Objective:To investigate whether the combination of the advantages of deep learining in image processing and radiotherapy will make the radiotherapy process more intelligent.Methods:The generative adversarial network (GAN) is a generation model using neural network. High-quality dose distribution images can be generated by inputting relevant features. Firstly, random unconditional GAN was utilized to verify the ideal data, then conditional GAN (cGAN) was employed to train DICOMRT data of tumor patients, and the target contour information was used to directly generate dose distribution images.Results:For the verification of ideal data, the generation of ideal distribution yielded good effect. By extracting target contour and real dose slice data and using cGAN training, the dose distribution maps of planning target volume (PTV) and organs at risk (OAR) of tumor patients could be obtained. The absolute error evaluation of the maximum and average values between the predicted value and the real dose was shown as[3.57%, 3.37%](PTV), [2.63%, 2.87%](brain), [1.50%, 2.70%](CTV), [3.87%, 1.79%](GTV), [3.60%, 3.23%](OAR1) and[4.40%, 3.13%](OAR2), respectively.Conclusions:GAN model can be used to generate ideal dose distribution data, and cGAN model with prior knowledge can be employed to establish the relationship between target information and dose distribution. Directly generating the corresponding dose distribution image by inputting the target contour information is an effective attempt for dose prediction.

With the application of magnetic resonance imaging (MRI)-guided photon therapy, the concept of combining real-time MRI guidance with proton therapy, namely, MRI-guided proton therapy (MRPT), has attracted widespread attention. It is expected that MRPT can mitigate the uncertainties during the treatment of proton therapy to make full use of the physical advantages of protons. However, multiple electromagnetic interactions between proton therapy and MRI-guided systems may lead to mutual interference between the two systems. This article review the research progress on the MRPT system, aiming to provide certain reference for the design of MRPT system.

Objective:Proton pencil beam (PB) dose calculation can achieve rapid dose calculation, whereas it is inaccurate due to the approximation in dealing with inhomogeneities. Monte Carlo (MC) dose calculation is recognized as the most accurate method, but it is extremely time consuming. The aim of this study was to apply deep-learning methods to improve the accuracy of PB dose calculation by learning the difference between the MC and PB dose distribution.Methods:A model which could convert the PB dose into the MC dose in lung cancer patients treated with intensity-modulated proton therapy (IMPT) was established based on the Hierarchically Densely Connected U-Net (HD U-Net) network. PB dose and CT images were used as model input to predict the MC dose for IMPT. The beam dose and CT images of 27 non-small cell lung cancer patients were preprocessed to the same angle and normalized, and then used as model input. The accuracy of the model was evaluated by comparing the mean square error and γ passing rate (1 mm/1%) results between the predicted dose and MC dose.Results:The predicted dose showed good agreement with MC dose. Using the 1 mm/1% criteria, the average γ passing rate (voxels receiving more than 10% of maximum MC dose) between the predicted and MC doses reached (92.8±3.4)% for the test patients. The average dose prediction time for test patients was (6.72±2.26) s.Conclusion:A deep-learning model that can accurately predict the MC dose based on the PB dose and CT images is successfully developed, which can be used as an efficient and practical tool to improve the accuracy of PB dose calculation for IMPT in lung cancer patients.

Objective:To demonstrate the concept of planning target volume (PTV) is not suitable for intensity proton therapy (IMPT) in lung cancer, plan differences were compared based on the concept of PTV and Internal target volume (ITV), aiming to provide clinical reference.Methods:Six patients were retrospectively selected and approved by the local ethics committee. Each of the six patients received two IMPT plans based on a synchronous accelerator model, developed by SINAP team (Shanghai Institute of Applied Physics, China Academy Science University) and commercial treatment system: one with the PTV-based robust IMPT (PTV-IMPT) plan and the other with ITV-based robust IMPT (ITV-IMPT) plan. Three beams were set in all plans, and the final dose was calculated using Monte Carlo dose algorithm. The plan quality and robustness of PTV-IMPT and ITV-IMPT plans were evaluated quantitatively.Results:Compared to the PTV-IMPT plan, ITV-IMPT plan showed better target conformity index (conformability index: 0.58 vs.0.43), better homogeneity index (homogeneity index: 0.96 vs.0.92), lower V 5Gy in normal lung tissue (13.1% vs.13.5%) and maximum dose in spinal cord (8.9 Gy vs. 9.5 Gy) as well as plan monitor unit (MU: 338 vs. 401) . In addition, ITV-IMPT plan showed more robust in target coverage (0.003-0.032 vs. 0.02-0.28), and normal lung tissue was also found a bit robust in the ITV-IMPT plan ( 0.06-0.11, 0.07-0.13). Conclusions:Compared with the PTV-IMPT plan, ITV-IMPT plan has the advantages of high planning quality, well robustness and better tumor motion mitigation. Therefore, ITV concept is recommended to be applied in the IMPT plan for lung cancer.

