Search

Background: Patients may undergo anatomical changes during radiotherapy, leading to an underdosing of the target or overdosing of the organs at risk (OARs)., Purpose: This study developed a deep-learning method to predict the tumor response of patients with nasopharyngeal carcinoma (NPC) during treatment. This method can predict the anatomical changes of a patient., Methods: The participants included 230 patients with NPC. The data included planning computed tomography (pCT) and routine cone-beam CT (CBCT) images. The CBCT image quality was improved to the CT level using an advanced method. A long short-term memory network-generative adversarial network (LSTM-GAN) is proposed, which can harness the forecasting ability of LSTM and the generation ability of GAN. Four models were trained to predict the anatomical changes that occurred in weeks 3-6 and named LSTM-GAN-week 3 to LSTM-GAN-week 6. The pCT and CBCT were used as input, and the tumor target volumes (TVs) and OARs were delineated on the predicted and real images (ground truth). Finally, the models were evaluated using contours and dosimetry parameters., Results: The proposed method predicted the anatomical changes, with a dice similarity coefficient above 0.94 and 0.90 for the TVs and surrounding OARs, respectively. The dosimetry parameters were close between the prediction and ground truth. The deviations in the prescription, minimum, and maximum doses of the tumor targets were below 0.5 Gy. For serial organs (brain stem and spinal cord), the deviations in the maximum dose were below 0.6 Gy. For parallel organs (bilateral parotid glands), the deviations in the mean dose were below 0.8 Gy., Conclusion: The proposed method can predict the tumor response to radiotherapy in the future such that adaptation can be scheduled on time. This study provides a proactive mechanism for planning adaptation, which can enable personalized treatment and save clinical time by anticipating and preparing for treatment strategy adjustments., (© 2024 American Association of Physicists in Medicine.)

Background: Volumetric modulated arc therapy (VMAT) planning optimization involves iterative adjustment of numerous parameters, and hence requires repeatedly dose recalculation. In this study, we used the deep learning method to develop a fast and accurate dose calculation method for VMAT., Methods: The classical 3D UNet was adopted and trained to learn the physics principle of dose calculation. The inputs included the projected fluence map (FM), computed tomography (CT) images, the radiological depth and the source-to-voxel distance (SVD). The projected FM was generated by projecting the accumulated FM between two consecutive control points (CPs) onto the patient's anatomy. The accumulated FM was calculated by simulating the movement of the multi-leaf collimator (MLC) from one CP to the next. The dose, calculated by the treatment planning system (TPS), was used as ground truth. 51 head and neck VMAT plans were used, with 43, 1 and 7 cases as training, validation, and testing datasets, respectively. Correspondingly, 7182, 180 and 1260 CP samples were included in the training, validation, and testing datasets., Results: This presented method was evaluated by comparing the derived dose distribution to the TPS calculated dose distribution. The dose profiles coincided for both the single CP and the entire plan (summation of all CPs). But the network derived dose was smoother than the TPS calculated dose. Gamma analysis was performed between the network derived dose and the TPS calculated dose. The average gamma pass rate was 96.56%, 98.75%, 98.03% and 99.30% under the criteria of 2% (tolerance) -2 mm (distance to agreement, DTA). 2%-3 mm, 3%-2 mm and 3%-3 mm. No significant difference was observed on the critical indices including the max, mean dose, and the relative volume covered by the 2000 cGy, 4000 cGy and the prescription dose. For one CP, the average computational time of the network and TPS was 0.09s and 0.53s. And for one patient, the average time was 16.51s and 95.60s., Conclusion: The dose distribution derived by the network showed good agreement with the TPS calculated dose distribution. The computational time was reduced to approximate one-sixth of its original duration. Therefore the presented deep learning-based dose calculation method has the potential to be used for planning optimization., (© 2024. The Author(s).)

Purpose: The purpose of the study is to investigate the clinical application of deep learning (DL)-assisted automatic radiotherapy planning for lung cancer., Methods: A DL model was developed for predicting patient-specific doses, trained and validated on a dataset of 235 patients with diverse target volumes and prescriptions. The model was integrated into clinical workflow with DL-predicted objective functions. The automatic plans were retrospectively designed for additional 50 treated manual volumetric modulated arc therapy (VMAT) plans. A comparison was made between automatic and manual plans in terms of dosimetric indexes, monitor units (MUs) and planning time. Plan quality metric (PQM) encompassing these indexes was evaluated, with higher PQM values indicating superior plan quality. Qualitative evaluations of two plans were conducted by four reviewers., Results: The PQM score was 40.7 ± 13.1 for manual plans and 40.8 ± 13.5 for automatic plans (P = 0.75). Compared to manual plans, the targets coverage and homogeneity of automatic plans demonstrated no significant difference. Manual plans exhibited better sparing for lung in V5 (difference: 1.8 ± 4.2 %, P = 0.02), whereas automatic plans showed enhanced sparing for heart in V30 (difference: 1.4 ± 4.7 %, P = 0.02) and for spinal cord in Dmax (difference: 0.7 ± 4.7 Gy, P = 0.04). The planning time and MUs of automatic plans were significantly reduced by 70.5 ± 20.0 min and 97.4 ± 82.1. Automatic plans were deemed acceptable in 88 % of the reviews (176/200)., Conclusions: The DL-assisted approach for lung cancer notably decreased planning time and MUs, while demonstrating comparable or superior quality relative to manual plans. It has the potential to provide benefit to lung cancer patients., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Associazione Italiana di Fisica Medica e Sanitaria. Published by Elsevier Ltd. All rights reserved.)

Purpose: To establish a deep learning-based model to predict radiotherapy-induced temporal lobe injury (TLI)., Materials and Methods: Spatial features of dose distribution within the temporal lobe were extracted using both the three-dimensional convolution (C3D) network and the dosiomics method. The Minimal Redundancy-Maximal-Relevance (mRMR) method was employed to rank the extracted features and select the most relevant ones. Four machine learning (ML) classifiers, including logistic regression (LR), k-nearest neighbors (kNN), support vector machines (SVM) and random forest (RF), were used to establish prediction models. Nested sampling and hyperparameter tuning methods were applied to train and validate the prediction models. For comparison, a prediction model base on the conventional D 0.5cc of the temporal lobe obtained from dose volume (DV) histogram was established. The area under the receiver operating characteristic (ROC) curve (AUC) was utilized to compare the predictive performance of the different models., Results: A total of 127 nasopharyngeal carcinoma (NPC) patients were included in the study. In the model based on C3D deep learning features, the highest AUC value of 0.843 was achieved with 5 features. For the dosiomics features model, the highest AUC value of 0.715 was attained with 1 feature. Both of these models demonstrated superior performance compared to the prediction model based on DV parameters, which yielded an AUC of 0.695., Conclusion: The prediction model utilizing C3D deep learning features outperformed models based on dosiomics features or traditional parameters in predicting the onset of TLI. This approach holds promise for predicting radiation-induced toxicities and guide individualized radiotherapy., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Associazione Italiana di Fisica Medica e Sanitaria. Published by Elsevier Ltd. All rights reserved.)

