Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·27d ago·5 min read
Physically Aware Radiomics Without Interpolation: Disentangling Voxel Geometry and Signal Modification in CT and MRI
TL;DR
A voxel-spacing-aware radiomic framework was developed to compute texture features in CT and MRI without resampling, achieving high agreement with native extraction methods. This approach effectively separates voxel geometry from signal modification, enhancing radiomic analysis in anisotropic images.
✦ Why It Matters
Implement voxel-spacing-aware extraction in your radiomic analyses to improve accuracy without resampling.
Key Takeaways
- Use voxel-spacing-aware extraction for accurate radiomic analysis.
- Avoid isotropic resampling to prevent signal distortion.
- Test multiple extraction configurations for robust results.
- Leverage high ICC values to validate your radiomic features.
How It Works
The voxel-spacing-aware framework modifies the PyRadiomics tool to account for the physical geometry of voxels during texture feature extraction. By doing so, it avoids the pitfalls of interpolation that can distort signal characteristics in anisotropic images, allowing for more accurate analysis without altering the original image data.