 5.612 slope units, 7.56 times the pointwise width. The reverse file direction covered 12 of 17 maps. The theoretical guarantee assumes exchangeable patient objects and fixed scoring. This audit reused earlier patients and does not establish population or clinical coverage. The package preserves code, compact features, raw retrieval records, all checks, and reproduction instructions.

SEEG · spectral slopes · conformal prediction · prediction bands · map coverage · patient calibration · epilepsy · coverage uncertainty