Background: Studies on computed tomography (CT) synthesis based on magnetic resonance imaging (MRI) have mainly focused on pixel-wise consistency, but the texture features of regions of interest (ROIs) have not received appropriate attention., Purpose: This study aimed to propose a novel loss function to reproduce texture features of ROIs and pixel-wise consistency for deep learning-based MRI-to-CT synthesis. The method was expected to assist the multi-modality studies for radiomics., Methods: The study retrospectively enrolled 127 patients with nasopharyngeal carcinoma. CT and MRI images were collected for each patient, and then rigidly registered as pre-procession. We proposed a gray-level co-occurrence matrix (GLCM)-based loss function to improve the reproducibility of texture features. This novel loss function could be embedded into the present deep learning-based framework for image synthesis. In this study, a typical image synthesis model was selected as the baseline, which contained a Unet trained mean square error (MSE) loss function. We embedded the proposed loss function and designed experiments to supervise different ROIs to prove its effectiveness. The concordance correlation coefficient (CCC) of the GLCM feature was employed to evaluate the reproducibility of GLCM features, which are typical texture features. Besides, we used a publicly available dataset of brain tumors to verify our loss function., Results: Compared with the baseline, the proposed method improved the pixel-wise image quality metrics (MAE: 107.5 to 106.8 HU; SSIM: 0.9728 to 0.9730). CCC values of the GLCM features in GTVnx were significantly improved from 0.78 ± 0.12 to 0.82 ± 0.11 (p < 0.05 for paired t-test). Generally, > 90% (22/24) of the GLCM-based features were improved compared with the baseline, where the Informational Measure of Correlation feature was improved the most (CCC: 0.74 to 0.83). For the public dataset, the loss function also shows its effectiveness. With our proposed loss function added, the ability to reproduce texture features was improved in the ROIs., Conclusions: The proposed method reproduced texture features for MRI-to-CT synthesis, which would benefit radiomics studies based on image multi-modality synthesis., (© 2023 American Association of Physicists in Medicine.)

Background: The application of cone-beam computed tomography (CBCT) in image-guided radiotherapy and adaptive radiotherapy remains limited due to its poor image quality., Purpose: In this study, we aim to develop a deep learning framework to generate high-quality CBCT images for therapeutic applications., Methods: The synthetic CT (sCT) generation from the CBCT was proposed using a transformer-based network with a hybrid loss function. The network was trained and validated using the data from 176 patients to produce a general model that can be extensively applied to enhance CBCT images. After the first therapy, each patient can receive paired CBCT/planning CT (pCT) scans, and the obtained data were used to fine-tune the general model for further improvement. For subsequent treatment, a patient-specific, personalized model was made available. In total, 34 patients were examined for general model testing, and another six patients who underwent rescanned pCT scan were used for personalized model training and testing., Results: The general model decreased the mean absolute error (MAE) from 135 HU to 59 HU as compared to the CBCT. The hybrid loss function demonstrated superior performance in CT number correction and noise/artifacts reduction. The proposed transformer-based network also showed superior power in CT number correction compared to the classical convolutional neural network. The personalized model showed improvement based on the general model in some details, and the MAE was reduced from 59 HU (for the general model) to 57 HU (p < 0.05 Wilcoxon signed-rank test)., Conclusion: We established a deep learning framework based on transformer for clinical needs. The deep learning model demonstrated potential for continuous improvement with the help of a suggested personalized training strategy compatible with the clinical workflow., (© 2023 American Association of Physicists in Medicine.)

Purpose: This study was to improve image quality for high-speed MR imaging using a deep learning method for online adaptive radiotherapy in prostate cancer. We then evaluated its benefits on image registration., Methods: Sixty pairs of 1.5 T MR images acquired with an MR-linac were enrolled. The data included low-speed, high-quality (LSHQ), and high-speed low-quality (HSLQ) MR images. We proposed a CycleGAN, which is based on the data augmentation technique, to learn the mapping between the HSLQ and LSHQ images and then generate synthetic LSHQ (synLSHQ) images from the HSLQ images. Five-fold cross-validation was employed to test the CycleGAN model. The normalized mean absolute error (nMAE), peak signal-to-noise ratio (PSNR), structural similarity index measurement (SSIM), and edge keeping index (EKI) were calculated to determine image quality. The Jacobian determinant value (JDV), Dice similarity coefficient (DSC), and mean distance to agreement (MDA) were used to analyze deformable registration., Results: Compared with the LSHQ, the proposed synLSHQ achieved comparable image quality and reduced imaging time by ~ 66%. Compared with the HSLQ, the synLSHQ had better image quality with improvement of 57%, 3.4%, 26.9%, and 3.6% for nMAE, SSIM, PSNR, and EKI, respectively. Furthermore, the synLSHQ enhanced registration accuracy with a superior mean JDV (6%) and preferable DSC and MDA values compared with HSLQ., Conclusion: The proposed method can generate high-quality images from high-speed scanning sequences. As a result, it shows potential to shorten the scan time while ensuring the accuracy of radiotherapy., (© 2023. The Author(s).)

Background: Recently, patient rotating devices for gantry-free radiotherapy, a new approach to implement external beam radiotherapy, have been introduced. When a patient is rotated in the horizontal position, gravity causes anatomic deformation. For treatment planning, one feasible method is to acquire simulation images at different horizontal rotation angles., Purpose: This study aimed to investigate the feasibility of synthesizing magnetic resonance (MR) images at patient rotation angles of 180° (prone position) and 90° (lateral position) from those at a rotation angle of 0° (supine position) using deep learning., Methods: This study included 23 healthy male volunteers. They underwent MR imaging (MRI) in the supine position and then in the prone (23 volunteers) and lateral (16 volunteers) positions. T1-weighted fast spin echo was performed for all positions with the same parameters. Two two-dimensional deep learning networks, pix2pix generative adversarial network (pix2pix GAN) and CycleGAN, were developed for synthesizing MR images in the prone and lateral positions from those in the supine position, respectively. For the evaluation of the models, leave-one-out cross-validation was performed. The mean absolute error (MAE), Dice similarity coefficient (DSC), and Hausdorff distance (HD) were used to determine the agreement between the prediction and ground truth for the entire body and four specific organs., Results: For pix2pix GAN, the synthesized images were visually bad, and no quantitative evaluation was performed. The quantitative evaluation metrics of the body outlines calculated for the synthesized prone and lateral images using CycleGAN were as follows: MAE, 35.63 ± 3.98 and 40.45 ± 5.83, respectively; DSC, 0.97 ± 0.01 and 0.94 ± 0.01, respectively; and HD (in pixels), 16.74 ± 3.55 and 31.69 ± 12.03, respectively. The quantitative metrics of the bladder and prostate performed were also promising for both the prone and lateral images, with mean values >0.90 in DSC (p > 0.05). The mean DSC and HD values of the bilateral femur for the prone images were 0.96 and 3.63 (in pixels), respectively, and 0.78 and 12.65 (in pixels) for the lateral images, respectively (p < 0.05)., Conclusions: The CycleGAN could synthesize the MRI at lateral and prone positions using images at supine position, and it could benefit gantry-free radiation therapy., (© 2022 American Association of Physicists in Medicine.)

