A contour-guided deformable image registration algorithm for adaptive radiotherapy.

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Medical Image Analysis

Background:

• Deformable image registration is crucial for adaptive radiotherapy, enabling dose accumulation and contour propagation.
• Auto-propagated contours often exhibit errors, particularly in low-contrast areas, necessitating manual correction by clinicians.
• Existing methods yield inaccurate deformation vector fields (DVFs) inconsistent with edited contours, impacting adaptive therapy precision.

Purpose of the Study:

• To develop a novel contour-guided deformable image registration (CG-DIR) algorithm.
• To enhance the accuracy and consistency of the deformation vector field (DVF) for adaptive radiotherapy applications.
• To ensure clinical requirements for adaptive therapy are met with improved image registration.

Main Methods:

• Developed CG-DIR by regularizing the demons algorithm's objective function with intensity matching of delineated structures.
• Incorporated clinician-edited contours into the registration process to guide DVF generation.
• Implemented the CG-DIR algorithm on graphics processing units (GPUs) for efficient computation.

Main Results:

• CG-DIR demonstrated improved accuracy and consistency of the DVF compared to the original demons algorithm.
• The algorithm maintained high computational efficiency, similar to the original GPU-based demons method.
• Evaluated on diverse patient data, including head-and-neck and pelvic cancers, confirming robust performance.

Conclusions:

• The contour-guided deformable image registration (CG-DIR) algorithm effectively improves DVF accuracy and consistency in adaptive radiotherapy.
• CG-DIR offers a viable solution for addressing contour inaccuracies and enhancing the reliability of adaptive treatment planning.
• The method provides a computationally efficient approach for clinical implementation in adaptive radiotherapy workflows.