Denoising proton therapy Monte Carlo dose distributions in multiple tumor sites: A comparative neural networks architecture study. (September 2021)

• Record Type: Journal Article
• Authors: Javaid, Umair; Souris, Kevin; Huang, Sheng; Lee, John A.
• Is Part Of: Physica medica. Volume 89(2021)
• Page Start: 93
• Page End: 103
• Publication Date: 2021-09
• Subjects: Artificial intelligence -- Convolutional neural networks -- Dose denoising -- Monte Carlo -- Proton therapy; Medical physics -- Periodicals; Biophysics -- Periodicals; Biophysics -- Periodicals; Imagerie médicale -- Périodiques; Radiothérapie -- Périodiques; Rayons X -- Sécurité -- Mesures -- Périodiques; Physique -- Périodiques; Médecine -- Périodiques; 610.153
• DOI: 10.1016/j.ejmp.2021.07.022
• Languages: English
• ISSNs: 1120-1797

Abstract

Highlights: Accelerate Monte Carlo dose calculation in multiple tumor sites for proton therapy. Predict dose maps similar to Monte Carlo simulations with 1 × 109 protons. Basic CNN (sNet) offers better performance than the complex dUNet for MC denoising. Scalability from 2.5D to 3D for better generalization and faster network training. Abstract: Introduction: Monte Carlo (MC) algorithms provide accurate modeling of dose calculation by simulating the delivery and interaction of many particles through patient geometry. Fast MC simulations using large number of particles are desirable as they can lead to reliable clinical decisions. In this work, we assume that faster simulations with fewer particles can approximate slower ones by denoising them with deep learning. Materials and methods: We use mean squared error (MSE) as loss function to train networks (sNet and dUNet), with 2.5D and 3D setups considering volumes of 7 and 24 slices. Our models are trained on proton therapy MC dose distributions of six different tumor sites acquired from 50 patients. We provide networks with input MC dose distributions simulated using 1 × 10 6 particles while keeping 1 × 10 9 particles as reference. Results: On average over 10 new patients with different tumor sites, in 2.5D and 3D, our models recover relative residual error on target volume, Δ D 95 TV of 0.67 ± 0.43% and 1.32 ± 0.87% for sNet vs. 0.83 ± 0.53% and 1.66 ± 0.98% for dUNet, compared to the noisy input at 12.40 ± 4.06%. Moreover, the denoising time for a dose distribution is: < 9s and < 1s for sNet vs. < 16s and < 1.5s for dUNet in 2.5D and 3D, in comparison to about 100 min (MC simulation using 1 × 10 9 particles). Conclusion: We propose a fast framework that can successfully denoise MC dose distributions. Starting from MC doses with 1 × 10 6 particles only, the networks provide comparable results as MC doses with1 × 10 9 particles, reducing simulation time significantly.