Deep nets vs expert designed features in medical physics: An IMRT QA case study

A unified deep-learning framework for enhanced patient-specific quality assurance of intensity-modulated radiation therapy plans

BACKGROUND

Modern radiation therapy techniques, such as intensity-modulated radiation therapy (IMRT) and volumetric-modulated arc therapy (VMAT), use complex fluence modulation strategies to achieve optimal patient dose distribution. Ensuring their accuracy necessitates rigorous patient-specific quality assurance (PSQA), traditionally done through pretreatment measurements with detector arrays. While effective, these methods are labor-intensive and time-consuming. Independent calculation-based methods leveraging advanced dose algorithms provide a reduced workload but cannot account for machine performance during delivery.

PURPOSE

This study introduces a novel unified deep-learning (DL) framework to enhance PSQA. The framework can combine the strengths of measurement- and calculation-based approaches.

METHODS

A comprehensive artificial training dataset, comprising 400,000 samples, was generated based on a rigorous mathematical model that describes the physical processes of radiation transport and interaction within both the medium and detector. This artificial data was used to pretrain the DL models, which were subsequently fine-tuned with a measured dataset of 400 IMRT segments to capture the machine-specific characteristics. Additional measurements of five IMRT plans were used as the unseen test dataset. Within the unified framework, a forward prediction model uses plan parameters to predict the measured dose distributions, while the backward prediction model reconstructs these parameters from actual measurements. The former enables a detailed control point (CP)-wise analysis. At the same time, the latter facilitates the reconstruction of treatment plans from the measurements and, subsequently, dose recalculation in the treatment planning system (TPS), as well as an independent second check software (VERIQA). This method has been tested with an OD 1600 SRS and an OD 1500 detector array with distinct spatial resolution and detector arrangement in combination with a dedicated upsampling model for the latter.

RESULTS

The final models could deliver highly accurate predictions of the measurements in the forward direction and the actual delivered plan parameters in the backward direction. In the forward direction, the test plans reached median gamma passing rates better than 94% for the OD 1600 SRS measurements. The upsampled OD 1500 measurements show similar performance with similar median gamma passing rates but a slightly higher variability. The 3D gamma passing rates from the comparisons between the original and reconstructed dose distributions in patients lie between 95.4% and 98.2% for the OD 1600 SRS and 94.7% and 98.5% for the interpolated OD 1500 measurements. The dose volume histograms (DVH) of the original and the reconstructed plans, recalculated in both the TPS and VERIQA, were evaluated for the organs at risk and targets based on clinical protocols and showed no clinically relevant deviations.

CONCLUSIONS

The flexibility of the implemented model architecture allows its adaptability to other delivery techniques and measurement modalities. Its utilization also reduces the requirements of the measurement devices. The proposed unified framework could play a decisive role in automating QA workflow, especially in the context of real-time adaptive radiation therapy (ART).

A StarGAN and transformer-based hybrid classification-regression model for multi-institution VMAT patient-specific quality assurance

BACKGROUND

The field of artificial intelligence (AI)-based patient-specific quality assurance (PSQA) for volumetric modulated arc therapy (VMAT) faces challenges in terms of developing general models across institutions due to the prevalence of multi-institution data collection and multivariate heterogeneity. Building a general model that is capable of handling diverse multi-institution data is critical for enabling large-scale integration and analysis.

PURPOSE

This study aims to develop a star generative adversarial network (StarGAN) and transformer-based hybrid classification-regression PSQA framework to address unification of heterogeneous data from different institutions.

METHODS

A StarGAN and transformer-based hybrid classification-regression model was developed as a general PSQA framework to predict gamma passing rates (GPRs) and classify quality assurance (QA) results as "Pass" or "Fail" at multiple institutions. A total of 1815 VMAT plans were collected from eight institutions to develop the general PSQA framework and perform clinical commissioning and implementation. Among them, 20 independent clinical plans from each of eight institutions, for a total of 160 plans, were used for the clinical commissioning, and 205 new clinical plans from eight institutions were used for clinical implementation.

RESULTS

For the 3%/3, 3%/2, and 2%/2 mm gamma criteria, the sensitivity of the proposed PSQA framework with pretraining was 90.13%, 92.03%, and 95.84%, respectively, while the specificity was 76.01%, 76.12%, and 85.34%, respectively. The mean absolute errors (MAEs) of the proposed PSQA framework with pretraining were 1.36%, 2.37%, and 3.96%, respectively, while the root-mean-square errors (RMSEs) were 2.31%, 3.89%, and 5.17%, respectively. The results demonstrated visible improvement at multiple institutions. For clinical commissioning, the deviations between the predicted and measured results were all within 3% for 3%/3 and 3%/2 mm at eight institutions. For clinical implementation, all failure plans were correctly identified by the proposed PSQA framework.

CONCLUSIONS

The general PSQA framework enables diverse clinical data sources to be handled to achieve enhanced model performance and generalizability, and provides a solution to the unification of heterogeneous data from different institutions to construct robust QA models. This approach can be clinically deployed for VMAT QA.

Technical note: A software tool to extract complexity metrics from radiotherapy treatment plans

BACKGROUND

Complexity metrics are mathematical quantities designed to quantify aspects of radiotherapy treatment plans that may affect both their deliverability and dosimetric accuracy. Despite numerous studies investigating their utility, there remains a notable absence of shared tools for their extraction.

PURPOSE

This study introduces UCoMX (Universal Complexity Metrics Extractor), a software package designed for the extraction of complexity metrics from the DICOM-RT plan files of radiotherapy treatments.

METHODS

UCoMX is developed around two extraction engines: VCoMX (VMAT Complexity Metrics Extractor) for VMAT/IMRT plans, and TCoMX (Tomotherapy Complexity Metrics Extractor) tailored for Helical Tomotherapy plans. The software, built using Matlab, is freely available in both Matlab-based and stand-alone versions. More than 90 complexity metrics, drawn from relevant literature, are implemented in the package: 43 for VMAT/IMRT and 51 for Helical Tomotherapy.

RESULTS

The package is designed to read DICOM-RT plan files generated by most commercially available Treatment Planning Systems (TPSs), across various treatment units. A reference dataset containing VMAT, IMRT, and Helical Tomotherapy plans is provided to serve as a reference for comparing UCoMX with other in-house systems available at other centers.

CONCLUSION

UCoMX offers a straightforward solution for extracting complexity metrics from radiotherapy plans. Its versatility is enhanced through different versions, including Matlab-based and stand-alone, and its compatibility with a wide range of commercially available TPSs and treatment units. UCoMX presents a free, user-friendly tool empowering researchers to compute the complexity of treatment plans efficiently.

Generating 3D images of VMAT plans for predictive models and activation maps associated with plan deliverability

BACKGROUND

Intensity modulation with dynamic multi-leaf collimator (MLC) and monitor unit (MU) changes across control points (CPs) characterizes volumetric modulated arc therapy (VMAT). The increased uncertainty in plan deliverability required patient-specific quality assurance (PSQA), which remained inefficient upon Quality Assurance (QA) failure. To prevent waste before QA, plan complexity metrics (PCMs) and machine learning models with the metrics were generated, which were lack of providing CP-specific information upon QA failures.

PURPOSE

By generating 3D images from digital imaging and comminications in medicine in radiation therapy (DICOM RT) plan, we proposed a predictive model that can estimate the deliverability of VMAT plans and visualize CP-specific regions associated with plan deliverability.

