Coronary Calcium Score Prediction

This project aimed to develop a Machine Learning model using Convolutional Neural Networks (CNNs) to classify chest radiography images and predict the Agatston coronary calcium score. The internally developed model achieved a 32% cost reduction and reached a ROC-AUC of 0.75 using the VGG-19 architecture, demonstrating its ability to distinguish patients with different levels of coronary calcification—an important indicator of cardiovascular risk.

Tools used:
Python, TensorFlow, Scikit-learn, VGG-19, Pandas, NumPy, and Seaborn.
Jupyter Notebooks.
Google Cloud Storage and Google Compute Engine.
GitLab.