Synthetic 3D anatomical changes help improve proton therapy plans
SynthRCT: Scalable Conditional Deformation Synthesis for Synthetic Repeat CT Generation
Computer Vision and Pattern RecognitionArtificial IntelligenceMachine Learning
Summary
In proton therapy, doctors rely on CT scans to plan how to treat tumors, but patients’ bodies can change in complex ways making the plan less accurate. The authors created SynthRCT, a computer program that can generate realistic, patient-specific changes in 3D body images based on past observations, mimicking how anatomy might shift during treatment. This helps test treatment plans against more realistic scenarios, potentially improving safety and effectiveness. The system works efficiently even with large 3D images.
What this means in practice
- •For radiation therapy planners: Generate patient-specific anatomical variations to test robustness of proton therapy plans against realistic body changes.
- •For medical imaging software developers: Incorporate scalable synthetic deformation models to create varied anatomical scenarios for large CT volumes.$Commercial implications: Enables new software features for patient-specific treatment planning tools that can be sold to hospitals and clinics.
Authors
Tomas Guija-Valiente, Blanca Rodriguez-Gonzalez, Norberto Malpica
Abstract
In proton therapy, plans are typically optimized on a single planning CT, making robustness evaluation essential under anatomical changes. However, current scenarios often rely on simplified perturbations that poorly capture complex, patient-specific variability. We propose SynthRCT, a scalable conditional generative framework for 3D anatomical deformation synthesis. Based on a conditional variational autoencoder, SynthRCT learns a latent deformation space and decodes sampled latent codes into local stationary velocity fields conditioned on an input anatomy. Local fields are assembled into coherent full-volume transformations, enabling memory-scalable generation for large field-of-view CT data. We validate the approach on respiratory 4DCT data with multiple breathing-phase anatomies per subject. SynthRCT enables patient-specific sampling of plausible anatomical transformations beyond predefined robustness scenarios. Code available at: https://github.com/TomasGuija/SynthRCT.