METHODS

The patient cohort consisted of 259 and 190 cases for left- and right-breast VMAT treatments, which were split into 235 and 166 cases for training and 24 cases from each treatment for testing the networks. Three-channel 3D images generated from DICOM RT plans were fed into a DenseNet-based deep learning network. To reflect VMAT plan complexity as an image, the first two channels described MLC and MU variations between two consecutive CPs, while the last channel assigned the beam field size. The network output was defined as binary classified PSQA results, indicating deliverability. The predictive performance was assessed by accuracy, sensitivity, specificity, F1-score, and area under the curve (AUC). The gradient-weighted class activation map (Grad-CAM) highlighted the regions of CPs in VMAT plans associated with deliverability, compared against PCMs by Spearman correlation.

RESULTS

The DenseNet-based predictive model yielded AUCs of 92.2% and 93.8%, F1-scores of 97.0% and 93.8% and accuracies of 95.8% and 91.7% for the left- and right-breast VMAT cases. Additionally, the specificity of 87.5% for both cases indicated that the predictive model accurately detected QA failing cases. The activation maps significantly differentiated QA failing-labeled from passing-labeled classes for the non-deliverable cases. The PCM with the highest correlation to the Grad-CAM varied from patient cases, implying that plan deliverability would be considered patient-specific.

CONCLUSION

This work demonstrated that the deep learning-based network based on visualization of dynamic VMAT plan information successfully predicted plan deliverability, which also provided control-point specific planning parameter information associated with plan deliverability in a patient-specific manner.

Deep learning for patient-specific quality assurance of volumetric modulated arc therapy: Prediction accuracy and cost-sensitive classification performance

PURPOSE

To evaluate a deep learning model's performance in predicting and classifying patient-specific quality assurance (PSQA) results for volumetric modulated arc therapy (VMAT), aiming to streamline PSQA workflows and reduce the onsite measurement workload.

METHODS

A total of 761 VMAT plans were analyzed using 3D-MResNet to process multileaf collimator images and monitor unit data, with the gamma passing rate (GPR) as the output. Thresholds for the predicted GPR (Th-p) and measured GPR (Th-m) were established to aid in PSQA decision-making, using cost curves and error rates to assess classification performance.

RESULTS

The mean absolute errors of the model for the test set were 1.63 % and 2.38 % at 3 %/2 mm and 2 %/2 mm, respectively. For the classification of the PSQA results, Th-m was 88.3 % at 2 %/2 mm and 93.3 % at 3 %/2 mm. The lowest cost-sensitive error rates of 0.0127 and 0.0925 were obtained when Th-p was set as 91.2 % at 2 %/2 mm and 96.4 % at 3 %/2 mm, respectively. Additionally, the 2 %/2 mm classifier also achieved a lower total expected cost of 0.069 compared with 0.110 for the 3 %/2 mm classifier. The deep learning classifier under the 2 %/2 mm gamma criterion had a sensitivity and specificity of 100 % (10/10) and 83.5 % (167/200), respectively, for the test set.

CONCLUSIONS

The developed 3D-MResNet model can accurately predict and classify PSQA results based on VMAT plans. The introduction of a deep learning model into the PSQA workflow has considerable potential for improving the VMAT PSQA process and reducing workloads.

From plan to delivery: Machine learning based positional accuracy prediction of multi-leaf collimator and estimation of delivery effect in volumetric modulated arc therapy

PURPOSE

The positional accuracy of MLC is an important element in establishing the exact dosimetry in VMAT. We comprehensively analyzed factors that may affect MLC positional accuracy in VMAT, and constructed a model to predict MLC positional deviation and estimate planning delivery quality according to the VMAT plans before delivery.

METHODS

A total of 744 "dynalog" files for 23 VMAT plans were extracted randomly from treatment database. Multi-correlation was used to analyzed the potential influences on MLC positional accuracy, including the spatial characteristics and temporal variability of VMAT fluence, and the mechanical wear parameters of MLC. We developed a model to forecast the accuracy of MLC moving position utilizing the random forest (RF) ensemble learning method. Spearman correlation was used to further investigate the associations between MLC positional deviation and dosage deviations as well as gamma passing rates.

RESULTS

The MLC positional deviation and effective impact factors show a strong multi-correlation (R = 0.701, p-value < 0.05). This leads to the development of a highly accurate prediction model with average variables explained of 95.03% and average MSE of 0.059 in the 5-fold cross-validation, and MSE of 0.074 for the test data was obtained. The absolute dose deviations caused by MLC positional deviation ranging from 12.948 to 210.235 cGy, while the relative volume deviation remained small at 0.470%-5.161%. The average MLC positional deviation correlated substantially with gamma passing rates (with correlation coefficient of -0.506 to -0.720 and p-value < 0.05) but marginally with dosage deviations (with correlation coefficient < 0.498 and p-value > 0.05).

CONCLUSIONS

The RF predictive model provides a prior tool for VMAT quality assurance.

Multi-granularity prior networks for uncertainty-informed patient-specific quality assurance

Deep Learning Automated Patient-Specific Quality Assurance (PSQA) aims to reduce clinical resource requirements. It is vital to ensure the safety and effectiveness of radiation therapy by predicting the dose difference metric (Gamma passing rate) and its distribution. However, current research overlooks uncertainty quantification in model predictions, limiting their trustworthiness in real clinical environments. This paper proposes a Multi-granularity Uncertainty Quantification (MGUQ) framework. A Bayesian framework that quantifies uncertainties at multiple granularities for multi-task PSQA, specifically Gamma Passing Rate (GPR) prediction and Dose Difference Prediction (DDP), integrates visualization-based interactive components. Using Bayesian theory, we derive a comprehensive multi-granularity loss function that comprises granularity-specific loss and coherence loss components. Additionally, we proposed Multi-granularity Prior Networks, a dual-stream network architecture, to infer the distributions of DDP (modeled as t-distributions) and GPR (modeled as Gaussian distributions) under specific statistical assumptions. Comprehensive evaluations are conducted on a dataset from ''Peeking Union Medical College Hospital'', and results show that our proposed method achieves a minimum MAE loss of 0.864 with a 2%/3 mm criterion and realizes the uncertainty visualization of dose difference. Further, it also achieves 100% Clinical Accuracy (CA) with a workload of 67.2%. Experiments demonstrate that the proposed framework can enhance the trustworthiness of deep learning applications in PSQA.

Deep learning applied to dose prediction in external radiation therapy: A narrative review

Over the last decades, the use of artificial intelligence, machine learning and deep learning in medical fields has skyrocketed. Well known for their results in segmentation, motion management and posttreatment outcome tasks, investigations of machine learning and deep learning models as fast dose calculation or quality assurance tools have been present since 2000. The main motivation for this increasing research and interest in artificial intelligence, machine learning and deep learning is the enhancement of treatment workflows, specifically dosimetry and quality assurance accuracy and time points, which remain important time-consuming aspects of clinical patient management. Since 2014, the evolution of models and architectures for dose calculation has been related to innovations and interest in the theory of information research with pronounced improvements in architecture design. The use of knowledge-based approaches to patient-specific methods has also considerably improved the accuracy of dose predictions. This paper covers the state of all known deep learning architectures and models applied to external radiotherapy with a description of each architecture, followed by a discussion on the performance and future of deep learning predictive models in external radiotherapy.

