CardioAGI-X

CardioAGI-X: AGI-Inspired Biomedical Intelligence System for Cardiovascular Risk Reasoning

This project upgrades the original heart disease notebook into a modular biomedical AI research platform with advanced ML, calibrated uncertainty, simulated longitudinal modeling, explainability, multimodal scaffolds, multi-agent reasoning, safety reports, and deployment interfaces.

Disclaimer

This project is for research and education only. It is not a medical diagnostic device, is not true AGI, and has no clinical certification.

Main Artifacts

• heartd.ipynb: full research notebook.
• cardioagi_x/: modular Python package.
• app/streamlit_app.py: physician/patient research dashboard.
• api/main.py: FastAPI backend.
• scripts/train_platform.py: reproducible training and report generation.
• models/: trained model artifacts.
• outputs/: prediction, calibration, uncertainty, fairness, explainability, and survival outputs.
• reports/: publication-style and governance reports.
• diagrams/: Mermaid architecture diagrams.

Run Training

C:\ProgramData\anaconda3\python.exe scripts\train_platform.py

Run Streamlit

streamlit run app\streamlit_app.py

Run API

uvicorn api.main:app --reload

Optional Advanced Dependencies

The platform runs with sklearn fallbacks when advanced libraries are missing. To enable native XGBoost, LightGBM, CatBoost, SHAP, lifelines, MLflow, and transformer experiments, install:

python -m pip install -r requirements-advanced.txt

Scientific Constraints

• No true AGI claim.
• No clinical certification claim.
• No autonomous diagnosis or treatment recommendation.
• random_state=42 is used for reproducibility.
• Preprocessing is fitted only inside model pipelines to reduce leakage.
• Explainability outputs are statistical explanations, not causal medical proof.