Mini-project
Healthcare & Hospitals
Chest X-ray pneumonia screener
Flag chest X-rays that likely show pneumonia, for a doctor to review.
The brief
Rural hospitals have X-ray machines but few radiologists. Train an image model that flags X-rays likely to show pneumonia, so a doctor reviews those first. It must never replace the doctor, and it must show where in the image it is looking. Good looks like: high sensitivity, honest metrics on a held-out set, heat maps that point at the lungs, and a clear limits statement.
What you will build
- Prepare the images and keep patients separate between splits
- Fine-tune a pre-trained image model
- Evaluate sensitivity, specificity and AUC
- Show heat maps (Grad-CAM) for each prediction
- Write the model's limits and intended use
Free, open data to build it with
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.