Background Purpose: This study focused on developing a fast Monte Carlo (MC) plan verification platform via a deep learning (DL)-based denoising approach. It can maintain the MC dose calculation accuracy while significantly reducing the computation time. We also investigated its potential applications for online adaptive proton therapy (APT)., Methods: First, we modeled an MC platform for proton therapy using the beam data library (BDL) required for treatment planning systems and then tested it with measured data. To accelerate the dose calculation, a dl-based denoising model with deep ResNet-deconvolution networks was developed. It was trained on the MC dose distribution of tumor sites obtained from 52 patients. The input MC dose distribution was with 1 × 10 6 simulated protons and the reference was 1 × 10 8 . Fivefold cross-validation was performed., Results: Comparing the MC model with measured data, the range agreement (point-to-point difference) was better than 0.85 mm, and the lateral dose profile difference was below 2.41 %. For the denoising approach, we found a significant improvement in the dose volume histogram (DVH) for predicted images compared with input images. The root mean squared error (RMSE) for predicted versus reference images was 3.94 times lower than that of the input versus reference images. Moreover, for the gamma passing rate (3 mm/3%), the predicted versus reference images have an average of 99 %, much higher than the 82 % of the input versus reference images. The MC model successfully denoised the test dose map (high noise) to approach the reference (low noise). The elapsed time can be reduced to < 60 s (simulation time [high noise] + predicted time), much lower than the simulation time of a low noise dose map (e.g., >100 min of simulation of 1 + E8 particles)., Conclusions: We propose an analogous end-to-end fast plan verification platform using the combination of MC and DL methods. The platform yields dose calculation accuracy similar to MC codes while significantly reducing the elapsed time and can be used for online APT as an alternative to online plan verification., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2022 Associazione Italiana di Fisica Medica e Sanitaria. Published by Elsevier Ltd. All rights reserved.)

Background and Purpose: Hybrid iterative reconstruction (HIR) is the most commonly used algorithm for four-dimensional computed tomography (4DCT) reconstruction due to its high speed. However, the image quality is worse than that of model-based iterative reconstruction (MIR). Different reconstruction methods affect the stability of radiomics features. Herein, we developed a deep learning method to improve the quality and radiomics reproducibility of the high-speed reconstruction., Materials and Methods: The 4DCT images of 70 patients were reconstructed using both the HIR and MIR algorithms. A cycle-consistent adversarial network was adopted to learn the mapping from HIR to MIR, and then generate synthetic MIR (sMIR) images from HIR. The performance was evaluated using the testing set (10 patients)., Results: The total reconstruction times for the HIR, MIR, and proposed sMIR images were approximately 2.5, 15, and 3.1 mins, respectively. The quality of sMIR images was close to that of MIR and was superior to that of HIR images, with noise reduced by 45-77% and contrast-to-noise ratio improved by 91-296%. The concordance correlation coefficients (CCC) of radiomic features improved from 0.89 ± 0.15 for HIR to 0.97 ± 0.07 for the proposed sMIR. The percentage of reproducible features (CCC ≥ 0.85) increased from 76.08% for HIR to 95.86% for sMIR, with an improvement of 19.78%., Conclusion: Compared to existing HIR algorithm, the proposed method improves the image quality and radiomics reproducibility of 4DCT images under high-speed reconstruction. It is computationally efficient and has potential to be integrated into any CT system., (Copyright © 2022 Elsevier B.V. All rights reserved.)

Purpose: We have developed a feasible method to evaluate deformable image registration using deep learning (DL)-based segmentation., Methods: Eighty patients with nasopharyngeal carcinoma were enrolled in this study. Two sets of fixed and moving computed tomography images acquired from each patient were input into the DL segmentation model to generate nine anatomic regions of interest (ROIs) separately and automatically. The ROIs generated in moving images were transferred to the fixed images using the registration transformation metric. The registration evaluation indexes, including the Dice similarity coefficient, derived from 60 well-registrated cases were then used to develop criteria for decision making. A double-blind study was performed to test the proposed method on quality assurance (QA) for image registration on a new test data set of 20 cases., Results: The values of evaluation indexes generated by our automated evaluation method were quite consistent with those from the manual method; however, the proposed method could save about 116 min per patient on average. The QA method achieved promising image registration error detection, with the following metrics for the nine ROIs: balanced accuracy, 0.946 ± 0.029; sensitivity, 0.959 ± 0.021; and specificity, 0.933 ± 0.050., Conclusions: The proposed method could potentially evaluate the deformable registration accuracy of specific areas. The preliminary NPC result shows that it has consistent performance with the conventional evaluation method with higher efficiency., (Copyright © 2022 Associazione Italiana di Fisica Medica e Sanitaria. Published by Elsevier Ltd. All rights reserved.)

Objective: Megavoltage computed tomography (MV-CT) is used for setup verification and adaptive radiotherapy in tomotherapy. However, its low contrast and high noise lead to poor image quality. This study aimed to develop a deep-learning-based method to generate synthetic kilovoltage CT (skV-CT) and then evaluate its ability to improve image quality and tumor segmentation. Approach: The planning kV-CT and MV-CT images of 270 patients with nasopharyngeal carcinoma (NPC) treated on an Accuray TomoHD system were used. An improved cycle-consistent adversarial network which used residual blocks as its generator was adopted to learn the mapping between MV-CT and kV-CT and then generate skV-CT from MV-CT. A Catphan 700 phantom and 30 patients with NPC were used to evaluate image quality. The quantitative indices included contrast-to-noise ratio (CNR), uniformity and signal-to-noise ratio (SNR) for the phantom and the structural similarity index measure (SSIM), mean absolute error (MAE), and peak signal-to-noise ratio (PSNR) for patients. Next, we trained three models for segmentation of the clinical target volume (CTV): MV-CT, skV-CT, and MV-CT combined with skV-CT. The segmentation accuracy was compared with indices of the dice similarity coefficient (DSC) and mean distance agreement (MDA). Main results: Compared with MV-CT, skV-CT showed significant improvement in CNR (184.0%), image uniformity (34.7%), and SNR (199.0%) in the phantom study and improved SSIM (1.7%), MAE (24.7%), and PSNR (7.5%) in the patient study. For CTV segmentation with only MV-CT, only skV-CT, and MV-CT combined with skV-CT, the DSCs were 0.75 ± 0.04, 0.78 ± 0.04, and 0.79 ± 0.03, respectively, and the MDAs (in mm) were 3.69 ± 0.81, 3.14 ± 0.80, and 2.90 ± 0.62, respectively. Significance: The proposed method improved the image quality of MV-CT and thus tumor segmentation in helical tomotherapy. The method potentially can benefit adaptive radiotherapy., (Creative Commons Attribution license.)

