CardioScan: An AI-Powered Heart Disease Prediction System Using Machine Learning A locally developed web application that uses machine learning to predict heart disease with 88.5% accuracy, providing instant risk assessment through an accessible browser-based interface.

Abstract

Heart disease is the leading cause of death worldwide, causing an estimated 17.9 million deaths each year. In Pakistan, cardiovascular disease accounts for about 30% of adult deaths, yet access to specialized heart care is unequal — in many regions, there is less than one cardiologist for every 100,000 people.

CardioScan is an open-source web application built locally to address this problem. It uses Python and Flask to train a Random Forest classifier on the UCI Heart Disease dataset and gives an instant cardiac risk score through a user-friendly browser interface. The system achieved 88.5% accuracy on the test set, with a 5-fold cross-validation score that remained stable across folds.

The system requires no special hardware, works on any web server, and generates a formatted clinical PDF report for patients to take to any doctor. The model inference itself takes less than 1 ms since it loads the trained model into memory at startup.

"Digitising cardiac screening does not need expensive imaging equipment; it needs a trained model, a browser, and blood test results."