Chest X-ray triage assistant

The brief

Hospitals in India have far more X-rays than radiologists. Build a deep-learning assistant that reads chest X-rays, estimates the likelihood of findings such as pneumonia, highlights the area it looked at, and reorders the reading list so urgent cases come first. Good looks like: strong AUC per finding, heat maps that point to sensible regions, a triage simulation showing time saved, and an honest safety discussion.

What you will build

1. Prepare the images and labels; split by patient
2. Train a CNN (transfer learning) and evaluate per finding
3. Add heat maps (Grad-CAM) and review them with your mentor
4. Simulate triage and measure time to read urgent cases
5. Build a reading-list demo
6. Report performance, bias and clinical safety limits

Free, open data to build it with

• NIH Chest X-rays ↗ Kaggle (NIH data)
• Chest X-Ray Images (Pneumonia) ↗ Kaggle

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.