Purpose: To develop a deep learning method to predict patient-specific dose volume histograms (DVHs) for radiotherapy planning., Methods: Patient data included 180 cases with nasopharyngeal cancer, of which 153 cases were used for training and 27 for testing. A network (named "DVHnet") based on a convolutional neural network (CNN) was designed for directly predicting DVHs of organs at risk (OARs). Two-channel images with contoured structures were generated as the inputs for training the model. A one-dimensional array consisting of 256 continuous volume percentages on a DVH curve for each slice was calculated as the corresponding output. The combined DVH was then calculated. Sixteen OARs were modeled in the study. Prediction accuracy was evaluated against the corresponding DVH curve of ground truth (GT) plans. A global DVH analysis and critical dosimetry metrics for each OAR were calculated for quantitative evaluation. The performance of DVHnet also was evaluated against two baselines: DosemapNet (developed by our research group) and commercial RapidPlan software., Results: The predicted mean difference in average dose of all OARs using DVHnet was 0.30 ± 0.95 Gy. And the predicted differences in D2% and D50 can be control within 2.32 and 0.69 Gy. For most OARs, there were no obvious differences between the dosimetric metrics of the predicted and GT values for both DVHnet and DosemapNet (P ≥ 0.05). Only the predicted D2% of the optic organs for DVHnet, and of brain stem PRV for DosemapNet displayed statistically significant differences. Except for the optic organs, DVHnet performs better than or comparably with RapidPlan. The mean difference in proportion of points of interest was 3.59% ± 7.78%., Conclusions: A deep learning network model was developed to automatically extract useful features for accurate prediction of patient-specific DVH curves directly. The performance of DVHnet was comparable to DosemapNet and RapidPlan., (© 2021 American Association of Physicists in Medicine.)

Purpose: To develop a deep learning-based model to predict achievable dose-volume histograms (DVHs) of organs at risk (OARs) for automation of inverse planning., Methods: The model was based on a connected residual deconvolution network (CResDevNet) and compared with UNet as a baseline. The DVHs of OARs are dependent on patient anatomical features of the planning target volumes and OARs, and these spatial relationships can be learned automatically from prior high-quality plans. The contours of planning target volumes and OARs were parsed from the plan database and used as the input to the model, and the dose-area histograms (DAHs) of the OARs were output from the model. The model was trained from scratch by correlating anatomical features with DAHs of OARs, then accumulating these histograms to obtain the final predicted DVH for each OAR. Helical tomotherapy plans for 170 nasopharyngeal cancer patients were used to train and validate the model. An additional 60 patient treatment plans were used to test the predictive accuracy of the model. The DVHs and dose-volume indices (DVIs) of clinical interest for each OAR in the testing dataset were predicted to evaluate the accuracy of the models. The mean absolute errors in the DVIs for each OAR were calculated using each model and statistically compared using a paired-samples t-test. Dice similarity coefficients for areas of the DVHs were also evaluated., Results: Dose-volume histograms of 21 OARs in nasopharyngeal cancer were predicted using the models. For each patient, 63 DVIs for all OARs were calculated. Using the 60 patient treatment plans in the testing dataset, 79% and 73% of the DVIs predicted using the CResDevNet and UNet models, respectively, were within 5% of the clinical values. The median value of the DVIs' mean absolute errors was 3.2 ± 2.5% and 3.7 ± 2.9% for the CResDevNet and UNet models, respectively. The average dice similarity coefficient for all OARs was 0.965 using the CResDevNet model and 0.958 using the UNet model., Conclusions: A deep learning model was developed for predicting achievable DVHs of OARs. The prediction accuracy of the CResDevNet model was evaluated using a planning database of nasopharyngeal cancer cases and shown to be more accurate than the UNet model. Prediction accuracy was also higher for larger-volume OARs. The model can be used for automation of inverse planning and quality assessment of individual treatment plans., (© 2020 American Association of Physicists in Medicine.)

Purpose: The purpose of this study is to develop a deep learning (DL) method for producing four-dimensional computed tomography (4DCT) ventilation imaging and to evaluate the accuracy of the DL-based ventilation imaging against single-photon emission-computed tomography (SPECT) ventilation imaging (SPECT-VI). The performance of the DL-based method is assessed by comparing with density change- and Jacobian-based (HU and JAC) methods., Materials and Methods: Fifty patients with esophagus or lung cancer who underwent thoracic radiotherapy were enrolled in this study. For each patient, 4DCT scans paired with 99mTc-Technegas SPECT/CT were acquired before the first radiotherapy treatment. 4DCT and SPECT/CT were first rigidly registered using MIMvista and converted to data matrix using MATLAB, and then transferred to a DL model based on U-net for correlating 4DCT features and SPECT-VI. Two forms of 4DCT dataset [(a) ten phases and (b) two phases of peak-exhalation and peak-inhalation] as input are studied. Tenfold cross-validation procedure was used to evaluate the performance of the DL model. For comparative evaluation, HU and JAC methodologies are used to calculate specific ventilation imaging based on 4DCT (CTVI) for each patient. The voxel-wise Spearman's correlation was evaluated over the whole lung between each of CTVI and corresponding SPECT-VI. The SPECT-VI and produced CTVIs were segmented into high, median, and low functional lung (HFL, MFL, and LFL) regions. The spatial overlap of corresponding HFL, MFL, and LFL for each CTVI against SPECT-VI was also evaluated using the dice similarity coefficient (DSC). The averaged DSC of functional lung regions was calculated and statistically analyzed with a one-factor ANONA model among different methods., Results: The voxel-wise Spearman r s values were (0.22 ± 0.31), (-0.09 ± 0.18), and (0.73 ± 0.16)/(0.71 ± 0.17) for the CTVI HU , CTVI JAC , and CTVI DL(1) /CTVI DL(2) . These results showed the DL method yielded the strongest correlation with SPECT-VI. Using the DSC as the spatial overlap metric, we found that the CTVI HU , CTVI JAC , and CTVI DL(1) /CTVI DL(2) methods achieved averaged DSC values for all patients to be (0.45 ± 0.08), (0.33 ± 0.04), and (0.73 ± 0.09)/(0.71 ± 0.09), respectively. The results demonstrated that the DL method yielded the highest similarity with SPECT-VI with the prominently significant difference (P < 10 -7 )., Conclusions: This study developed a DL method for producing CTVI and performed a validation against SPECT-VI. The results demonstrated that DL method can derive CTVI with greatly improved accuracy in comparison to HU and JAC methods. The produced ventilation images can be more accurate and useful for lung functional avoidance radiotherapy and treatment response modeling., (© 2019 American Association of Physicists in Medicine.)

