Study of three-dimensional dose distribution based-deep learning in predicting distant metastasis in head and neck cancer / 中华放射肿瘤学杂志

Objective:To investigate the role of three-dimensional dose distribution-based deep learning model in predicting distant metastasis of head and neck cancer.Methods:Radiotherapy and clinical follow-up data of 237 patients with head and neck cancer undergoing intensity-modulated radiotherapy (IMRT) from 4 different institutions were collected. Among them, 131 patients from HGJ and CHUS institutions were used as the training set, 65 patients from CHUM institution as the validation set, and 41 patients from HMR institution as the test set. Three-dimensional dose distribution and GTV contours of 131 patients in the training set were input into the DM-DOSE model for training and then validated with validation set data. Finally, the independent test set data were used for evaluation. The evaluation content included the area under receiver operating characteristic curve (AUC), balanced accuracy, sensitivity, specificity, concordance index and Kaplan-Meier survival curve analysis.Results:In terms of prognostic prediction of distant metastasis of head and neck cancer, the DM-DOSE model based on three-dimensional dose distribution and GTV contours achieved the optimal prognostic prediction performance, with an AUC of 0.924, and could significantly distinguish patients with high and low risk of distant metastasis (log-rank test, P<0.001). Conclusion:Three-dimensional dose distribution has good predictive value for distant metastasis in head and neck cancer patients treated with IMRT, and the constructed prediction model can effectively predict distant metastasis.

Dosimetric effects of isocenter on intensity-modulated radiotherapy for MR-Linac / 中华放射医学与防护杂志

Objective:To analyze the dosimetric effects on off-center tumour treatment plan resulting from the MR-Linac-based isocenter position radiotherapy plan.Methods:The cases of 19 patients who were treated in Sun Yat-sen University Cancer Center in 2020 were collected in this study. Two different IMRT plans were designed for each patient with off-center tumor both for group A with planned isocenter position as IMRT and group B with planed target center position as geometric center. The conformity index and homogeneity index of target, the dose normal tissue and the number of MU were compared between two plans.Results:The two IMRT plans met clinical dosimetric requirements. No statistical differences were found both in homogeneity index and conformity index ( P>0.05). Also there was no differences found in doses to normal tissues. However, the MU number (1 149±903, t=2.804, P=0.012) in group A was higher than that in group B (970±652). Conclusions:It is feasible to perform MR-Linac-based off-center treatment plan.

A preliminary study on the out-of-field in-air electron streaming effect in MRI guided radiotherapy / 中华放射医学与防护杂志

Objective:To investigate the impacts of electron streaming effect (ESE) on out-of-field dose distribution in 1.5 T MRI-guided radiotherapy.Methods:Firstly, the Monaco v5.40.1 (Elekta AB, Stockholm, Sweden) treatment planning system (TPS) was implemented to investigate the ESE in a square field (5 cm × 5 cm) at the entry and exit sides of a special homogeneous water phantom. Afterward, a retrospective investigation was conducted into one laryngeal cancer case and one breast cancer case who had been treated on a conventional linear particle accelerator (linac). Then doses were recalculated in the Monaco system using a Unity machine model. Meanwhile, the out-of-field skin dose enhancement induced by ESE was investigated.Results:ESE-induced dose variations were observed at both the entry and exit sides of the phantom surface in the presence of a magnetic field, with the ESE on the exit side notably stronger than that on the entry side. For the laryngeal cancer case, the ESE was not notable and had insignificant impacts on the out-of-field skin dose. In contrast, ESE-induced in-air high-dose region outside the body stretched to the chin area for the breast cancer case. This led to the skin dose escalation of the chin at D1 cm 3 454.6 cGy. After the application of 1 cm bolus, the corresponding skin dose of the chin D1 cm 3 reduced to as low as 113.6 cGy, which is almost equivalent to that in the absence of a magnetic field ( D1 cm 3=92.5 cGy). Conclusions:The ESE in a magnetic field can alter out-of-field dose and lead to local dose enhancement along the electron path. Although the ESE had insignificant impacts on the out-of-field dose of the laryngeal cancer case, it reached the chin area of the breast cancer case. ESE can be effectively shielded by adding protective bolus.