Objective:The purpose of current study was to evaluate the interplay effects in intensity-modulated proton therapy (IMPT) for lung cancer and compare the results of different Iso-energy layer repainting techniques in the mitigation of interplay effects.Methods:Eight patients with lung cancer who underwent 4DCT were retrospectively selected. A robust CTV-based IMPT plan was generated for each based on commercial TPS, considering patient setup errors ±5 mm, range uncertainties ±3.5%, and CTV time structure motion in 4DCT image. Monte Carlo dose engines were used for all IMPT plans in the final dose calculation. The 4D static dose (4DSD) and 4D dynamic dose (4DDD) were calculated using a hybrid deformable algorithm and simulated proton delivery system for interplay effects. An index[ΔI(ROI, DVH)] was developed to quantitatively evaluate the interplay effects. We applied Iso-energy layer repainting techniques with different numbers of repainting (3, 4, 5, 6, 7) to the robust IMPT plans and evaluated the difference in the mitigation of interplay effects based on the ΔI(ROI, DVH) index.Results:Due to interplay effects, the mean values of target coverage, conformity and homogeneity index reduced by 13.7%, 12.7% and 24.6%, respectively. The mean values of lung V 5Gy and V 20Gy improved by 0.8%, 3.4% and 2.6%. Compared to the IMPT plans without layer repainting, Multiple iso-energy layers repainting techniques improved the mean values of CTV coverage by 4.5%, 3.8%, 3.8%, 3.6% and 5.7%, respectively. The average values of lung V 20Gy reduced by 1.5%, 1.8%, 1.7%, 1.6% and 1.9%, respectively. Conclusions:In the robust CTV-based IMPT plans, the interplay effects degraded the target dose distribution but were mitigated using iso-energy layer repainting techniques. We recommended to use the layer repainting technique according to the characteristics of the patient.

Objective:To propose a new robust optimization method, known as modified worst case method, was proposed, which can enable users to control the trade-off between nominal plan quality and plan robustness.Methods:In each iteration of the plan optimization process, the dose value of each voxel in nine scenarios, which corresponded to a nominal scenario and eight perturbed scenarios with range or set-up uncertainties, were calculated and the maximum of deviations of each scenario voxel dose from that of the nominal scenario was included as an additive robust optimization term in the objective function. A weighting factor p robust was used to this robust optimization term to balance the nominal plan quality and plan robustness. Results:The robust optimization methods were implemented and compared in an in-house developed robust optimization module. When p robust=0.8, compared with conventional optimization, the ΔD 95% of CTV was reduced from 9.8 Gy to 7.6 Gy. When p robust was reduced from 1 to 0, ΔD 95% was increased from 7.0 Gy to 9.8 Gy, whereas the D 95% and D max of CTV, and the D 5% and D max of organs at risk (OAR) in the nominal scenario were reduced. Conclusions:The proposed modified worst case method can effectively improve the robustness of the plan to the range and set-up uncertainties. Besides, the weighting factor p robust in this method can be adopted to control the trade-off between nominal plan quality and plan robustness.

