This project is a lightweight web application for running patchwise inference of diagnostics and BRAF V600E mutation on 3D volumetric micro-CT scans of thyroid tumors.

• Patch-based Inference on 3D micro-CT images.
• Multi-task Prediction: BRAF V600E mutation status and diagnostic classification.
• TIFF Export of prediction maps.
• End-to-End Pipeline from upload → inference → export

git clone https://github.com/kiataj/ThyVision.git
cd ThyVision

docker build -t thyvision .

docker run -p 8000:8000 -v C:\User\app:/data thyvision 

You can replace C:\User\app with a local derive of your preference for saving the predictions.

Once the Docker container is running, the web interface will be available at:
http://localhost:8000/

1. Upload a 3D .tiff volume.
2. Configure patch and stride size.
3. Click Run Inference - wait until the buttons are activated again, the progress bar can be seen in the command prompt that the docker container is running in.
4. Click Save Predictions to save the predictions in the mounted derive.

#### Directory structure
├── main.py                   # FastAPI backend
├── inference_module.py       # Inference logic (with tqdm)
├── app/                      # React frontend (Vite)
└── Input example/            # An input example .tif file you can use to run inference.