Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·19h ago
AI Research
arXiv CS.AI·Jun 10·5 min read
Deep Slice Interpolation for Reducing Through-Plane Anisotropy and Noise in Head CT
TL;DR
Head CT scans often suffer from through-plane anisotropy, which leads to inconsistent image quality and increased noise. A new technique called Deep Slice Interpolation was developed to address these issues by enhancing the quality of CT images.
✦ Why It Matters
Engineers can implement Deep Slice Interpolation to improve image quality in medical imaging applications.
Key Takeaways
- Utilize deep learning for improved CT slice interpolation.
- Evaluate models using a combination of loss functions for best results.
- Implement implicit denoising techniques in imaging workflows.
- Test new models on out-of-distribution data for robustness.
How It Works
The model synthesizes intermediate slices by analyzing pairs of neighboring axial slices, effectively reducing the through-plane spacing from 2-5 mm to 1-2.5 mm. It employs a combination of pixel-wise losses and structural similarity metrics to optimize image quality, ensuring that the generated slices maintain high fidelity and low noise.