Purpose: To develop a deep learning method for prediction of three-dimensional (3D) voxel-by-voxel dose distributions of helical tomotherapy (HT)., Methods: Using previously treated HT plans as training data, a deep learning model named U-ResNet-D was trained to predict a 3D dose distribution. First, the contoured structures and dose volumes were converted from plan database to 3D matrix with a program based on a developed visualization toolkit (VTK), then transferred to U-ResNet-D for correlating anatomical features and dose distributions at voxel-level. One hundred and ninety nasopharyngeal cancer (NPC) patients treated by HT with multiple planning target volumes (PTVs) in different prescription patterns were studied. The model was typically trained from scratch with weights randomly initialized rather than using transfer-learning method, and used to predict new patient's 3D dose distributions. The predictive accuracy was evaluated with three methods: (a) The dose difference at the position r, δ(r, r) = D c (r) - D p (r), was calculated for each voxel. The mean (μ δ(r,r) ) and standard deviation (σ δ(r,r) ) of δ(r, r) were calculated to assess the prediction bias and precision; (b) The mean absolute differences of dosimetric indexes (DIs) including maximum and mean dose, homogeneity index, conformity index, and dose spillage for PTVs and organ at risks (OARs) were calculated and statistically analyzed with the paired-samples t test; (c) Dice similarity coefficients (DSC) between predicted and clinical isodose volumes were calculated., Results: The U-ResNet-D model predicted 3D dose distribution accurately. For twenty tested patients, the prediction bias ranged from -2.0% to 2.3% and prediction error varied from 1.5% to 4.5% (relative to prescription) for 3D dose differences. The mean absolute dose differences for PTVs and OARs are within 2.0% and 4.2%, and nearly all the DIs for PTVs and OARs had no significant differences. The averaged DSC ranged from 0.95 to 1 for different isodose volumes., Conclusions: The study developed a new deep learning method for 3D voxel-by-voxel dose prediction, and shown to be able to produce accurately dose predictions for nasopharyngeal patients treated by HT. The predicted 3D dose map can be useful for improving radiotherapy planning design, ensuring plan quality and consistency, making clinical technique comparison, and guiding automatic treatment planning., (© 2019 American Association of Physicists in Medicine.)

Purpose: To develop a method for predicting optimal dose distributions, given the planning image and segmented anatomy, by applying deep learning techniques to a database of previously optimized and approved Intensity-modulated radiation therapy treatment plans., Methods: Eighty cases of early-stage nasopharyngeal cancer (NPC) were included in the study. Seventy cases were chosen randomly as the training set and the remaining as the test set. The inputs were the images with structures, with each target and organs at risk (OARs) assigned a unique label. The outputs were dose maps, including coarse dose maps and converted fine dose maps (FDM) from convolution. Two types of input images with structures were used in the model building. One type of input included the images (with associated structures) without manipulation. The second type of input involved modifying the image gray label with information from radiation beam geometry. ResNet101 was chosen as the deep learning network for both. The accuracy of predicted dose distributions was evaluated against the corresponding dose as used in the clinic. A global three-dimensional gamma analysis was calculated for the evaluation., Results: The proposed model trained with the two different sets of input images and structures could both predict patient-specific dose distributions accurately. For the out-of-field dose distributions, the model obtained from the input with radiation geometry performed better (dose difference in %, 4.7 ± 6.1% vs 5.5 ± 7.9%, P < 0.05). The mean Gamma pass rates of dose distributions predicted with both types of input were comparable for most OARs (P > 0.05), except for the bilateral optic nerves and the optic chiasm., Conclusions: The proposed system with radiation geometry added to the input is a promising method to generate patient-specific dose distributions for radiotherapy. It can be applied to obtain the dose distributions slice-by-slice for planning quality assurance and for guiding automated planning., (© 2018 The Authors Medical Physics published by Wiley Periodicals, Inc. on behalf of American Association of Physicists in Medicine.)

Purpose: To develop a non‐coplanar cone‐beam computed tomography (CBCT) image reconstruction method using projections within a limited angle range for non‐coplanar radiotherapy. Methods: A generative adversarial network (GAN) was utilized to reconstruct non‐coplanar CBCT images. Data from 40 patients with brain tumors and two head phantoms were used in this study. In the training stage, the generator of the GAN used coplanar CBCT and non‐coplanar projections as the input, and an encoder with a dual‐branch structure was utilized to extract features from the coplanar CBCT and non‐coplanar projections separately. Non‐coplanar CBCT images were then reconstructed using a decoder by combining the extracted features. To improve the reconstruction accuracy of the image details, the generator was adversarially trained using a patch‐based convolutional neural network as the discriminator. A newly designed joint loss was used to improve the global structure consistency rather than the conventional GAN loss. The proposed model was evaluated using data from eight patients and two phantoms at four couch angles (±45°, ±90°) that are most commonly used for brain non‐coplanar radiotherapy in our department. The reconstructed accuracy was evaluated by calculating the root mean square error (RMSE) and an overall registration error ε, computed by integrating the rigid transformation parameters. Results: In both patient data and phantom data studies, the qualitative and quantitative metrics results indicated that ± 45° couch angle models performed better than ±90° couch angle models and had statistical differences. In the patient data study, the mean RMSE and ε values of couch angle at 45°, −45°, 90°, and −90° were 58.5 HU and 0.42 mm, 56.8 HU and 0.41 mm, 73.6 HU and 0.48 mm, and 65.3 HU and 0.46 mm, respectively. In the phantom data study, the mean RMSE and ε values of couch angle at 45°, −45°, 90°, and −90° were 91.2 HU and 0.46 mm, 95.0 HU and 0.45 mm, 114.6 HU and 0.58 mm, and 102.9 HU and 0.52 mm, respectively. Conclusions: The results show that the reconstructed non‐coplanar CBCT images can potentially enable intra‐treatment three‐dimensional position verification for non‐coplanar radiotherapy. [ABSTRACT FROM AUTHOR]

Background: It is necessary to contour regions of interest (ROIs) for online magnetic resonance imaging (MRI)‐guided adaptive radiotherapy (MRIgART). These updated contours are used for online replanning to obtain maximum dosimetric benefits. Contouring can be accomplished using deformable image registration (DIR) and deep learning (DL)‐based autosegmentation methods. However, these methods may require considerable manual editing and thus prolong treatment time. Purpose: The present study aimed to improve autosegmentation performance by integrating patients' pretreatment information in a DL‐based segmentation algorithm. It is expected to improve the efficiency of current MRIgART process. Methods: Forty patients with prostate cancer were enrolled retrospectively. The online adaptive MR images, patient‐specific planning computed tomography (CT), and contours in CT were used for segmentation. The deformable registration of planning CT and MR images was performed first to obtain a deformable CT and corresponding contours. A novel DL network, which can integrate such patient‐specific information (deformable CT and corresponding contours) into the segmentation task of MR images was designed. We performed a four‐fold cross‐validation for the DL models. The proposed method was compared with DIR and DL methods on segmentation of prostate cancer. The ROIs included the clinical target volume (CTV), bladder, rectum, left femur head, and right femur head. Dosimetric parameters of automatically generated ROIs were evaluated using a clinical treatment planning system. Results: The proposed method enhanced the segmentation accuracy of conventional procedures. Its mean value of the dice similarity coefficient (93.5%) over the five ROIs was higher than both DIR (87.5%) and DL (87.2%). The number of patients (n = 40) that required major editing using DIR, DL, and our method were 12, 18, and 7 (CTV); 17, 4, and 1 (bladder); 8, 11, and 5 (rectum); 2, 4, and 1 (left femur head); and 3, 7, and 1 (right femur head), respectively. The Spearman rank correlation coefficient of dosimetry parameters between the proposed method and ground truth was 0.972 ± 0.040, higher than that of DIR (0.897 ± 0.098) and DL (0.871 ± 0.134). Conclusion: This study proposed a novel method that integrates patient‐specific pretreatment information into DL‐based segmentation algorithm. It outperformed baseline methods, thereby improving the efficiency and segmentation accuracy in adaptive radiotherapy. [ABSTRACT FROM AUTHOR]

