Heart Failure Prediction: Clinical AI Monitor

An enterprise-grade, real-time Cardiac Clinical Decision Support System (CDSS). This project integrates machine learning with a professional dashboard to assist medical professionals in identifying high-risk cardiovascular patients based on clinical vitals.

The system utilizes an XGBoost Classifier, deployed via a high-performance FastAPI backend, and visualized through an advanced, intuitive Streamlit clinical suite.

Core Features

• Clinical Risk Analysis: Calculates the probability of heart failure using multi-factor clinical inputs (Age, Blood Pressure, Cholesterol, etc.).
• Explainable AI (XAI) Dashboard: Features a Radar Chart (Vitals Profile Analysis) to visually represent patient health imbalances, allowing doctors to quickly identify which specific vitals are at risk.
• Real-time Inference: A robust API that processes patient vitals and returns risk probabilities in milliseconds.
• Professional Metrics: High-precision clinical monitoring with low False Positive rates.

---

Model Evaluation Metrics

The model was validated on a comprehensive dataset of cardiovascular vitals, achieving the following performance metrics:

Metric	Score
Accuracy	86.4%
Precision	91.8%
Recall	84.1%
F1-Score	87.8%

The model demonstrates high precision, minimizing false-alarm fatigue, which is critical for implementation in clinical environments.

---

Dashboard Interpretation

The dashboard is designed to reduce cognitive load for medical staff:

1. Patient Vitals Entry: Sidebar inputs for capturing precise clinical parameters.
2. Calculated Risk Gauge: A visual representation of cardiac failure probability (0-100%).
3. Vitals Profile Analysis (Radar Chart): Normalizes patient data to visualize the "shape" of the patient's health profile. An asymmetric polygon indicates an imbalance in critical health markers.

---

Architecture

1. Machine Learning: XGBoost Classifier trained on 11 critical cardiovascular features.
2. Backend API: FastAPI server managing data validation and inference.
3. Frontend Dashboard: Streamlit application featuring radar visualization and clinical risk assessment.

Directory Structure

Heart_Failure_Prediction/
├── app/
│   ├── main.py                 # FastAPI backend
│   └── dashboard.py            # Streamlit clinical UI
├── data/
│   ├── heart.csv               # Raw dataset
│   └── dashboard_sample.csv    # Processed sample data
├── models/
│   └── heart_failure_model.pkl # Serialized XGBoost model
├── notebooks/
│   └── 01_training_and_eda.ipynb # Pipeline
├── requirements.txt
└── README.md

Installation & Setup

1. Clone and Install:

git clone https://github.com/Arya-azimi/Heart-Failure-Prediction.git
cd Heart-Failure-Prediction
pip install -r requirements.txt

2. Start Services:

• Backend:

cd app
uvicorn main:app --reload

• Frontend:

cd app
streamlit run dashboard.py

License

This project is licensed under the MIT License.