Objective:To develop and validate the accuracy of an independent dose calculation toolkit for the horizontal beamline of Shanghai Advanced Proton Therapy (SAPT) facility based on an open-source dose calculation engine.Methods:Machine data, such as absolute integral depth doses (IDDs) and lateral profiles in air were measured and lateral profiles in water were derived by Monte-Carlo simulations. The dose computation models for SAPT horizontal beamline pencil beams in water were achieved by combining machine data and dose calculation engine. The verification of the dose reconstruction toolkit included absolute dose verification and relative dose verification. The absolute dose verification is performed to mainly compare the reconstructed value and the measured value at different depths along the center axis of the beam direction of a cube plan. The relative dose verification is conducted to mainly compare the lateral profile or two-dimensional dose distribution between the measured value and the reconstructed value. Meanwhile, the precision of double-gaussian and single-gaussian lateral beam models was compared.Results:The deviations of the absolute dose between the calculated and measured values were basically within 2%. The deviations of 20%-80% penumbra between the measured and the calculated values were within 1 mm, and deviations of the full width at half height were within 2 mm. For 3 cube plans and 2 clinical cases, the two-dimensional gamma pass rates (3 mm/3%) between the measured and calculated dose distributions at the corresponding depths were greater than 95%. The double-gaussian lateral beam model was more accurate in the high dose gradient region and deeper depth.Conclusion:The precision of independent dose calculation toolkit is acceptable for clinical requirements, which can be employed to investigate other dose-related issues.

Objective:In Shanghai Advanced Proton Therapy Facility (SAPT) of Ruijin Hospital Proton Therapy Center, the calculation accuracy of the commercial proton treatment planning system RayStation (V10), especially the accuracy of the proton range calculation, was measured and verified, aiming to provide reference for the clinical application of the treatment planning system.Methods:A head phantom was used to verify the calculation accuracy of RayStation. The phantom CT was imported into treatment planning system (TPS). The phantom was followed closely by a water tank with a one-liter cubic target. A single field verification plan with the prescribed dose of 200 cGy (relative biological effectiveness) was designed and implemented. Then, the measured distribution results were compared with the calculation results.Results:When the verification plan of the phantom was designed with the default settings of RayStation, the measured longitudinal dose distribution was approximately 4 mm deeper than that of TPS, indicating that RayStation overestimated the water equivalent thickness (WET) of the tissue substitute materials in the phantom. To study the range error, the actual beam was used to measure the WET of the soft tissue substitute material. The default setting of RayStation was fine-tuned according to the measured results. It was found that the error between the measured SOBP and TPS calculations was reduced to only 2 mm.Conclusions:Using the default setting of RayStation to calculate the stopping power of the phantom may cause a large range error. A method that combines tissue segmentation with the measured WET of the tissue substitute material is proposed to improve the range calculation accuracy of the TPS. The results show that the proposed method can improve the dose and range accuracy of the commercial TPS including RayStation for tissue substitute materials.

Objective:To calculate the single-event dose-averaged specific energy of particles delivered in spherical domains based on the track structure model and using triple integration, and to investigate the influence of the domain shape on the key model parameters of microdosimetric kinetic model (MKM) and its corresponding physical significance.Methods:The domains are assumed to be cylinders and spheres, respectively. With α 0, domain radius, rd, and nucleus radius, Rn, as undetermined coefficients, the nuclear charge numbers, kinetic energies and their corresponding LETs of three kinds of charged particles ( 3He, 12C, 20Ne) as independent variables, D10 as dependent variable, the mean value of squared residuals, J2, between the D10 calculated values and D10 experimental values as the optimization objective, the final fitting values of the above undetermined coefficients of human salivary gland (HSG) cells and Chinese hamster lung (V79) cells obtained after iteration by the robust least square method are the optimal model parameter values of MKM. Results:For HSG cells, cylindrical domain: α 0=0.073/Gy, rd=0.29 μm, Rn=4.1 μm, J2=0.039 7 Gy 2; spherical domain: α 0=0.023/Gy, rd=0.29 μm, Rn=4.4 μm, J2=0.039 3 Gy 2; For V79 cells, cylindrical domain: α 0=0.114/Gy, rd=0.25 μm, Rn=3.8 μm, J2=0.097 4 Gy 2; spherical domain: α 0=0.095/Gy, rd=0.26 μm, Rn=4.1 μm, J2=0.096 9 Gy 2. Conclusions:For the same type of cells, cylindrical and spherical domains were selected respectively, and there are significant differences in MKM parameters obtained by fitting. The fitting values of the domain radius, rd of the two shapes of domains show no significant difference, while the fitting values of α0 of spherical domains are smaller than those of cylindrical domains, the fitting values of nucleus radius, Rn, of spherical domain are larger than those of cylindrical domains, closer to the nucleus radius observed by fluorescence microscopy. In the low LET (<20 keV/μm) region, D10 calculated according to the parameters of the two different shapes of domains are different, so the selection of the domain shape will cause differences in the relative biological effectiveness(RBE) calculation of proton in the region near Bragg peak.