About

What This Tool Does

This web application is a research prototype for binary chest imaging classification. It returns a model prediction (`Fibrosis` / `Normal`), class probabilities, and a Grad-CAM overlay for supportive review.

Current Repository Scope

• Inputs: Single uploaded DICOM image or chest X-ray image
• Model family: EfficientNet-B3 (via timm)
• Output: Binary label, class probabilities, Grad-CAM heatmap, PDF report
• Validation status: Not validated for clinical diagnosis or full CT study interpretation
• Intended use: Research, experimentation, and engineering review only

Datasets & References

• OSIC Pulmonary Fibrosis Progression (Kaggle)
• Pulmonary Fibrosis Dataset (Kaggle)
• ChestX-ray14 (NIH)
• Fibrosis-Net (DarwinAI) — Frontiers in AI, 2021
• FibroVit — PMC, 2023