Purpose: Respiration has a major impact on the accuracy of radiation treatment for thorax and abdominal tumours. Instantaneous volumetric imaging could provide precise knowledge of tumour and normal organs' three‐dimensional (3D) movement, which is the key to reducing the negative effect of breathing motion. Therefore, this study proposed a real‐time 3D MRI reconstruction method from cine‐MRI using an unsupervised network. Methods and materials: Cine‐MRI and setup 3D‐MRI from eight patients with liver cancer were utilized to establish and validate the deep learning network for 3D‐MRI reconstruction. Unlike previous methods that required 4D‐MRI for network training, the proposed method utilized a reference 3D‐MRI and cine‐MRI to generate the training data. Then, a network was trained in an unsupervised manner to estimate the relationship between the cine‐MRI acquired on coronal plane and deformation vector field (DVF) that describes the patient's breathing motion. After the training process, the coronal cine‐MRI were inputted into the network, and the corresponding DVF was obtained. By wrapping the reference 3D‐MRI with the generated DVF, the 3D‐MRI could be reconstructed. Results: The reconstructed 3D‐MRI slices were compared with the corresponding phase‐sorted cine‐MRI using dice similarity coefficients (DSCs) of liver contours and blood vessel localization error. In all patients, the liver DSC had mean value >96.1% and standard deviation < 1.3%; the blood vessel localization error had mean value <2.6 mm, and standard deviation was <2.0 mm. Moreover, the time for 3D‐MRI reconstruction was approximately 100 ms. These results indicated that the proposed method could accurately reconstruct the 3D‐MRI in real time. Conclusions: The proposed method could accurately reconstruct the 3D‐MRI from cine‐MRI in real time. This method has great potential in improving the accuracy of radiotherapy for moving tumours. [ABSTRACT FROM AUTHOR]

Purpose: The current algorithms for measuring ventilation images from 4D cone-beam computed tomography (CBCT) are affected by the accuracy of deformable image registration (DIR). This study proposes a new deep learning (DL) method that does not rely on DIR to derive ventilation images from 4D-CBCT (CBCT-VI), which was validated with the gold-standard single-photon emission-computed tomography ventilation image (SPECT-VI). Materials and Methods: This study consists of 4D-CBCT and 99mTc-Technegas SPECT/CT scans of 28 esophagus or lung cancer patients. The scans were rigidly registered for each patient. Using these data, CBCT-VI was derived using a deep learning-based model. Two types of model input data are studied, namely, (a) 10 phases of 4D-CBCT and (b) two phases of peak-exhalation and peak-inhalation of 4D-CBCT. A sevenfold cross-validation was applied to train and evaluate the model. The DIR-dependent methods (density-change-based and Jacobian-based methods) were used to measure the CBCT-VIs for comparison. The correlation was calculated between each CBCT-VI and SPECT-VI using voxel-wise Spearman's correlation. The ventilation images were divided into high, medium, and low functional lung regions. The similarity of different functional lung regions between SPECT-VI and each CBCT-VI was evaluated using the dice similarity coefficient (DSC). One-factor ANONA model was used for statistical analysis of the averaged DSC for the different methods of generating ventilation images. Results: The correlation values were 0.02 ± 0.10, 0.02 ± 0.09, and 0.65 ± 0.13/0.65 ± 0.15, and the averaged DSC values were 0.34 ± 0.04, 0.34 ± 0.03, and 0.59 ± 0.08/0.58 ± 0.09 for the density change, Jacobian, and deep learning methods, respectively. The strongest correlation and the highest similarity with SPECT-VI were observed for the deep learning method compared to the density change and Jacobian methods. Conclusion: The results showed that the deep learning method improved the accuracy of correlation and similarity significantly, and the derived CBCT-VIs have the potential to monitor the lung function dynamic changes during radiotherapy. [ABSTRACT FROM AUTHOR]

Background: Radical radiotherapy is the main treatment modality for early and locally advanced nasopharyngeal carcinoma (NPC). Magnetic resonance imaging (MRI) has the advantages of no ionizing radiation and high soft-tissue resolution compared to computed tomography (CT), but it does not provide electron density (ED) information for radiotherapy planning. Therefore, in this study, we developed a pseudo-CT (pCT) generation method to provide necessary ED information for MRI-only planning in NPC radiotherapy. Methods: Twenty patients with early-stage NPC who received radiotherapy in our hospital were investigated. First, 1433 sets of paired T1 weighted magnetic resonance (MR) simulation images and CT simulation images were rigidly registered and preprocessed. A 16-layer U-Net was used to train the pCT generative model and a "pix2pix" generative adversarial network (GAN) was also trained to compare with the pure U-Net regrading pCT quality. Second, the contours of all target volumes and organs at risk in the original CT were transferred to the pCT for planning, and the beams were copied back to the original CT for reference dose calculation. Finally, the dose distribution calculated on the pCT was compared with the reference dose distribution through gamma analysis and dose-volume indices. Results: The average time for pCT generation for each patient was 7.90 ± 0.47 seconds. The average mean (absolute) error was −9.3 ± 16.9 HU (102.6 ± 11.4 HU), and the mean-root-square error was 209.8 ± 22.6 HU. There was no significant difference between the pCT quality of pix2pix GAN and that of pure U-Net (p > 0.05). The dose distribution on the pCT was highly consistent with that on the original CT. The mean gamma pass rate (2 mm/3%, 10% low dose threshold) was 99.1% ± 0.3%, and the mean absolute difference of nasopharyngeal PGTV D99% and PTV V95% were 0.4% ± 0.2% and 0.1% ± 0.1%. Conclusion: The proposed deep learning model can accurately predict CT from MRI, and the generated pCT can be employed in precise dose calculations. It is of great significance to realize MRI-only planning in NPC radiotherapy, which can improve structure delineation and considerably reduce additional imaging dose, especially when an MR-guided linear accelerator is adopted for treatment. [ABSTRACT FROM AUTHOR]