Forecasting model for qualitative prediction of the results of patient-specific quality assurance based on planning and complexity metrics and their interrelations. Pilot study

Background

The purpose was to analyse the interrelations between planning and complexity metrics and gamma passing rates (GPRs) obtained from VMAT treatments and build the forecasting models for qualitative prediction (QD) of GPRs results.

Materials and method

802 treatment arcs from the plans prepared for the head and neck, thorax, abdomen, and pelvic cancers were analysed. The plans were verified by portal dosimetry and analysed twice using the gamma method with 3%|2mm and 2%|2mm acceptance criteria. The tolerance limit of GPR was 95%. Red, yellow, and green QDs were established for GPR examination. The interrelations were examined, as well as the analysis of effective differentiation of QD. Three models for QD forecasting based on discriminant analysis (DA), random decision forest (RDF) methods, and the hybrid model (HM) were built and evaluated.

Results

Most of the interrelations were small or moderate. The exception is correlations of the join function with the average number of monitor units per control point (R = 0.893) and the beam aperture with planning target volume (R = 0.897). While many metrics allow for the effective separation of the QDs from each other, the study shows that predicting the values of the QD is possible only through multi-component forecasting models, of which the HM is the most accurate (0.894).

Conclusion

Of the three models explored in this study, the HM, which uses DA methods to predict red QD and RDF methods to predict green and yellow QDs, is the most promising one.

Assessment of the deep learning-based gamma passing rate prediction system for 1.5 T magnetic resonance-guided linear accelerator

Measurement-based verification is impossible for the patient-specific quality assurance (QA) of online adaptive magnetic resonance imaging-guided radiotherapy (oMRgRT) because the patient remains on the couch throughout the session. We assessed a deep learning (DL) system for oMRgRT to predict the gamma passing rate (GPR). This study collected 125 verification plans [reference plan (RP), 100; adapted plan (AP), 25] from patients with prostate cancer treated using Elekta Unity. Based on our previous study, we employed a convolutional neural network that predicted the GPRs of nine pairs of gamma criteria from 1%/1 mm to 3%/3 mm. First, we trained and tested the DL model using RPs (n = 75 and n = 25 for training and testing, respectively) for its optimization. Second, we tested the GPR prediction accuracy using APs to determine whether the DL model could be applied to APs. The mean absolute error (MAE) and correlation coefficient (r) of the RPs were 1.22 ± 0.27% and 0.29 ± 0.10 in 3%/2 mm, 1.35 ± 0.16% and 0.37 ± 0.15 in 2%/2 mm, and 3.62 ± 0.55% and 0.32 ± 0.14 in 1%/1 mm, respectively. The MAE and r of the APs were 1.13 ± 0.33% and 0.35 ± 0.22 in 3%/2 mm, 1.68 ± 0.47% and 0.30 ± 0.11 in 2%/2 mm, and 5.08 ± 0.29% and 0.15 ± 0.10 in 1%/1 mm, respectively. The time cost was within 3 s for the prediction. The results suggest the DL-based model has the potential for rapid GPR prediction in Elekta Unity.

Neural Network Model Based on Branch Architecture for the Quality Assurance of Volumetric Modulated Arc Therapy

Radiation therapy relies on quality assurance (QA) to verify dose delivery accuracy. However, current QA methods suffer from operation lag as well as inaccurate performance. Hence, to address these shortcomings, this paper proposes a QA neural network model based on branch architecture, which is based on the analysis of the category features of the QA complexity metrics. The designed branch network focuses on category features, which effectively improves the feature extraction capability for complexity metrics. The branch features extracted by the model are fused to predict the GPR for more accurate QA. The performance of the proposed method was validated on the collected dataset. The experiments show that the prediction performance of the model outperforms other QA methods; the average prediction errors for the test set are 2.12% (2%/2 mm), 1.69% (3%/2 mm), and 1.30% (3%/3 mm). Moreover, the results indicate that two-thirds of the validation samples' model predictions perform better than the clinical evaluation results, suggesting that the proposed model can assist physicists in the clinic.

A Review of Artificial Intelligence Application for Radiotherapy

Background and Purpose

Artificial intelligence (AI) is a technique which tries to think like humans and mimic human behaviors. It has been considered as an alternative in a lot of human-dependent steps in radiotherapy (RT), since the human participation is a principal uncertainty source in RT. The aim of this work is to provide a systematic summary of the current literature on AI application for RT, and to clarify its role for RT practice in terms of clinical views.

Materials and Methods

A systematic literature search of PubMed and Google Scholar was performed to identify original articles involving the AI applications in RT from the inception to 2022. Studies were included if they reported original data and explored the clinical applications of AI in RT.

Results

The selected studies were categorized into three aspects of RT: organ and lesion segmentation, treatment planning and quality assurance. For each aspect, this review discussed how these AI tools could be involved in the RT protocol.

Conclusions

Our study revealed that AI was a potential alternative for the human-dependent steps in the complex process of RT.

Plan complexity metrics for head and neck VMAT competition plans

Previous plan competitions have largely focused on dose metric assessments. However, whether the submitted plans were realistic and reasonable from a quality assurance (QA) perspective remains unclear. This study aimed to investigate the relationship between aperture-based plan complexity metrics (PCM) in volumetric modulated arc therapy (VMAT) competition plans and clinical treatment plans verified through patient-specific QA (PSQA). In addition, the association of PCMs with plan quality was examined. A head and neck (HN) plan competition was held for Japanese institutions from June 2019 to July 2019, in which 210 competition plans were submitted. Dose distribution quality was quantified based on dose-volume histogram (DVH) metrics by calculating the dose distribution plan score (DDPS). Differences in PCMs between the two VMAT treatment plan groups (HN plan competitions held in Japan and clinically accepted HN VMAT plans through PSQA) were investigated. The mean (± standard deviation) DDPS for the 98 HN competition plans was 158.5 ± 20.6 (maximum DDPS: 200). DDPS showed a weak correlation with PCMs with a maximum r of 0.45 for monitor unit (MU); its correlation with some PCMs was "very weak." Significant differences were found in some PCMs between plans with the highest 20% DDPSs and the remaining plans. The clinical VMAT and competition plans revealed similar distributions for some PCMs. Deviations in PCMs for the two groups were comparable, indicating considerable variability among planners regarding planning skills. The plan complexity for HN VMAT competition plans increased for high-quality plans, as shown by the dose distribution. Direct comparison of PCMs between competition plans and clinically accepted plans showed that the submitted HN VMAT competition plans were realistic and reasonable from the QA perspective. This evaluation may provide a set of criteria for evaluating plan quality in plan competitions.

