Repository navigation
fastrad v1.0.0: Initial Clinical Release
FastRadiomics (fastrad) v1.0.0 is officially released!
This marks the first stable clinical deployment of the fastrad library.
Core Features
- Full IBSI Compliance: Matches Image Biomarker Standardisation Initiative reference values natively.
- PyRadiomics Parity: Verified 100% numerical parity against PyRadiomics for First Order, Shape, GLCM, GLRLM, GLSZM, GLDM, and NGTDM features.
- Hardware Acceleration: Native PyTorch backend seamlessly leverages CUDA for massive texture matrix evaluations, performing up to 30x faster than legacy libraries.
- Graceful OOM Logic: Inbuilt fallback protection routing VRAM overflows to system CPU dynamically to prevent clinical pipeline crash loops.
- Awesome Documentation: Completely overhauled, mathematically rigorous ReadTheDocs deployment.
This release integrates directly with PyPI and supports pip install fastrad.