Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Aug 4·5 min read
Anticipatory Digital Twins for Online Head-and-Neck Adaptive Proton Therapy via Foundation-Model Registration
TL;DR
A digital twin framework was developed to predict treatment-day anatomy for head-and-neck proton therapy, improving treatment accuracy without the need for repeated imaging. This approach demonstrated significant enhancements in anatomical matching, particularly for patients experiencing major changes.
✦ Why It Matters
Implement this digital twin approach to enhance accuracy in adaptive proton therapy without additional imaging delays.
Key Takeaways
- Utilize digital twin technology for real-time anatomical predictions.
- Reduce reliance on repeated imaging during treatment.
- Enhance treatment accuracy for patients with significant anatomical changes.
- Implement foundation-model registration for efficient data processing.
How It Works
The framework employs a two-step registration process. First, it aligns a prior patient's planning CT to the current target anatomy.
Then, it estimates the anatomical changes from the prior's quality assurance CT (QACT) and applies these changes to the current planning CT, generating predicted CTs (pdCTs) that reflect the anticipated treatment-day anatomy.