Invertible and Variable Augmented Network for Pretreatment Patient-Specific Quality Assurance Dose Prediction

Pretreatment patient-specific quality assurance (prePSQA) is conducted to confirm the accuracy of the radiotherapy dose delivered. However, the process of prePSQA measurement is time consuming and exacerbates the workload for medical physicists. The purpose of this work is to propose a novel deep learning (DL) network to improve the accuracy and efficiency of prePSQA. A modified invertible and variable augmented network was developed to predict the three-dimensional (3D) measurement-guided dose (MDose) distribution of 300 cancer patients who underwent volumetric modulated arc therapy (VMAT) between 2018 and 2021, in which 240 cases were randomly selected for training, and 60 for testing. For simplicity, the present approach was termed as "IVPSQA." The input data include CT images, radiotherapy dose exported from the treatment planning system, and MDose distribution extracted from the verification system. Adam algorithm was used for first-order gradient-based optimization of stochastic objective functions. The IVPSQA model obtained high-quality 3D prePSQA dose distribution maps in head and neck, chest, and abdomen cases, and outperformed the existing U-Net-based prediction approaches in terms of dose difference maps and horizontal profiles comparison. Moreover, quantitative evaluation metrics including SSIM, MSE, and MAE demonstrated that the proposed approach achieved a good agreement with ground truth and yield promising gains over other advanced methods. This study presented the first work on predicting 3D prePSQA dose distribution by using the IVPSQA model. The proposed method could be taken as a clinical guidance tool and help medical physicists to reduce the measurement work of prePSQA.

A TPS integrated machine learning tool for predicting patient-specific quality assurance outcomes in volumetric-modulated arc therapy

PURPOSE

Machine learning (ML) models have been demonstrated to be beneficial for optimizing the workload of patient-specific quality assurance (PSQA). Implementing them in clinical routine frequently requires third-party applications beyond the treatment planning system (TPS), slowing down the workflow. To address this issue, a PSQA outcomes predictive model was carefully selected and validated before being fully integrated into the TPS.

MATERIALS AND METHODS

Nine ML algorithms were evaluated using cross-validation. The learning database was built by calculating complexity metrics (CM) and binarizing PSQA results into "pass"/"fail" classes for 1767 VMAT arcs. The predictive performance was evaluated using area under the ROC curve (AUROC), sensitivity, and specificity. The ML model was integrated into the TPS via a C# script. Script-guided reoptimization impact on PSQA and dosimetric results was evaluated on ten VMAT plans with "fail"-predicted arcs. Workload reduction potential was also assessed.

RESULTS

The selected model exhibited an AUROC of 0.88, with a sensitivity and specificity exceeding 50 % and 90 %, respectively. The script-guided reoptimization of the ten evaluated plans led to an average improvement of 1.4 ± 0.9 percentage points in PSQA results, while preserving the quality of the dose distribution. A yearly savings of about 140 h with the use of the script was estimated.

CONCLUSIONS

The proposed script is a valuable complementary tool for PSQA measurement. It was efficiently integrated into the clinical workflow to enhance PSQA outcomes and reduce PSQA workload by decreasing the risk of failing QA and thereby, the need for repeated replanning and measurements.

Advancing equitable and personalized cancer care: Novel applications and priorities of artificial intelligence for fairness and inclusivity in the patient care workflow

Patient care workflows are highly multimodal and intertwined: the intersection of data outputs provided from different disciplines and in different formats remains one of the main challenges of modern oncology. Artificial Intelligence (AI) has the potential to revolutionize the current clinical practice of oncology owing to advancements in digitalization, database expansion, computational technologies, and algorithmic innovations that facilitate discernment of complex relationships in multimodal data. Within oncology, radiation therapy (RT) represents an increasingly complex working procedure, involving many labor-intensive and operator-dependent tasks. In this context, AI has gained momentum as a powerful tool to standardize treatment performance and reduce inter-observer variability in a time-efficient manner. This review explores the hurdles associated with the development, implementation, and maintenance of AI platforms and highlights current measures in place to address them. In examining AI's role in oncology workflows, we underscore that a thorough and critical consideration of these challenges is the only way to ensure equitable and unbiased care delivery, ultimately serving patients' survival and quality of life.

A feasibility study to predict 3D dose delivery accuracy for IMRT using DenseNet with log files

OBJECTIVE

This study aims to explore the feasibility of DenseNet in the establishment of a three-dimensional (3D) gamma prediction model of IMRT based on the actual parameters recorded in the log files during delivery.

METHODS

A total of 55 IMRT plans (including 367 fields) were randomly selected. The gamma analysis was performed using gamma criteria of 3% /3 mm (Dose Difference/Distance to Agreement), 3% /2 mm, 2% /3 mm, and 2% /2 mm with a 10% dose threshold. In addition, the log files that recorded the gantry angle, monitor units (MU), multi-leaf collimator (MLC), and jaws position during delivery were collected. These log files were then converted to MU-weighted fluence maps as the input of DenseNet, gamma passing rates (GPRs) under four different gamma criteria as the output, and mean square errors (MSEs) as the loss function of this model.

RESULTS

Under different gamma criteria, the accuracy of a 3D GPR prediction model decreased with the implementation of stricter gamma criteria. In the test set, the mean absolute error (MAE) of the prediction model under the gamma criteria of 3% /3 mm, 2% /3 mm, 3% /2 mm, and 2% /2 mm was 1.41, 1.44, 3.29, and 3.54, respectively; the root mean square error (RMSE) was 1.91, 1.85, 4.27, and 4.40, respectively; the Sr was 0.487, 0.554, 0.573, and 0.506, respectively. There was a correlation between predicted and measured GPRs (P < 0.01). Additionally, there was no significant difference in the accuracy between the validation set and the test set. The accuracy in the high GPR group was high, and the MAE in the high GPR group was smaller than that in the low GPR group under four different gamma criteria.

CONCLUSIONS

In this study, a 3D GPR prediction model of patient-specific QA using DenseNet was established based on log files. As an auxiliary tool for 3D dose verification in IMRT, this model is expected to improve the accuracy and efficiency of dose validation.

Prediction of patient-specific quality assurance for volumetric modulated arc therapy using radiomics-based machine learning with dose distribution

PURPOSE

We sought to develop machine learning models to predict the results of patient-specific quality assurance (QA) for volumetric modulated arc therapy (VMAT), which were represented by several dose-evaluation metrics-including the gamma passing rates (GPRs)-and criteria based on the radiomic features of 3D dose distribution in a phantom.

METHODS

A total of 4,250 radiomic features of 3D dose distribution in a cylindrical dummy phantom for 140 arcs from 106 clinical VMAT plans were extracted. We obtained the following dose-evaluation metrics: GPRs with global and local normalization, the dose difference (DD) in 1% and 2% passing rates (DD1% and DD2%) for 10% and 50% dose threshold, and the distance-to-agreement in 1-mm and 2-mm passing rates (DTA1 mm and DTA2 mm) for 0.5%/mm and 1.0%.mm dose gradient threshold determined by measurement using a diode array in patient-specific QA. The machine learning regression models for predicting the values of the dose-evaluation metrics using the radiomic features were developed based on the elastic net (EN) and extra trees (ET) models. The feature selection and tuning of hyperparameters were performed with nested cross-validation in which four-fold cross-validation is used within the inner loop, and the performance of each model was evaluated in terms of the root mean square error (RMSE), the mean absolute error (MAE), and Spearman's rank correlation coefficient.

RESULTS

The RMSE and MAE for the developed machine learning models ranged from <1% to nearly <10% depending on the dose-evaluation metric, the criteria, and dose and dose gradient thresholds used for both machine learning models. It was advantageous to focus on high dose region for predicating global GPR, DDs, and DTAs. For certain metrics and criteria, it was possible to create models applicable for patients' heterogeneity by training only with dose distributions in phantom.

CONCLUSIONS

The developed machine learning models showed high performance for predicting dose-evaluation metrics especially for high dose region depending on the metric and criteria. Our results demonstrate that the radiomic features of dose distribution can be considered good indicators of the plan complexity and useful in predicting measured dose evaluation metrics.

