Worst-case guarantees and target accuracy in respiratory motion-model selection — code, protocols and derived results

Code, frozen protocols, lock files and derived results for the manuscript "Worst-Case Guarantees and Target Accuracy in Respiratory Motion-Model Selection" (submitted to IEEE Journal of Biomedical and Health Informatics). The deposit contains: (1) code for the public POPI 4DCT benchmark, the controlled nonlinear-surface benchmark and its hash-locked replication, the calibration-consistency gate, and the TrackRAD2025 cine analysis kit; (2) protocols with their SHA-256 hashes and the lock files for each locked analysis, including the long-sequence files exactly as locked; (3) derived results behind every table and figure. No source data are redistributed. POPI and TrackRAD2025 data must be obtained from their original providers. No image data or landmark/tracking coordinates are included. This deposit postdates the analysis locks. It publishes the recorded hashes but does not by itself establish when they were created; see preregistration/PROJECT_LOG_EXCERPT.md. Research only — not for clinical use.