GPU-accelerated Monte Carlo-based online adaptive proton therapy: A feasibility study.

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Imaging

Background:

• Interfraction anatomical changes in patients undergoing pencil beam scanning (PBS) proton therapy can degrade treatment plan quality.
• Online adaptive radiation therapy (ART) is crucial for addressing these dynamic changes.
• Monte Carlo (MC) methods offer high accuracy for dose calculation but are computationally intensive.

Purpose of the Study:

• To develop and evaluate an online graphic processing unit (GPU)-accelerated Monte Carlo-based ART workflow for PBS proton therapy.
• To address interfraction anatomical variations by enabling rapid plan adaptation.
• To ensure treatment plan quality and accuracy throughout the therapy course.

Main Methods:

• A four-step workflow integrating GPU-accelerated MC dose calculation and deformable image registration (DIR) was developed.
• The workflow includes contour propagation, verification dose calculation, plan reoptimization, and two-stage patient-specific quality assurance (PSQA).
• MCsquare and ClearCheck™ were utilized for independent dose verification and plan evaluation, respectively.

Main Results:

• The ART workflow demonstrated efficient contour propagation with good volume conformance but suboptimal boundary coincidence for organs-at-risk.
• Reoptimization significantly improved target coverage and plan quality to clinically acceptable levels after interfraction changes were detected.
• 3D Gamma analyses confirmed the accuracy of the reoptimized plan, with mean Gamma indices of 98.74% (pre-delivery) and 99.05% (post-delivery).

Conclusions:

• The proposed online ART workflow for PBS proton therapy is efficient and effective.
• The GPU-accelerated MC approach successfully generates reoptimized plans, significantly improving treatment quality.
• This workflow provides a viable solution for managing anatomical changes during PBS proton therapy.