CardioVision AI

Inspiration

Cardiovascular diseases are a major health concern, and understanding medical reports can be difficult for non-specialists. We were inspired to combine AI with interactive 3D visualization to make cardiovascular risk information easier to understand and explore.

What it does

CardioVision AI aims to predict coronary artery disease (CAD) risk using patient clinical and physiological data. It visualizes predicted stenosis probabilities for the LAD, LCX, and RCA arteries on an interactive 3D heart model. The dashboard is designed to display risk estimates and explain the factors influencing predictions.

How we built it

We designed the project around a machine learning pipeline using Python, pandas, and scikit-learn to process clinical data and train prediction models. We plan to use FastAPI for the backend and React with Three.js for interactive 3D visualization. Explainable AI techniques such as SHAP will help communicate the factors contributing to model predictions.

Challenges we ran into

Some key challenges include working with a relatively small medical dataset, avoiding data leakage, evaluating model performance reliably, and representing vessel-specific predictions accurately on a 3D heart. We also need to ensure that visualizations do not imply actual plaque locations when the model only predicts risk probabilities.

Accomplishments that we're proud of

We are proud of developing a project concept that brings together machine learning, explainable AI, medical data, and interactive 3D visualization. Our goal is to make complex cardiovascular risk information more understandable through a clear, user-friendly interface.

What we learned

We learned about the importance of data preprocessing, feature selection, model evaluation, and explainability when working with medical datasets. We also explored how frontend visualization and backend machine learning can work together to create a more interactive and informative experience.

What's next for CardioVision AI

Our next steps are to complete the prediction pipeline, evaluate multiple models, integrate the trained models with the web application, and implement interactive 3D artery visualization. We also aim to improve the dashboard, add clear explanations of predictions, and test the prototype thoroughly. The system will remain an educational and decision-support prototype, not a replacement for professional medical diagnosis.