Search

Background Purpose Methods Results Conclusion Respiratory motion is a challenge for accurate radiotherapy that may be mitigated by real‐time tracking. Commercial tracking systems utilize a hybrid external‐internal correlation model (ECM), integrating continuous external breathing monitoring with sparse X‐ray imaging of the internal tumor position.This study investigates the feasibility of using the next generation reservoir computing (NG‐RC) model as a hybrid ECM to transform measured external motions into estimated 3D internal motions.The NG‐RC model utilizes the nonlinear vector autoregressive (NVAR) machine to account for the hysteresis or phase differences between external and internal motions. The datasets used to evaluate the efficacy of the NG‐RC model include 57 motion traces from the CyberKnife system. The datasets were divided into three regions (central, lower, and upper livers) and three motion patterns. These patterns include linear and nonlinear motion patterns (Group A), hysteresis motion patterns (Group B), and all motion patterns (Group C). Moreover, various updating techniques were examined, such as continuously updating the NG‐RC model using the first‐in‐first‐out (FIFO) approach and sampling the internal tumor position every 0 s (strategy A), 60 s (strategy B), 30 s (strategy C), and 50 s (strategy D).The NG‐RC model combined with strategy C resulted in better estimation accuracy than the reported CyberKnife cases (Wilcoxon signed rank p < 0.05). For linear and nonlinear motion patterns, the 3D radial estimation accuracy (mean ± SD) using the NG‐RC model combined with strategy C and the CyberKnife system was 1.20 ± 0.78 and 1.1 ± 0.20 mm in the central liver, 0.66 ± 0.25 and 1.49 ± 0.50 mm in the lower liver, and 1.73 ± 0.86 and 1.61 ± 0.42 mm in the upper liver. For hysteresis motion patterns, the corresponding values were 1.13 ± 0.37 and 1.45 ± 0.33 mm, 1.43 ± 1.30 and 1.67 ± 0.42 mm, and 1.20 ± 0.68 and 1.46 ± 0.54 mm in the central, lower, and upper livers, respectively.This study proposed a new hybrid correlation model for real‐time tumor tracking, which can be used to account for both linear and nonlinear motion patterns, as well as hysteresis motion patterns. Additionally, the NG‐RC model required shorter training data sets (15 s) during pre‐treatment and short internal motion sampling (every 30 s) during treatment compared to other ECMs. [ABSTRACT FROM AUTHOR]

Background: This study investigates different strategies for estimating internal liver tumor motion during radiotherapy based on continuous monitoring of external respiratory motion combined with sparse internal imaging., Methods: Fifteen patients underwent three-fraction stereotactic liver radiotherapy. The 3D internal tumor motion (INT) was monitored by electromagnetic transponders while a camera monitored the external marker block motion (EXT). The ability of four external-internal correlation models (ECM) to estimate INT as function of EXT was investigated: a simple linear model (ECM1), an augmented linear model (ECM2), an augmented quadratic model (ECM3), and an extended quadratic model (ECM4). Each ECM was constructed by fitting INT and EXT during the first 60s of each fraction. The fit accuracy was calculated as the root-mean-square error (RMSE) between ECM-estimated and actual tumor motion. Next, the RMSE of the ECM-estimated tumor motion throughout the fractions was calculated for four simulated ECM update strategies: (A) no update, 0.33Hz internal sampling with continuous update of either (B) all ECM parameters based on the last 2 minutes samples or (C) only the baseline term based on the last 5 samples, (D) full ECM update every minute using 20s continuous internal sampling., Results: The augmented quadratic ECM3 had best fit accuracy with mean (± SD)) RMSEs of 0.32 ± 0.11mm (left-right, LR), 0.79 ± 0.30mm (cranio-caudal, CC) and 0.56 ± 0.31mm (anterior-posterior, AP). However, the simpler augmented linear ECM2 combined with frequent baseline updates (update strategy C) gave best motion estimations with mean RMSEs of 0.41 ± 0.14mm (LR), 1.02 ± 0.33mm (CC) and 0.78 ± 0.48mm (AP). This was significantly better than all other ECM-update strategy combinations for CC motion (Wilcoxon signed rank p<0.05)., Conclusion: The augmented linear ECM2 combined with frequent baseline updates provided the best compromise between fit accuracy and robustness towards irregular motion. It allows accurate internal motion monitoring by combining external motioning with sparse 0.33Hz kV imaging, which is available at conventional linacs., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2024 Samadi Miandoab, Worm, Hansen, Weber, Høyer, Saramad, Setayeshi and Poulsen.)

Background: During X-ray imaging, pulmonary movements can cause many image artifacts. To tackle this issue, several studies, including mathematical algorithms and 2D-3D image registration methods, have been presented. Recently, the application of deep artificial neural networks has been considered for image generation and prediction., Objective: In this study, a convolutional long short-term memory (ConvLSTM) neural network is used to predict spatiotemporal 4DCT images., Material and Methods: In this analytical analysis study, two ConvLSTM structures, consisting of stacked ConvLSTM models along with the hyperparameter optimizer algorithm and a new design of the ConvLSTM model are proposed. The hyperparameter optimizer algorithm in the conventional ConvLSTM includes the number of layers, number of filters, kernel size, epoch number, optimizer, and learning rate. The two ConvLSTM structures were also evaluated through six experiments based on Root Mean Square Error (RMSE) and structural similarity index (SSIM)., Results: Comparing the two networks demonstrates that the new design of the ConvLSTM network is faster, more accurate, and more reliable in comparison to the tuned-stacked ConvLSTM model. For all patients, the estimated RMSE and SSIM were 3.17 and 0.988, respectively, and a significant improvement can be observed in comparison to the previous studies., Conclusion: Overall, the results of the new design of the ConvLSTM network show excellent performances in terms of RMSE and SSIM. Also, the generated CT images with the new design of the ConvLSTM model show a good consistency with the corresponding references regarding registration accuracy and robustness., Competing Interests: None, (Copyright: © Journal of Biomedical Physics and Engineering.)

Background: In external beam radiotherapy, a prediction model is required to compensate for the temporal system latency that affects the accuracy of radiation dose delivery. This study focused on a thorough comparison of seven deep artificial neural networks to propose an accurate and reliable prediction model., Methods: Seven deep predictor models are trained and tested with 800 breathing signals. In this regard, a nonsequential-correlated hyperparameter optimization algorithm is developed to find the best configuration of parameters for all models. The root mean square error (RMSE), mean absolute error, normalized RMSE, and statistical F-test are also used to evaluate network performance., Results: Overall, tuning the hyperparameters results in a 25%-30% improvement for all models compared to previous studies. The comparison between all models also shows that the gated recurrent unit (GRU) with RMSE = 0.108 ± 0.068 mm predicts respiratory signals with higher accuracy and better performance., Conclusion: Overall, tuning the hyperparameters in the GRU model demonstrates a better result than the hybrid predictor model used in the CyberKnife VSI system to compensate for the 115 ms system latency. Additionally, it is demonstrated that the tuned parameters have a significant impact on the prediction accuracy of each model., (© 2022 The Authors. Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.)