Dose evaluation of adaptive radiotherapy using high-field MR-Linac systems for head and neck cancers / 中华放射医学与防护杂志

Objective:To investigate the feasibilityof the adaptive radiotherapy using high-field MR-Linac systems for head neck cancers and perform the evaluation of target coverage and dose criteria.Methods:This study investigated 128 treatment plans of six patients who were treated on 1.5T MR-Linacsystems in Sun Yat-sen University Cancer Center in 2019, compared the differences in target coverage and dose criteria between the dose accumulation in the adaptive radiotherapy using MR-Linac systems and the reference plans, and evaluated the target coverage and dose criteria of each fraction of adaptive plan based on daily MRI anatomy.Results:There was no significant change in the target coverage and dose criteria for each treatment fraction(<1%). However, the change of lens dose was significant (maximum 98%). In addition, the result showed that there was no significant difference in target coverage and dose criteria between the dose accumulation in adaptive radiotherapy using MR-Linac systems and reference plans.In contrast, the average dose to lens was increased by 31.7%.Conclusions:It is feasible to perform adaptive radiotherapy using 1.5T MR-Linacsystems for head neck cancers according tothe evaluation of target coverage and dose criteria. Additionally, since the actual dose tolens was quite different from the reference plan, the lens exposure should be considered in clinical practice.

Different receptive fields-based automatic segmentation network for gross target volume and organs at risk of patients with nasopharyngeal carcinoma / 中华放射肿瘤学杂志

Objective:To establish an automatic segmentation network based on different receptive fields for gross target volume (GTV) and organs at risk in patients with nasopharyngeal carcinoma.Methods:Radiotherapy data of 100 cases of nasopharyngeal carcinoma including CT images and GTV and organs at risk delineated by the physicians were collected. Ninety plans were randomly selected as the training dataset, and the other 10 plans as the validation dataset. Firstly, the images were subject to three data augmentation methods including center cropping, vertical flipping and rotation (-30°to 30°), and then input into MA_net networks proposed in this study for training. The model performance of networks was assessed by the number of network parameters (NP), floating-point number (FPN), the running memory (RM) and Dice index (DI), and eventually compared with DeeplabV3+ , PSP_net, UNet+ + and U_Net networks.Results:When the input image was in the size of 240×240, MA_net had a NP of 23.20%, 20.10%, 25.55% and 27.11% of these 4 networks, 50.02%, 19.86%, 6.37% and 13.44% for the FPN, 40.63%, 23.60%, 11.58% and 14.99% for the RM, respectively. For the DI of GTV, MA_net was 1.16%, 2.28%, 1.27% and 3.59% higher than these 4 networks. For the average DI of GTV and OAR, MA_net was 0.16%, 1.37%, 0.30% and 0.97% higher than these 4 networks.Conclusion:Compared with those four networks, the proposed MA_net network has slightly higher Dice index with fewer parameters, lower FPN and smaller RM.

Characterization of imaging distortion of high-field magnetic resonance imaging-guided linear accelerator (MR-Linac) and its influencing factors / 中华放射肿瘤学杂志

Objective:To characterize the imaging distortion of the 1.5T magnetic resonance imaging-guided linear accelerator (MR-Linac) and to analyze the influence of MR-Linac and peripheral devices on the geometric distortion.Methods:Specialized MRI imaging distortion phantom and analysis software were applied. The baseline of imaging distortion within diameter spherical volume (DSV) around the center of the magnet was established. The influence of the beam generation system, mechanical system and peripheral devices on the imaging distortion was analyzed. The long-term stability of imaging distortion was tested on the MR-Linac.Results:Imaging distortion of the MR-Linac was increased with the increasing distance to the center of the magnet. Within DSV 400 mm, few test points surpassed 1 mm imaging distortion in 3D directions. However, imaging distortion surpassed 2 mm in part of region within DSV 400-500 mm, with the largest distortion over 7 mm. Imaging distortion of the MR-Linac remained unchanged within 7 months after installation. And the influence of the MR-Linac and peripheral devices on the imaging distortion was only observed in the overall largest distortion within DSV 400-500 mm.Conclusions:Cautions should be taken during the application of high-field MR-Linac in patients whose tumor location is over 20 cm from the ISO center. Imaging distortion of the MR-Linac remains stable within 7 months after installation. The influence of the MR-Linac and peripheral devices on the imaging distortion is trivial, which can be neglected in clinical practice.

