Dynamic Isolation Forest for Anomaly Detection in Post-PCI Myocardial Infarction Patients: Code and De-identified Development Cohort Data

This record contains the de-identified local development cohort data, analysis scripts, result tables, figures, and supporting documentation for the study “Dynamic Isolation Forest for Anomaly Detection in Post-PCI Myocardial Infarction Patients”. The dataset includes de-identified inpatient laboratory data from patients with myocardial infarction who underwent percutaneous coronary intervention (PCI). The accompanying scripts reproduce the fixed three-time-point window construction, feature extraction, model comparison, KNN-imputation sensitivity analysis, patient-level bootstrap stability analysis, and figure generation used in the local development cohort analysis. This release package includes:1. de-identified raw and processed development cohort data;2. fixed-window manifests and expert-review related data;3. Python scripts for preprocessing, model evaluation, sensitivity analyses, and bootstrap analyses;4. main and supplementary result tables;5. figures generated from the analysis scripts;6. documentation files, including README and data dictionary files. All direct patient identifiers have been removed before release. Dates and admission-related identifiers have been de-identified or transformed for privacy protection. The external MIMIC-IV supportive prognostic analysis is not included in this package.