Background and Purpose: Automatic segmentation model is proven to be efficient in delineation of organs at risk (OARs) in radiotherapy; its performance is usually evaluated with geometric differences between automatic and manual delineations. However, dosimetric differences attract more interests than geometric differences in the clinic. Therefore, this study aimed to evaluate the performance of automatic segmentation with dosimetric metrics for volumetric modulated arc therapy of esophageal cancer patients. Methods: Nineteen esophageal cancer cases were included in this study. Clinicians manually delineated the target volumes and the OARs for each case. Another set of OARs was automatically generated using convolutional neural network models. The radiotherapy plans were optimized with the manually delineated targets and the automatically delineated OARs separately. Segmentation accuracy was evaluated by Dice similarity coefficient (DSC) and mean distance to agreement (MDA). Dosimetric metrics of manually and automatically delineated OARs were obtained and compared. The clinically acceptable dose difference and volume difference of OARs between manual and automatic delineations are supposed to be within 1 Gy and 1%, respectively. Results: Average DSC values were greater than 0.92 except for the spinal cord (0.82), and average MDA values were <0.90 mm except for the heart (1.74 mm). Eleven of the 20 dosimetric metrics of the OARs were not significant (P > 0.05). Although there were significant differences (P < 0.05) for the spinal cord (D2%), left lung (V10, V20, V30, and mean dose), and bilateral lung (V10, V20, V30, and mean dose), their absolute differences were small and acceptable for the clinic. The maximum dosimetric metrics differences of OARs between manual and automatic delineations were ΔD2% = 0.35 Gy for the spinal cord and ΔV30 = 0.4% for the bilateral lung, which were within the clinical criteria in this study. Conclusion: Dosimetric metrics were proposed to evaluate the automatic delineation in radiotherapy planning of esophageal cancer. Consequently, the automatic delineation could substitute the manual delineation for esophageal cancer radiotherapy planning based on the dosimetric evaluation in this study. [ABSTRACT FROM AUTHOR]

Purpose: More and more automatic segmentation tools are being introduced in routine clinical practice. However, physicians need to spend a considerable amount of time in examining the generated contours slice by slice. This greatly reduces the benefit of the tool's automaticity. In order to overcome this shortcoming, we developed an automatic quality assurance (QA) method for automatic segmentation using convolutional neural networks (CNNs). Materials and Methods: The study cohort comprised 680 patients with early-stage breast cancer who received whole breast radiation. The overall architecture of the automatic QA method for deep learning-based segmentation included the following two main parts: a segmentation CNN model and a QA network that was established based on ResNet-101. The inputs were from computed tomography, segmentation probability maps, and uncertainty maps. Two kinds of Dice similarity coefficient (DSC) outputs were tested. One predicted the DSC quality level of each slice ([0.95, 1] for "good," [0.8, 0.95] for "medium," and [0, 0.8] for "bad" quality), and the other predicted the DSC value of each slice directly. The performances of the method to predict the quality levels were evaluated with quantitative metrics: balanced accuracy, F score, and the area under the receiving operator characteristic curve (AUC). The mean absolute error (MAE) was used to evaluate the DSC value outputs. Results: The proposed methods involved two types of output, both of which achieved promising accuracy in terms of predicting the quality level. For the good, medium, and bad quality level prediction, the balanced accuracy was 0.97, 0.94, and 0.89, respectively; the F score was 0.98, 0.91, and 0.81, respectively; and the AUC was 0.96, 0.93, and 0.88, respectively. For the DSC value prediction, the MAE was 0.06 ± 0.19. The prediction time was approximately 2 s per patient. Conclusions: Our method could predict the segmentation quality automatically. It can provide useful information for physicians regarding further verification and revision of automatic contours. The integration of our method into current automatic segmentation pipelines can improve the efficiency of radiotherapy contouring. [ABSTRACT FROM AUTHOR]

Purpose: To investigate whether a deep learning-assisted contour (DLAC) could provide greater accuracy, inter-observer consistency, and efficiency compared with a manual contour (MC) of the clinical target volume (CTV) for non-small cell lung cancer (NSCLC) receiving postoperative radiotherapy (PORT). Materials and Methods: A deep dilated residual network was used to achieve the effective automatic contour of the CTV. Eleven junior physicians contoured CTVs on 19 patients by using both MC and DLAC methods independently. Compared with the ground truth, the accuracy of the contour was evaluated by using the Dice coefficient and mean distance to agreement (MDTA). The coefficient of variation (CV) and standard distance deviation (SDD) were rendered to measure the inter-observer variability or consistency. The time consumed for each of the two contouring methods was also compared. Results: A total of 418 CTV sets were generated. DLAC improved contour accuracy when compared with MC and was associated with a larger Dice coefficient (mean ± SD: 0.75 ± 0.06 vs. 0.72 ± 0.07, p < 0.001) and smaller MDTA (mean ± SD: 2.97 ± 0.91 mm vs. 3.07 ± 0.98 mm, p < 0.001). The DLAC was also associated with decreased inter-observer variability, with a smaller CV (mean ± SD: 0.129 ± 0.040 vs. 0.183 ± 0.043, p < 0.001) and SDD (mean ± SD: 0.47 ± 0.22 mm vs. 0.72 ± 0.41 mm, p < 0.001). In addition, a value of 35% of time saving was provided by the DLAC (median: 14.81 min vs. 9.59 min, p < 0.001). Conclusions: Compared with MC, the DLAC is a promising strategy to obtain superior accuracy, consistency, and efficiency for the PORT-CTV in NSCLC. [ABSTRACT FROM AUTHOR]

Purpose To train and evaluate a very deep dilated residual network (DD-ResNet) for fast and consistent auto-segmentation of the clinical target volume (CTV) for breast cancer (BC) radiotherapy with big data. Methods DD-ResNet was an end-to-end model enabling fast training and testing. We used big data comprising 800 patients who underwent breast-conserving therapy for evaluation. The CTV were validated by experienced radiation oncologists. We performed a fivefold cross-validation to test the performance of the model. The segmentation accuracy was quantified by the Dice similarity coefficient (DSC) and the Hausdorff distance (HD). The performance of the proposed model was evaluated against two different deep learning models: deep dilated convolutional neural network (DDCNN) and deep deconvolutional neural network (DDNN). Results Mean DSC values of DD-ResNet (0.91 and 0.91) were higher than the other two networks (DDCNN: 0.85 and 0.85; DDNN: 0.88 and 0.87) for both right-sided and left-sided BC. It also has smaller mean HD values of 10.5 mm and 10.7 mm compared with DDCNN (15.1 mm and 15.6 mm) and DDNN (13.5 mm and 14.1 mm). Mean segmentation time was 4 s, 21 s and 15 s per patient with DDCNN, DDNN and DD-ResNet, respectively. The DD-ResNet was also superior with regard to results in the literature. Conclusions The proposed method could segment the CTV accurately with acceptable time consumption. It was invariant to the body size and shape of patients and could improve the consistency of target delineation and streamline radiotherapy workflows. [ABSTRACT FROM AUTHOR]