The clinical application of magnetic resonance-guided radiotherapy / 中华放射肿瘤学杂志

Objective:To investigate the clinical feasibility of the Unity radiotherapy system guided by magnetic resonance imaging.Methods:Twenty-four patients were enrolled and received a total of 384 fractions of treatment at Unity system. According to the treatment site, all patients were divided into head-neck, abdomen-thorax, pelvic, spine and limb groups. The patients were set-up without external laser. And then, the time required at different stages in online treatment process and the registration error of each fraction were separately calculated. The geometric deformations of MR images were weekly measured by using MR geometric deformation phantom. At last, the Arccheck was used to perform the dose verification of reference plan, online plan and offline plan.Results:The mean duration of radiotherapy in the five groups were 29.1, 27.6, 26.6, 25.6 and 32.0 min, respectively. The set-up errors in the left-right, superior-inferior and anterior-posterior direction in the five groups were: head-neck group (0.08±0.06 cm, 0.16±0.13 cm, 0.08±0.05 cm), abdomen-thorax group (0.23±0.18 cm, 0.50±0.47 cm, 0.12±0.1 cm), pelvic group (0.25±0.19 cm, 0.32±0.25 cm, 0.11±0.09 cm), spine group (0.46±0.38 cm, 0.26±0.26 cm, 0.13±0.07 cm) and limb group (0.33±0.30 cm, 0.34±0.23 cm, 0.08±0.06 cm), respectively. In the central region, the geometric deformation of MR was less than 0.3 mm, and that of the sphere with a diameter of 500 mm was less than 2.1 mm. The meanγ pass rate of the reference plan, online plan and offline plan were 97.92%, 97.84% and 94.58%, respectively.Conclusions:MR-guided radiotherapy has great potential for clinical application, whereas the process of Unity system is relatively complex. The synergy of different departments has a great impact on the treatment, which needs further optimization.

Evaluation of three predictive models of knowledge-based treatment strategies for radiotherapy / 中华放射肿瘤学杂志

Objective:To compare the accuracy and generalized robustness of three predictive models of knowledge-based treatment strategies for radiotherapy for optimized model selection.Methods:The clinical radiotherapy plans of 45 prostate cancer (PC) cases and 25 nasopharyngeal cancer (NPC) cases were collected, and analyzed using three models (Z, L and S model), proposed by Zhu et al, Appenzoller et al and Shiraishi et al, respectively, to predict the dose-volume histogram (DVH) of bladder and rectum on PC cases and that of left and right parotid on NPC cases. The prediction error was measured by the difference of area under the predicted DVH and the clinical DVH curves (|V (pre_DVH)-V (clin_DVH)|), where a smaller prediction error implies a greater prediction accuracy. The accuracies of these three models were compared on the single organ at risk (OAR), and the generalized robustness of models was evaluated and compared by calculating the standard deviation of the prediction accuracy on different OAR. Results:For bladder and rectum, the prediction error of L model (0.114 and 0.163, respectively) was significantly higher than those values of Z and S models (≤0.071, P<0.05); for left parotid gland, the predicted error of S model (0.033) did not present significant difference from those values of Z and L models (≤0.025, P>0.05); for right parotid gland, S model (0.033) demonstrated significantly higher prediction error than those of Z and L models (≤0.028, P<0.05). Regarding different OAR, S model showed a lower standard deviation of prediction accuracy when comparing to Z and L models (0.016, 0.018 and 0.060, respectively). Conclusions:In the prediction of DVH in bladder and rectum of PC, Z and S models were more accurate than L model. In contrast, Z and L models demonstrated higher accuracy than S model in the prediction of left and right parotid glands of NPC. In respect to different OAR, the generalized robustness of S model was superior than the other two models.