Virtual pretreatment patient-specific quality assurance of volumetric modulated arc therapy using deep learning

BACKGROUND

Automatic patient-specific quality assurance (PSQA) is recently explored using artificial intelligence approaches, and several studies reported the development of machine learning models for predicting the gamma pass rate (GPR) index only.

PURPOSE

To develop a novel deep learning approach using a generative adversarial network (GAN) to predict the synthetic measured fluence.

METHODS AND MATERIALS

A novel training method called "dual training," which involves the training of the encoder and decoder separately, was proposed and evaluated for cycle GAN (cycle-GAN) and conditional GAN (c-GAN). A total of 164 VMAT treatment plans, including 344 arcs (training data: 262, validation data: 30, and testing data: 52) from various treatment sites, were selected for prediction model development. For each patient, portal-dose-image-prediction fluence from TPS was used as input, and measured fluence from EPID was used as output/response for model training. Predicted GPR was derived by comparing the TPS fluence with the synthetic measured fluence generated by the DL models using gamma evaluation of criteria 2%/2 mm. The performance of dual training was compared against the traditional single-training approach. In addition, we also developed a separate classification model specifically designed to detect automatically three types of errors (rotational, translational, and MU-scale) in the synthetic EPID-measured fluence.

RESULTS

Overall, the dual training improved the prediction accuracy of both cycle-GAN and c-GAN. Predicted GPR results of single training were within 3% for 71.2% and 78.8% of test cases for cycle-GAN and c-GAN, respectively. Moreover, similar results for dual training were 82.7% and 88.5% for cycle-GAN and c-GAN, respectively. The error detection model showed high classification accuracy (>98%) for detecting errors related to rotational and translational errors. However, it struggled to differentiate the fluences with "MU scale error" from "error-free" fluences.

CONCLUSION

We developed a method to automatically generate the synthetic measured fluence and identify errors within them. The proposed dual training improved the PSQA prediction accuracy of both the GAN models, with c-GAN demonstrating superior performance over the cycle-GAN. Our results indicate that the c-GAN with dual training approach combined with error detection model, can accurately generate the synthetic measured fluence for VMAT PSQA and identify the errors. This approach has the potential to pave the way for virtual patient-specific QA of VMAT treatments.

Pretreatment patient-specific quality assurance prediction based on 1D complexity metrics and 3D planning dose: classification, gamma passing rates, and DVH metrics

PURPOSE

Highly modulated radiotherapy plans aim to achieve target conformality and spare organs at risk, but the high complexity of the plan may increase the uncertainty of treatment. Thus, patient-specific quality assurance (PSQA) plays a crucial role in ensuring treatment accuracy and providing clinical guidance. This study aims to propose a prediction model based on complexity metrics and patient planning dose for PSQA results.

MATERIALS AND METHODS

Planning dose, measurement-based reconstructed dose and plan complexity metrics of the 687 radiotherapy plans of patients treated in our institution were collected for model establishing. Global gamma passing rate (GPR, 3%/2mm,10% threshold) of 90% was used as QA criterion. Neural architecture models based on Swin-transformer were adapted to process 3D dose and incorporate 1D metrics to predict QA results. The dataset was divided into training (447), validation (90), and testing (150) sets. Evaluation of predictions was performed using mean absolute error (MAE) for GPR, planning target volume (PTV) HI and PTV CI, mean absolute percentage error (MAPE) for PTV D95, PTV D2 and PTV Dmean, and the area under the receiver operating characteristic (ROC) curve (AUC) for classification. Furthermore, we also compare the prediction results with other models based on either only 1D or 3D inputs.

RESULTS

In this dataset, 72.8% (500/687) plans passed the pretreatment QA under the criterion. On the testing set, our model achieves the highest performance, with the 1D model slightly surpassing the 3D model. The performance results are as follows (combine, 1D, and 3D transformer): The AUCs are 0.92, 0.88 and 0.86 for QA classification. The MAEs of prediction are 0.039, 0.046, and 0.040 for 3D GPR, 0.018, 0.021, and 0.019 for PTV HI, and 0.075, 0.078, and 0.084 for PTV CI. Specifically, for cases with 3D GPRs greater than 90%, the MAE could achieve 0.020 (combine). The MAPE of prediction is 1.23%, 1.52%, and 1.66% for PTV D95, 2.36%, 2.67%, and 2.45% for PTV D2, and 1.46%, 1.70%, and 1.71% for PTV Dmean.

CONCLUSION

The model based on 1D complexity metrics and 3D planning dose could predict pretreatment PSQA results with high accuracy and the complexity metrics play a leading role in the model. Furthermore, dose-volume metric deviations of PTV could be predicted and more clinically valuable information could be provided.

A new architecture combining convolutional and transformer-based networks for automatic 3D multi-organ segmentation on CT images

PURPOSE

Deep learning-based networks have become increasingly popular in the field of medical image segmentation. The purpose of this research was to develop and optimize a new architecture for automatic segmentation of the prostate gland and normal organs in the pelvic, thoracic, and upper gastro-intestinal (GI) regions.

METHODS

We developed an architecture which combines a shifted-window (Swin) transformer with a convolutional U-Net. The network includes a parallel encoder, a cross-fusion block, and a CNN-based decoder to extract local and global information and merge related features on the same scale. A skip connection is applied between the cross-fusion block and decoder to integrate low-level semantic features. Attention gates (AGs) are integrated within the CNN to suppress features in image background regions. Our network is termed "SwinAttUNet." We optimized the architecture for automatic image segmentation. Training datasets consisted of planning-CT datasets from 300 prostate cancer patients from an institutional database and 100 CT datasets from a publicly available dataset (CT-ORG). Images were linearly interpolated and resampled to a spatial resolution of (1.0 × 1.0× 1.5) mm3 . A volume patch (192 × 192 × 96) was used for training and inference, and the dataset was split into training (75%), validation (10%), and test (15%) cohorts. Data augmentation transforms were applied consisting of random flip, rotation, and intensity scaling. The loss function comprised Dice and cross-entropy equally weighted and summed. We evaluated Dice coefficients (DSC), 95th percentile Hausdorff Distances (HD95), and Average Surface Distances (ASD) between results of our network and ground truth data.

RESULTS

SwinAttUNet, DSC values were 86.54 ± 1.21, 94.15 ± 1.17, and 87.15 ± 1.68% and HD95 values were 5.06 ± 1.42, 3.16 ± 0.93, and 5.54 ± 1.63 mm for the prostate, bladder, and rectum, respectively. Respective ASD values were 1.45 ± 0.57, 0.82 ± 0.12, and 1.42 ± 0.38 mm. For the lung, liver, kidneys and pelvic bones, respective DSC values were: 97.90 ± 0.80, 96.16 ± 0.76, 93.74 ± 2.25, and 89.31 ± 3.87%. Respective HD95 values were: 5.13 ± 4.11, 2.73 ± 1.19, 2.29 ± 1.47, and 5.31 ± 1.25 mm. Respective ASD values were: 1.88 ± 1.45, 1.78 ± 1.21, 0.71 ± 0.43, and 1.21 ± 1.11 mm. Our network outperformed several existing deep learning approaches using only attention-based convolutional or Transformer-based feature strategies, as detailed in the results section.

CONCLUSIONS

We have demonstrated that our new architecture combining Transformer- and convolution-based features is able to better learn the local and global context for automatic segmentation of multi-organ, CT-based anatomy.

