Four learned patterns had 2.34 percent higher mean relative error than four cosine patterns in five SEEG patients with separate shafts

We tested whether learned time patterns improve prediction of a five pulse SEEG mean. SEEG means stereoelectroencephalography. We used OpenNeuro ds004457 version 1.0.2 and five patients. We learned patterns from other stimulation shafts in each patient. The fixed main model used four patterns. The comparator used four fixed cosine patterns. The test used 1081 selected links, 50 to 250 ms, and all 252 complementary five pulse groups. The mean patient error reduction was -2.34 percent. The descriptive patient resample limits were -10.50 to 3.87 percent. The conditional two sided sign flip p value was 0.6875. Two patients improved. The two pattern model had a mean 7.70 percent reduction. The eight pattern model had a mean 3.29 percent increase. An exact finite pool calculation separated observed mean loss from pulse split variation. The package contains source locks, pulse data, learned patterns, all results, and independent checks. A fresh raw rebuild matched every saved waveform value. These results apply to the selected pulse pool in known patients.

stereoelectroencephalography · electrical brain stimulation · evoked response · principal component analysis · pulse prediction · finite pool loss · OpenNeuro