Fast ripple duration lowered loss for supplied overlap labels on later dates in seven SEEG patients

We tested whether fast ripple duration adds information about supplied ripple overlap labels on later dates. We used public sleep SEEG marks from OpenNeuro ds003498 version 1.0.1. The seven selected patients had depth electrodes in the source paper. The BIDS files call their channels ECOG; we preserve this conflict. We fitted on earlier calendar dates and tested 59,714 fast ripples on later dates. The comparison model used each event length's exact uniform timing overlap probability and electrode pair intercepts. Adding one duration term per patient lowered mean patient loss from 0.389944 to 0.208061 natural log units per event. The mean gain was 0.181883. All seven patients and all 15 test patient dates had positive gains. The 95 percent patient t interval was 0.134896 to 0.228869. The exact two-sided sign reversal value was 0.015625 under independent patients and symmetric null gains. Four fixed checks also had positive mean gains. Every one of 31,040 overlap-labeled fast ripples had the same start and duration as a ripple mark. Thus the result concerns supplied annotation information. It does not establish separate neural events, a neural cause, or clinical accuracy. This is a new analysis of a reused cohort. Independent timing, score, and optimizer checks passed. A fresh source retrieval and a separate process reproduced all 15 main numerical files byte for byte. The package preserves 655 source files, hashes, code, features, predictions, coefficients, results, figures, the local plan, and repeat instructions.

fast ripples · ripple overlap · stereoelectroencephalography · event duration · annotation boundaries · epilepsy