CardioSense – AI-Powered Heart Disease Risk Prediction

• 

  Dashboard

• 

  Metrics detail

• 

  Batch Files Analysis

About CardioSense: AI-Powered Cardiovascular Risk Prediction Inspiration Every 34 seconds, someone dies from heart disease often because risk is detected too late. As a student passionate about healthcare AI, we wanted to build a tool that not only predicts cardiovascular risk but also explains its predictions. The UCI Heart Disease dataset (303 patients, 14 features) provided a perfect starting point to demonstrate how machine learning can assist clinicians.

What we Learned Model comparison matters – different algorithms (Random Forest, Gradient Boosting, Logistic Regression) have different strengths; showing all three builds trust.

Interpretability – feature importance (e.g., thalach and cp as top predictors) is as valuable as accuracy.

Full‑stack integration – connecting a Flask API with a React frontend, containerising with Docker, and deploying on Hugging Face Spaces taught me practical MLOps.

How we Built It Backend: Flask REST API serving three scikit‑learn models trained on the Cleveland dataset.

Frontend: React + Tailwind CSS + Recharts for a responsive, medical‑friendly dashboard.

Deployment: Dockerised and hosted on Hugging Face Spaces (free tier).

Batch mode: CSV upload for bulk predictions, returning a downloadable results table.

Challenges Faced Class imbalance – addressed with class_weight='balanced' and stratified cross‑validation.

Model consistency – ensuring all three models receive the same scaled features via a shared StandardScaler.

Real‑time performance – optimised inference to <50 ms per request by pre‑loading models into memory.

Frontend‑backend communication – handling CORS and asynchronous file uploads with Axios.

Disclaimer: CardioSense is for research and education only—not a certified medical device. Always consult a healthcare professional.

Built With

• docker
• flask-3.0
• hf
• pytorch-2.2
• react-19
• resnet50