TransQA: deep hybrid transformer network for measurement-guided volumetric dose prediction of pre-treatment patient-specific quality assurance

Objective. Performing pre-treatment patient-specific quality assurance (prePSQA) is considered an essential, time-consuming, and resource-intensive task for volumetric modulated arc radiotherapy (VMAT) which confirms the dose accuracy and ensure patient safety. Most current machine learning and deep learning approaches stack excessive convolutional/pooling operations (CPs) to predict prePSQA with two-dimensional or one-dimensional information input. However, these models generally present limitations in explicitly modeling long-range dependency for volumetric dose prediction due to the loss of spatial dose features and the inherent locality of CPs. The purpose of this work is to construct a deep hybrid network by combining the self-attention mechanism-based Transformer with modified U-Net for predicting measurement-guided volumetric dose (MDose) of prePSQA.Approach. The enrolled 307 cancer patients underwent VMAT were randomly divided into 246 and 61 cases for training and testing the model. The input data included computed tomography images, radiotherapy dose images exported from the treatment planning system, as well as the MDose distribution from the verification system. The output was the predicted high-quality voxel-wise prePSQA dose distribution.Main results: qualitative and quantitative experimental results show that the proposed prediction method could achieve comparable or better performance on MDose prediction over other approaches in terms of spatial dose distribution, dose-volume histogram metrics, gamma passing rates, mean absolute error, root mean square error, and structural similarity.Significance. The preliminary results on multiple cancer sites show that our approach can be taken as a clinical guidance tool and help medical physicists to reduce the measurement work of prePSQA.

Improvement of deep learning prediction model in patient-specific QA for VMAT with MLC leaf position map and patient's dose distribution

PURPOSE

Deep learning-based virtual patient-specific quality assurance (QA) is a novel technique that enables patient QA without measurement. However, this method could be improved by further evaluating the optimal data to be used as input. Therefore, a deep learning-based model that uses multileaf collimator (MLC) information per control point and dose distribution in patient's CT as inputs was developed.

METHODS

Overall, 96 volumetric-modulated arc therapy plans generated for prostate cancer treatment were used. We developed a model (Model 1) that can predict measurement-based gamma passing rate (GPR) for a treatment plan using data stored as a map reflecting the MLC leaf position at each control point (MLPM) and data of the dose distribution in patient's CT as inputs. The evaluation of the model was based on the mean absolute error (MAE) and Pearson's correlation coefficient (r) between the measured and predicted GPR. For comparison, we also analyzed models trained with the dose distribution in patient's CT alone (Model 2) and with dose distributions recalculated on a virtual phantom CT (Model 3).

RESULTS

At the 2%/2 mm criterion, MAE[%] and r for Model 1, Model 2, and Model 3 were 2.32% ± 0.43% and 0.54 ± 0.03, 2.70% ± 0.26%, and 0.32 ± 0.08, and 2.96% ± 0.23% and 0.24 ± 0.22, respectively; at the 3%/3 mm criterion, these values were 1.25% ± 0.05% and 0.36 ± 0.18, 1.57% ± 0.35% and 0.19 ± 0.20, and 1.39% ± 0.32% and 0.17 ± 0.22, respectively. This result showed that Model 1 exhibited the lowest MAE and highest r at both criteria of 2%/2 mm and 3%3 mm.

CONCLUSIONS

These findings showed that a model that combines the MLPM and dose distribution in patient's CT exhibited a better GPR prediction performance compared with the other two studied models.

Evaluation of the dataset quality in gamma passing rate predictions using machine learning methods

OBJECTIVE

Gamma passing rate (GPR) predictions using machine learning methods have been explored for treatment verification of radiotherapy plans. However, these methods presented datasets with unbalanced number of plans having different treatment conditions (heterogeneous datasets), such as anatomical sites or dose per fractions, leading to lower model interpretability and prediction performance.

METHODS

We investigated the impact of the dataset composition on GPR binary classification (pass/fail) using random forest (RF), XG-boost, and neural network (NN) models. 945 plans were used to create one reference dataset (randomly assembled) and 24 customized datasets that considered four heterogeneity factors independently (anatomical region, number of arcs, dose per fraction, and treatment unit). 309 predictor features were extracted and calculated from plan parameters, modulation complexity metrics, and radiomic analysis (leave-trajectory maps, 3D dose distributions, and portal dosimetry images). The models' performances were measured using the area under the curve from the receiver operating characteristic (ROC-AUC).

RESULTS

Radiomics features for reference models increased ROC-AUC values up to 13%, 15%, and 5% for RF, XG-Boost, and NN, respectively. The datasets with higher heterogeneous conditions presented the lower ROC-AUC values (RF: 0.72 ± 0.11, XG-Boost: 0.67 ± 0.1, NN: 0.89 ± 0.05) compared to models with less heterogeneous treatment conditions (RF: 0.88 ± 0.06, XG-Boost: 0.89 ± 0.07, NN: 0.98 ± 0.01). The ten most important features for each heterogeneity dataset group demonstrated their correlation with the treatments' physical aspects and GPR prediction.

CONCLUSION

Improvements in data generalization and model performances can be associated with datasets having similar treatment conditions. This analysis might be implemented to evaluate the dataset quality and model consistency of further ML applications in radiotherapy.

ADVANCES IN KNOWLEDGE

Dataset heterogeneities decrease ML model performance and reliability.

Multicentric evaluation of a machine learning model to streamline the radiotherapy patient specific quality assurance process

PURPOSE

Patient-specific quality assurance (PSQA) is performed to ensure that modulated treatment plans can be delivered as intended, but constitutes a substantial workload that could slow down the radiotherapy process and delay the start of clinical treatments. In this study, we investigated a machine learning (ML) tree-based ensemble model to predict the gamma passing rate (GPR) for volumetric modulated arc therapy (VMAT) plans.

MATERIALS AND METHODS

5622 VMAT plans from multiple treatment sites were selected from a database of Institution 1 and the ML model trained using 19 metrics. PSQA analyses were performed automatically using criteria 3%/1 mm (global normalization, absolute dose, 10% threshold) and 95% action limit. Model's performance was evaluated on an out-of-sample test set of Institution 1 and on two independent sets of measurements collected at Institution 2 and Institution 3. Mean absolute error (MAE), as well as the model's sensitivity and specificity, were computed.

RESULTS

The model obtained a MAE of 2.33%, 2.54% and 3.91% for the three Institutions, with a specificity of 0.90, 0.90 and 0.68, and a sensitivity of 0.61, 0.25, and 0.55, respectively. Small positive median values of the residuals (i.e., the difference between measurements and predictions) were observed for each Institution (0.95%, 1.66%, and 3.42%). Thus, the model's predictions were, on average, close to the real values and provided a conservative estimation of the GPR.

CONCLUSIONS

ML models can be integrated into clinical practice to streamline the radiotherapy workflow, but they should be center-specific or thoroughly verified within centers before clinical use.

IMRT QA result prediction via MLC transmission decomposition

BACKGROUND

Quality assurance measurement of IMRT/VMAT treatment plans is resource intensive, and other more efficient methods to achieve the same confidence are desirable.

PURPOSE

We aimed to analyze treatment plans in the context of the treatment planning systems that created them, in order to predict which ones will fail a standard quality assurance measurement. To do so, we sought to create a tool external to the treatment planning system that could analyze a set of MLC positions and provide information that could be used to calculate various evaluation metrics.

