Analyzing a full 3D micro-CT volume of a thyroid tumor manually is not practical. ThyVision is a web application that automates this process: it takes a 3D micro-CT scan, runs patchwise inference across the entire volume, and produces spatial prediction maps for both tumor diagnostics and BRAF V600E mutation status.
The tool was developed as part of the PHRT thyroid cancer initiative to make the radiomics and deep learning models from the broader thyroid project accessible without requiring any programming knowledge. A pathologist or researcher can upload a scan, configure the patch size and stride, and export the resulting prediction maps directly from the browser.
Key features:
- Patchwise inference over full 3D micro-CT volumes
- Simultaneous prediction of tumor type and BRAF V600E mutation status
- Spatial output maps exported as TIFF for downstream analysis
- Web interface built with React and FastAPI, deployable via Docker
The source code is available on GitHub.