v1.0.0-demo – Time-Series ICU Patient Deterioration Predictor

v1.0.0-demo – Initial Public Demo Release

This release marks the first complete public demo version of the Time-Series ICU Patient Deterioration Predictor.

Included

• Dual-architecture deterioration-risk modelling system using LightGBM and Temporal CNNs
• NEWS2-derived clinical risk target generation
• ICU time-series preprocessing and feature engineering pipeline
• Timestamp-level temporal features and patient-level aggregated features
• Model training and evaluation workflows
• SHAP and saliency-based interpretability analyses
• Batch inference and per-patient lookup interface
• Full project documentation and Zenodo archive

Notes

This release uses the MIMIC-IV Clinical Demo dataset and is intended for methods demonstration, reproducibility, and educational use. Reported metrics are illustrative and should not be interpreted as clinically validated estimates of real-world performance.