METHODS

The tool was created in Python to read in DICOM plan files and determine the beam fluence fraction incident on each of seven different zones, each classified based on the RayStation MLC model. The fractions, termed grid point fractions, were validated by analyzing simple test plans. The average grid point fractions, over all control points for 46 plans were then computed. These values were then compared with gamma analysis pass percentages and median dose differences to determine if any significant correlations existed.

RESULTS

Significant correlation was found between the grid point fraction metrics and median dose differences, but not with gamma analysis pass percentages. Correlations were positive or negative, suggesting differing model parameter value sensitivities, as well as potential insight into the treatment planning system dose model.

CONCLUSIONS

By decomposing MLC control points into different transmission zones, it is possible to create a metric that predicts whether the analyzed plan will pass a quality assurance measurement from a dose calculation accuracy standpoint. The tool and metrics developed in this work have potential applications in comparing clinical beam models or identifying their weak points. Implementing the tool within a treatment planning system would also provide more potential plan optimization parameters.

Deep Hybrid Learning Prediction of Patient-Specific Quality Assurance in Radiotherapy: Implementation in Clinical Routine

BACKGROUND

Arc therapy allows for better dose deposition conformation, but the radiotherapy plans (RT plans) are more complex, requiring patient-specific pre-treatment quality assurance (QA). In turn, pre-treatment QA adds to the workload. The objective of this study was to develop a predictive model of Delta4-QA results based on RT-plan complexity indices to reduce QA workload.

METHODS

Six complexity indices were extracted from 1632 RT VMAT plans. A machine learning (ML) model was developed for classification purpose (two classes: compliance with the QA plan or not). For more complex locations (breast, pelvis and head and neck), innovative deep hybrid learning (DHL) was trained to achieve better performance.

RESULTS

For not complex RT plans (with brain and thorax tumor locations), the ML model achieved 100% specificity and 98.9% sensitivity. However, for more complex RT plans, specificity falls to 87%. For these complex RT plans, an innovative QA classification method using DHL was developed and achieved a sensitivity of 100% and a specificity of 97.72%.

CONCLUSIONS

The ML and DHL models predicted QA results with a high degree of accuracy. Our predictive QA online platform is offering substantial time savings in terms of accelerator occupancy and working time.

Prediction models as decision-support tools for virtual patient-specific quality assurance of helical tomotherapy plans

Background and purpose

Prediction models may be reliable decision-support tools to reduce the workload associated with the measurement-based patient-specific quality assurance (PSQA) of radiotherapy plans. This study compared the effectiveness of three different models based on delivery parameters, complexity metrics and sinogram radiomics features as tools for virtual-PSQA (vPSQA) of helical tomotherapy (HT) plans.

Materials and methods

A dataset including 881 RT plans created with two different treatment planning systems (TPSs) was collected. Sixty-five indicators including 12 delivery parameters (DP) and 53 complexity metrics (CM) were extracted using a dedicated software library. Additionally, 174 radiomics features (RF) were extracted from the plans' sinograms. Three groups of variables were formed: A (DP), B (DP + CM) and C (DP + CM + RF). Regression models were trained to predict the gamma index passing rate P R γ (3%G, 2mm) and the impact of each group of variables was investigated. ROC-AUC analysis measured the ability of the models to accurately discriminate between 'deliverable' and 'non-deliverable' plans.

Results

The best performance was achieved by model C which allowed detecting around 16% and 63% of the 'deliverable' plans with 100% sensitivity for the two TPSs, respectively. In a real clinical scenario, this would have decreased the whole PSQA workload by approximately 35%.

Conclusions

The combination of delivery parameters, complexity metrics and sinogram radiomics features allows for robust and reliable PSQA gamma passing rate predictions and high-sensitivity detection of a fraction of deliverable plans for one of the two TPSs. Promising yet improvable results were obtained for the other one. The results foster a future adoption of vPSQA programs for HT.

Patient-specific quality assurance prediction models based on machine learning for novel dual-layered MLC linac

BACKGROUND

Patient-specific quality assurance (PSQA) is an indispensable and essential procedure in radiotherapy workflow, and several studies have been done to develop prediction models based on the conventional C-arm linac of single-layered multileaf collimator (MLC) with machine learning (ML) and deep learning techniques to predict PSQA results and improve efficiency. Recently, a newly designed O-ring gantry linac "Halcyon" equipped with unique jawless stacked-and-staggered dual-layered MLC was released. However, few studies have focused on developing PSQA prediction models for this novel dual-layered MLC system.

PURPOSE

To evaluate the performance of ML to predict PSQA results of fixed field intensity-modulated radiation therapy (FF-IMRT) plans for linac equipped with dual-layered MLC.

METHODS AND MATERIALS

A total of 213 FF-IMRT treatment plans, including 1383 beams from various treatment sites, were selected and delivered with portal dosimetry verification on Halcyon linac. Gamma analysis was performed using 1%/1, 2%/2, and 3%/2 mm criteria with a 10% threshold. The training set (TS) of ML models consisted of 1106 beams, and an independent evaluation set (ES) consisted of 277 beams. For each beam, 33 complexity metrics were extracted as input data for training models. Three ML algorithms (gradient boosting decision tree/GBDT, random forest/RF, and Poisson Lasso/PL) were utilized to build the models and predict gamma passing rates (GPRs). To improve the prediction accuracy in the rare region, a method of reweighting for TS has been performed and compared to the unweighted results. The importance of complexity metrics was studied by permuted interesting features.

RESULTS

The GBDT model had the best performance in this study. In ES, the minimal mean prediction error for unweighted results was 1.93%, 1.16%, 0.78% under 1%/1, 2%/2, and 3%/2 mm criteria, respectively, from GBDT model. Comparing to the unweighted results, the models after reweighting gained up to 30% improvement in the rare region, whereas the overall prediction error was slightly worse depending on the kind of models. For feature importance, 2 tree-based models (GBDT and RF) had in common the top 10 most important metrics as well as the same metric with the largest impact.

CONCLUSION

For linac equipped with novel dual-layered MLC, the ML model based on GBDT algorithm shows a certain degree of accuracy for GPRs prediction. The specific ML model for dual-layered MLC configuration could be a useful tool for physicists detecting PSQA measurement failures.

A synthesized gamma distribution-based patient-specific VMAT QA using a generative adversarial network

BACKGROUND

Artificial intelligence (AI)-based gamma passing rate (GPR) prediction has been proposed as a time-efficient virtual patient-specific QA method for the delivery of volumetric modulation arc therapy (VMAT). However, there is a limitation that the GPR value loses the locational information of dose accuracy.

PURPOSE

The objective was to predict the failing points in the gamma distribution and the GPR using a synthesized gamma distribution of VMAT QA with a deep convolutional generative adversarial network (GAN).

METHODS

The fluence maps of 270 VMAT beams for prostate cancer were measured using an electronic portal imaging device and analyzed using gamma evaluation with 3%/2-mm, 2%/1-mm, 1%/1-mm, and 1%/0.5-mm tolerances. The 270 gamma distributions were divided into two datasets: 240 training datasets for creating a model and 30 test datasets for evaluation. The image prediction network for the fluence maps calculated by the treatment planning system (TPS) to the gamma distributions was created using a GAN. The sensitivity, specificity, and accuracy of detecting failing points were evaluated using measured and synthesized gamma distributions. In addition, the difference between measured GPR (mGPR) and predicted GPR (pGPR) values calculated from the synthesized gamma distributions was evaluated.

