Analysis code and de-identified dataset for: Muscle-ultrasound texture reveals information beyond conventional measures in children receiving palliative care

Complete analysis code and the de-identified participant-level dataset underlying a cross-sectional radiomics study of muscle-ultrasound texture in 77 children, conducted at Hospital Infantil Universitario Niño Jesús and Hospital Universitario La Paz, Madrid, Spain. Running "pip install -r requirements.txt" followed by "python run_all.py" reproduces every number reported in the manuscript: the eight feature counts, the three Cliff's delta estimates (derivation 0.88; cross-validated 0.86; frozen cross-phenotype application 0.90), the robustness analyses, and both main figures. Runtime is under one minute on a laptop. The dataset is deliberately minimised. Free-text diagnoses, genetic findings, anthropometry and per-image identifiers are withheld, and exact age is replaced by four-year bands, to prevent re-identification in a small rare-disease paediatric cohort. See the "Privacy note" sheet of the dataset file. Ultrasound images and segmentation masks are not shared.