shipr1611/xray-explainer

Clinical AI Explainability Dashboard

Upload a chest X-ray → get AI-powered triage + Grad-CAM explainability + clinician feedback loop.

Model

• —Architecture: DenseNet121 fine-tuned with class-imbalance weighting
• —Pre-training: SimCLR self-supervised contrastive learning
• —Dataset: COVID-19 Radiography Database (~21,000 images)
• —Classes: COVID-19 · Viral Pneumonia · Lung Opacity · Normal
• —Accuracy: 91.4% | F1: 0.89

Features

• —🩻 Grad-CAM heatmap overlay (which regions drove the prediction)
• —🚨 Triage levels: CRITICAL / URGENT / HIGH / NORMAL
• —📝 Plain-language clinical summary
• —💬 Clinician feedback loop (agree/disagree logging)
• —📊 Feedback agreement rate tracking

Disclaimer

For research and educational purposes only. Not a certified medical device.