Adaptive radiotherapy in locally advanced prostate cancer using a statistical deformable motion model

Area of Science:

• Radiation Oncology
• Medical Imaging
• Biomedical Engineering

Background:

• Adaptive radiotherapy relies on daily treatment plan selection to manage anatomical variations.
• The accuracy of this strategy hinges on initial images representing later treatment course variations.
• Prostate cancer treatment requires precise targeting of the prostate (CTV-p) and surrounding lymph nodes (CTV-ln, CTV-sv).

Purpose of the Study:

• To evaluate the efficacy of a statistical deformable model in prospectively accounting for residual prostate and elective target motion.
• To determine if models based on early treatment images can predict later anatomical variations.
• To assess the feasibility of using principal component analysis (PCA) for motion modeling in locally advanced prostate cancer.

Main Methods:

• Utilized repeat CT scans from 13 locally advanced prostate cancer patients.
• Developed patient-specific statistical motion models using PCA on displacement vector fields (DVFs) from deformable image registration.
• Created PCA models with varying numbers of DVFs (4, 5, 6, and all available) to simulate target shapes and assess coverage.

Main Results:

• Simulated shapes from PCA models using 4-6 DVFs covered at least 95% of evaluated CTV volumes.
• Sensitivity decreased with higher iso-coverage levels, particularly for larger target movements.
• The CTV-sv demonstrated the most significant influence on model geometry and showed the greatest range in sensitivity and precision.

Conclusions:

• PCA-based simulations using 4-6 DVFs effectively account for the majority of target shape variations during prostate cancer radiotherapy.
• The CTV-sv exhibits the widest variability in both sensitivity and precision, highlighting its importance in motion management.
• This approach supports the prospective adaptation of radiotherapy plans for improved treatment accuracy.