Purpose Delineation of the clinical target volume ( CTV) and organs at risk ( OARs) is very important for radiotherapy but is time-consuming and prone to inter-observer variation. Here, we proposed a novel deep dilated convolutional neural network ( DDCNN)-based method for fast and consistent auto-segmentation of these structures. Methods Our DDCNN method was an end-to-end architecture enabling fast training and testing. Specifically, it employed a novel multiple-scale convolutional architecture to extract multiple-scale context features in the early layers, which contain the original information on fine texture and boundaries and which are very useful for accurate auto-segmentation. In addition, it enlarged the receptive fields of dilated convolutions at the end of networks to capture complementary context features. Then, it replaced the fully connected layers with fully convolutional layers to achieve pixel-wise segmentation. We used data from 278 patients with rectal cancer for evaluation. The CTV and OARs were delineated and validated by senior radiation oncologists in the planning computed tomography ( CT) images. A total of 218 patients chosen randomly were used for training, and the remaining 60 for validation. The Dice similarity coefficient ( DSC) was used to measure segmentation accuracy. Results Performance was evaluated on segmentation of the CTV and OARs. In addition, the performance of DDCNN was compared with that of U-Net. The proposed DDCNN method outperformed the U-Net for all segmentations, and the average DSC value of DDCNN was 3.8% higher than that of U-Net. Mean DSC values of DDCNN were 87.7% for the CTV, 93.4% for the bladder, 92.1% for the left femoral head, 92.3% for the right femoral head, 65.3% for the intestine, and 61.8% for the colon. The test time was 45 s per patient for segmentation of all the CTV, bladder, left and right femoral heads, colon, and intestine. We also assessed our approaches and results with those in the literature: our system showed superior performance and faster speed. Conclusions These data suggest that DDCNN can be used to segment the CTV and OARs accurately and efficiently. It was invariant to the body size, body shape, and age of the patients. DDCNN could improve the consistency of contouring and streamline radiotherapy workflows. [ABSTRACT FROM AUTHOR]

Background Purpose Methods Results Conclusions Respiratory motion during radiotherapy (RT) may reduce the therapeutic effect and increase the dose received by organs at risk. This can be addressed by real‐time tracking, where respiration motion prediction is currently required to compensate for system latency in RT systems. Notably, for the prediction of future images in image‐guided adaptive RT systems, the use of deep learning has been considered.This study proposed a modified generative adversarial network (GAN) for predicting cine‐MR images in real time.Sagittal cine magnetic resonance (cine‐MR) images of 15 patients with liver cancer who received RT were collected. The image series length of each patient was 300, and each series was divided into training, validation, and test sets. The datasets were further divided using a sliding window size of 10 and a stride of 1. A pix2pix GAN with the generator replaced by convolutional long short‐term memory (ConvLSTM) was proposed herein. A five‐frame cine‐MR image series was inputted into the network, which predicted the next five frames. The proposed network was compared with three advanced networks: ConvLSTM, Eidetic 3D LSTM (E3D–LSTM), and SwinLSTM. Personalized models were trained for each patient. The peak signal‐to‐noise ratio (PSNR), structural similarity index measure (SSIM), visual information fidelity (VIF), Pearson correlation coefficient (Pearson corr), and respiratory motion accuracy of the predicted images were used to evaluate the methods.The proposed network demonstrated optimal performance in the four networks across various indicators. The proposed method provided better SSIM values than ConvLSTM at time steps 1, 2, 3, and 4, and outperformed E3DLSTM at all time steps. In terms of the VIF, the proposed method outperformed E3D–LSTM at all time steps and SwinLSTM at time steps 2, 3, 4, and 5. The proposed method was not significantly different from other methods in terms of Pearson correlation values except that it outperformed E3DLSTM at time step 1. In terms of the Pearson corr, the proposed method consistently achieves better values, especially in the high‐frequency components. Low average landmark tracking errors were provided by the proposed method at time steps 4 and 5 (2.42 ± 0.91 and 2.44 ± 0.96 mm, respectively).The GAN–ConvLSTM network can generate high‐acutance real‐time cine‐MR images and predict respiratory motion with better accuracy. [ABSTRACT FROM AUTHOR]

Background: The dynamic tracking of tumors with radiation beams in radiation therapy requires the prediction of real-time target locations prior to beam delivery, as treatment involving radiation beams and gating tracking results in time latency.Objective: In this study, a deep learning model that was based on a temporal convolutional neural network was developed to predict internal target locations by using multiple external markers.Methods: Respiratory signals from 69 treatment fractions of 21 patients with cancer who were treated with the CyberKnife Synchrony device (Accuray Incorporated) were used to train and test the model. The reported model's performance was evaluated by comparing the model to a long short-term memory model in terms of the root mean square errors (RMSEs) of real and predicted respiratory signals. The effect of the number of external markers was also investigated.Results: The average RMSEs of predicted (ahead time=400 ms) respiratory motion in the superior-inferior, anterior-posterior, and left-right directions and in 3D space were 0.49 mm, 0.28 mm, 0.25 mm, and 0.67 mm, respectively.Conclusions: The experiment results demonstrated that the temporal convolutional neural network-based respiratory prediction model could predict respiratory signals with submillimeter accuracy. [ABSTRACT FROM AUTHOR]

• A patient-specific auto-planning method for MRIgART is validated. • The method improves the quality and efficiency of adaptive treatment planning. • It can be integrated into the current workflow for MRIgART. Fast and automated generation of treatment plans is desirable for magnetic resonance imaging (MRI)-guided adaptive radiotherapy (MRIgART). This study proposed a novel patient-specific auto-planning method and validated its feasibility in improving the existing online planning workflow. Data from 40 patients with prostate cancer were collected retrospectively. A patient-specific auto-planning method was proposed to generate adaptive treatment plans. First, a population dose-prediction model (M 0) was trained using data from previous patients. Second, a patient-specific model (M ps) was created for each new patient by fine-tuning M 0 with the patient's data. Finally, an auto plan was optimized using the parameters derived from the predicted dose distribution by M ps. The auto plans were compared with manual plans in terms of plan quality, efficiency, dosimetric verification, and clinical evaluation. The auto plans improved target coverage, reduced irradiation to the rectum, and provided comparable protection to other organs-at-risk. Target coverage for the planning target volume (+0.61 %, P = 0.023) and clinical target volume 4000 (+1.60 %, P < 0.001) increased. V 2900cGy (−1.06 %, P = 0.004) and V 1810cGy (−2.49 %, P < 0.001) to the rectal wall and V 1810cGy (−2.82 %, P = 0.012) to the rectum were significantly reduced. The auto plans required less planning time (−3.92 min, P = 0.001), monitor units (−46.48, P = 0.003), and delivery time (−0.26 min, P = 0.004), and their gamma pass rates (3 %/2 mm) were higher (+0.47 %, P = 0.014). The proposed patient-specific auto-planning method demonstrated a robust level of automation and was able to generate high-quality treatment plans in less time for MRIgART in prostate cancer. [ABSTRACT FROM AUTHOR]