COVID-19 X-ray Classification

Deep learning for medical imaging with 92% ROC-AUC and Grad-CAM explainability for clinical decision support

Technology Stack

PyTorchResNet50EfficientNetDenseNetGrad-CAMCLAHEStreamlitAlbumentations
ML Frameworks
Deep Learning
NLP
Deployment

ROI

$1.05M per year

Savings

$700K per radiology department

Accuracy

68.7% overall, 92.1% ROC-AUC

Key Performance Indicators

COVID-19 Sensitivity
67.0%
Time Savings
90% (5 min → 30 sec)
Throughput Increase
+60% (250 → 400 studies/day)

Business Impact Summary

Return on Investment$1.05M per year
Cost Savings$700K per radiology department
Model Accuracy68.7% overall, 92.1% ROC-AUC
COVID-19 Sensitivity67.0%
Time Savings90% (5 min → 30 sec)
Throughput Increase+60% (250 → 400 studies/day)

Overview

State-of-the-art deep learning system classifying chest X-rays into COVID-19, Viral Pneumonia, Lung Opacity, and Normal with 92.1% ROC-AUC. Grad-CAM visualizations provide clinical explainability. Designed for emergency department triage and radiology workflow optimization.

Live Demo

Live Demo

Interactive Streamlit application

📊

COVID-19 X-ray Classification Demo

Click below to load the interactive demo

This is a live, interactive demo deployed on Streamlit Cloud. You can:

  • Upload your own data for predictions
  • Explore model performance metrics
  • View interactive visualizations
  • Understand model predictions with explainability