Estimation of dose and linear energy transfer for multi-spot single-energy proton probing beams using in-beam PET and neural network.
Authors
Affiliations (11)
Affiliations (11)
- Biomedical Engineering Department, Huazhong University of Science and Technology, Luoyu Road 1037#, Wuhan, Hubei, 430074, China.
- Department of biomedical engineering, Huazhong University of Science and Technology, Luoyu Road 1037#, Wuhan, Hubei, 430074, China.
- Cancer Center, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, No. 53 Huanhu Road, Dongxihu District, Wuhan, 430048, China.
- Cancer Center, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, No. 53 Huanhu Road, Dongxihu District, Wuhan, Hubei, 430048, China.
- School of Software Engineering, Huazhong University of Science and Technology, Luoyu Road 1037#, Wuhan, Hubei, 430074, China.
- Department of Biomedical Engineering, Huazhong University of Science and Technology, Luoyu Road 1037#, Wuhan, 430074, China.
- Department of Electronic Engineering and Information Science, University of Science and Technology of China, No. 96 Jinzhai Road, Hefei, 230026, China.
- Department of Electronic Engineering and Information Science, University of Science and Technology of China, No. 96 Jinzhai Road, Hefei, Anhui, 230026, China.
- Department of Biomedical Engineering, Huazhong University of Science and Technology, Luoyu Road 1037#, Wuhan, Hubei, 430074, China.
- Huazhong University of Science and Technology School of Mechanical Science and Engineering, Luoyu Road 1037, Wuhan, China, Wuhan, Hubei, 430074, China.
- Cancer Center, Union Hospital, Tong ji Medical College, Huazhong University of Science and Technology, No. 53 Huanhu Road, Dongxihu District, WUHAN, 430023, China.
Abstract
In-beam positron emission tomography (PET) enables online measurements of proton-induced positron-emitting activity and has been widely investigated for in vivo range verification in proton therapy. This study proposes a neural network-based method for the simultaneous estimation of three-dimensional dose and linear energy transfer (LET) distributions from in-beam PET data following single-energy probing-beam irradiation. The performance and feasibility of the proposed method were investigated through Monte Carlo simulations and experimental measurements. 
Approach: Single-energy proton beams were delivered as probes to head computed tomography (CT) phantoms, with spot patterns of 3×3, 5×5, 7×7, and 9×9. Each individual spot delivered 2×107 protons, ensuring that the maximum physical dose per probing-beam irradiation remained below 0.5 Gy. A dual-panel in-beam PET system was simulated to reconstruct three-dimensional distributions of proton-induced positron-emitting activity. The reconstructed PET images, together with the corresponding CT images, served as inputs to neural network models that simultaneously predicted dose and LET distributions. The simulated datasets generated using nine CT phantoms were randomly divided into training (80%), validation (10%), and test (10%) sets. Model performance was quantitatively assessed using the mean relative error (MRE) and absolute range error (ARE). External generalization test and robustness evaluation were performed using two additional CT phantoms not included in model training. A preliminary feasibility evaluation using experimentally measured PET data was performed with an all-digital PET prototype and polymethyl methacrylate (PMMA) phantoms.
Main results: Across all spot patterns, even under adverse perturbation conditions, the models generated reasonable dose and LET predictions. The maximum median ARE, dose MRE, and LET MRE were 0.97 mm, 1.68%, and 1.84%, respectively. For the experimentally measured data, the median ARE ranged from 0.38 mm to 2.37 mm, while the median dose and LET MRE varied from 2.05% to 4.06% and 2.06% to 3.02%, respectively.
Significance: This study demonstrated the preliminary feasibility of simultaneously estimating three-dimensional dose and proton dose-averaged LET distributions from reconstructed in-beam PET and CT images under multi-spot single-energy probing-beam conditions. Further studies are required to establish quantitative dose-LET deviation criteria and to evaluate the additional value of the joint estimation framework for probing-beam delivery assessment.
.