Generative Adversarial Networks based synthetic-CT generation for patients with nasopharyngeal carcinoma / 中华放射肿瘤学杂志

Objective:To establish a correlation model between MRI and CT images to generate synthetic-CT (sCT) of head and neck cancer during MRI-guided radiotherapy by using generative adversarial networks (GAN).Methods:Images and IMRT plans of 45 patients with nasopharyngeal carcinoma were collected before treatment. Firstly, the MRI (T1) and CT images were preprocessed, including rigid registration, clipping, background removal and data enhancement, etc. Secondly, the cases were trained by GAN, of which 30 cases were randomly selected and put into the network as training set images for modeling and learning, and the other 15 cases were used for testing. The image quality of predicted sCT and real CT were statistically compared, and the dose distribution recalculated upon predicted sCT was statistically compared with that of real planned dose distribution.Results:The mean absolute error of the predicted sCT of the testing set was (79.15±11.37) HU, and the SSIM value was 0.83±0.03. The MAE values of dose distribution difference at different regional levels were less than 1% compared to the prescription dose. The gamma passing rate of the sCT dose distribution was higher than 92% and 98% under the 2mm/2% and 3mm/3% criteria.Conclusions:We have successfully proposed and realized the generation of sCT for head and neck cancer using GAN, which lays a foundation for the implementation of MRI-guided radiotherapy. The comparison of image quality and dosimetry shows the feasibility and accuracy of this method.

Design of a semi-spherical applicator for intraoperative radiotherapy with a linear accelerator and assessment of its dosemetric characteristics based on Monte Carlo simulation / 中华放射医学与防护杂志

Objective:To design a semi-spherical applicator for delivery of semi-spherical dose distributions and assess its dosemetric characteristics.Methods:The applicator was designed in the following way. First, the scattering angle and dose rate of the electron beam having passed through a series of scattering foils of different thicknesses were calculated to determine the thickness of the scattering foil. And then, a series of location model was designed, and the variances of the mean electron energy on the surface of these models were calculated to determine the foil location. Finally, the relationship between the geometric characteristics of the layer and the surface dose on the applicator was established to design the modulator. Monte Carlo (MC) codes EGSnrc/BEAMnrc and EGS4/DOSXYZ were employed to model the head of the Mobetron, the location model, the layer, the semi-spherical applicator, and to calculate the dose distributions.Results:A semi-spherical applicator was designed for electron beam of energy 12 MeV, which consisted of a 2.5 cm diametre cylindrical collimator with 0.5 cm thick wall made of 0.3 cm thick steel and 0.2 cm thick water equivalent material (WEM), a 0.14 cm-thick foil made of tansgen, and a 2.5 cm diametre hollow semi-sphere containing a crescent modulator made of WEM. The dose rate was about 160 cGy/min, and the depth of the 50% isodose curve was 0.85 cm.Conclutions:We designed and performed a MC simulation of a semi-spherical applicator to deliver a semi-spherical dose distribution from a high energy electron beam.

Multi-task learning-based three-dimensional dose distribution prediction for multiple organs in a single model / 中华放射肿瘤学杂志

Objective To establish a three-dimensional (3D) dose prediction model,which can predict multiple organs simultaneously in a single model and automatically learn the effect of the geometric anatomical structure on dose distribution.Methods Clinical radiotherapy plans of patients diagnosed with the same type of tumors were collected and retrospectively analyzed.For every plan,each organs at risk (OAR) voxel was regarded as the study sample and its deposited dose was considered as the dosimetric feature.A regularized multi-task learning method than could learn the relationship among different tasks was employed to establish the relationship matrix among tasks and the correlation between geometric structure and dose distribution among organs.In this experiment,the spinal cord,brainstem and bilateral parotids involved in the intensity-modulated radiotherapy (IMRT) plan of 15 nasopharyngeal cancer patients were utilized to establish the multi-organ prediction model.The relative percentage error between the predicted dose of voxel and the clinical planning dose was calculated to assess the feasibility of the model.Results Ten cases receiving IMRT plans were utilized as the training data,and the remaining five cases were used as the test data.The test results demonstrated a higher prediction accuracy and less data demand.And the average voxel dose errors among the spinal cord,brainstem and the left and right parotids were (2.01±0.02)％,(2.65± 0.02) ％,(2.45± 0.02) ％ and (2.55± 0.02) ％,respectively.Conclusion The proposed model can accurately predict the dose of multiple organs in a single model and avoid the establishment of multiple single-organ prediction models,laying a solid foundation for patient-specific plan quality control and knowledge-based treatment planning.