RESULTS

The root mean squared errors between mGPR and pGPR were 1.0%, 2.1%, 3.5%, and 3.6% for the 3%/2-mm, 2%/1-mm, 1%/1-mm, and 1%/0.5-mm tolerances, respectively. The accuracies for detecting failing points were 98.9%, 96.9%, 94.7%, and 93.7% for 3%/2-mm, 2%/1-mm, 1%/1-mm, and 1%/0.5-mm tolerances, respectively. The sensitivity and specificity were the highest for 1%/0.5-mm and 3%/2-mm tolerances, which were 82.7% and 99.6%, respectively.

CONCLUSIONS

We developed a novel system using a GAN to generate a synthesized gamma distribution-based patient-specific VMAT QA. The system is promising from the point of view of quality assurance in radiotherapy because it shows high performance and can detect failing points.

Machine learning-based predictions of gamma passing rates for virtual specific-plan verification based on modulation maps, monitor unit profiles, and composite dose images

Machine learning (ML) methods have been implemented in radiotherapy to aid virtual specific-plan verification protocols, predicting gamma passing rates (GPR) based on calculated modulation complexity metrics because of their direct relation to dose deliverability. Nevertheless, these metrics might not comprehensively represent the modulation complexity, and automatically extracted features from alternative predictors associated with modulation complexity are needed. For this reason, three convolutional neural networks (CNN) based models were trained to predict GPR values (regression and classification), using respectively three predictors: (1) the modulation maps (MM) from the multi-leaf collimator, (2) the relative monitor units per control point profile (MUcp), and (3) the composite dose image (CDI) used for portal dosimetry, from 1024 anonymized prostate plans. The models' performance was assessed for classification and regression by the area under the receiver operator characteristic curve (AUC_ROC) and Spearman's correlation coefficient (r). Finally, four hybrid models were designed using all possible combinations of the three predictors. The prediction performance for the CNN-models using single predictors (MM, MUcp, and CDI) were AUC_ROC = 0.84 ± 0.03, 0.77 ± 0.07, 0.75 ± 0.04, andr= 0.6, 0.5, 0.7. Contrastingly, the hybrid models (MM + MUcp, MM + CDI, MUcp+CDI, MM + MUcp+CDI) performance were AUC_ROC = 0.94 ± 0.03, 0.85 ± 0.06, 0.89 ± 0.06, 0.91 ± 0.03, andr= 0.7, 0.5, 0.6, 0.7. The MP, MUcp, and CDI are suitable predictors for dose deliverability models implementing ML methods. Additionally, hybrid models are susceptible to improving their prediction performance, including two or more input predictors.

Predictive gamma passing rate of 3D detector array-based volumetric modulated arc therapy quality assurance for prostate cancer via deep learning

To predict the gamma passing rate (GPR) of the three-dimensional (3D) detector array-based volumetric modulated arc therapy (VMAT) quality assurance (QA) for prostate cancer using a convolutional neural network (CNN) with the 3D dose distribution. One hundred thirty-five VMAT plans for prostate cancer were selected: 110 plans were used for training and validation, and 25 plans were used for testing. Verification plans were measured using a helical 3D diode array (ArcCHECK). The dose distribution on the detector element plane of these verification plans was used as input data for the CNN model. The measured GPR (mGPR) values were used as the training data. The CNN model comprises eighteen layers and predicted GPR (pGPR) values. The mGPR and pGPR values were compared, and a cumulative frequency histogram of the prediction error was created to clarify the prediction error tendency. The correlation coefficients of pGPR and mGPR were 0.67, 0.69, 0.66, and 0.73 for 3%/3-mm, 3%/2-mm, 2%/3-mm, and 2%/2-mm gamma criteria, respectively. The respective mean±standard deviations of pGPR-mGPR were -0.87±2.18%, -0.65±2.93%, -0.44±2.53%, and -0.71±3.33%. The probabilities of false positive error cases (pGPR < mGPR) were 72%, 60%, 68%, and 56% for each gamma criterion. We developed a deep learning-based prediction model of the 3D detector array-based VMAT QA for prostate cancer, and evaluated the accuracy and tendency of prediction GPR. This model can provide a proactive estimation for the results of the patient-specific QA before the verification measurement.

A brain-like classification method for computed tomography images based on adaptive feature matching dual-source domain heterogeneous transfer learning

Transfer learning can improve the robustness of deep learning in the case of small samples. However, when the semantic difference between the source domain data and the target domain data is large, transfer learning easily introduces redundant features and leads to negative transfer. According the mechanism of the human brain focusing on effective features while ignoring redundant features in recognition tasks, a brain-like classification method based on adaptive feature matching dual-source domain heterogeneous transfer learning is proposed for the preoperative aided diagnosis of lung granuloma and lung adenocarcinoma for patients with solitary pulmonary solid nodule in the case of small samples. The method includes two parts: (1) feature extraction and (2) feature classification. In the feature extraction part, first, By simulating the feature selection mechanism of the human brain in the process of drawing inferences about other cases from one instance, an adaptive selected-based dual-source domain feature matching network is proposed to determine the matching weight of each pair of feature maps and each pair of convolution layers between the two source networks and the target network, respectively. These two weights can, respectively, adaptive select the features in the source network that are conducive to the learning of the target task, and the destination of feature transfer to improve the robustness of the target network. Meanwhile, a target network based on diverse branch block is proposed, which made the target network have different receptive fields and complex paths to further improve the feature expression ability of the target network. Second, the convolution kernel of the target network is used as the feature extractor to extract features. In the feature classification part, an ensemble classifier based on sparse Bayesian extreme learning machine is proposed that can automatically decide how to combine the output of base classifiers to improve the classification performance. Finally, the experimental results (the AUCs were 0.9542 and 0.9356, respectively) on the data of two center data show that this method can provide a better diagnostic reference for doctors.

What is the optimal input information for deep learning-based pre-treatment error identification in radiotherapy?

•

The choice of dose comparison method impacts deep learning error identification accuracy most.

•

Simple dose comparison methods are more beneficial than gamma analysis and alternative methods.

•

Mean/standard deviation normalization and high image resolution improve error identification.

Background and purpose

Deep learning (DL) provides high sensitivity for detecting and identifying errors in pre-treatment radiotherapy quality assurance (QA). This work’s objective was to systematically evaluate the impact of different dose comparison and image preprocessing methods on DL model performance for error identification in pre-treatment QA.

Materials and methods

For 53 volumetric modulated arc therapy (VMAT) and 69 stereotactic body radiotherapy (SBRT) treatment plans of lung cancer patients, mechanical errors were simulated (MLC leaf positions, monitor unit scaling, collimator rotation). Two classification levels were assessed: error type (Level 1) and error magnitude (Level 2). Portal dose images with and without errors were compared using standard (gamma analysis), simple (absolute/relative dose difference, ratio) and alternative (distance-to-agreement, structural similarity index, gradient) dose comparison methods. For preprocessing, different normalization methods (min/max and mean/standard deviation) and image resolutions (32 × 32, 64 × 64 and 128 × 128) were evaluated. All possible combinations of classification level, dose comparison, normalization method and image size resulted in 144 input datasets for DL networks for error identification.

Results

Average accuracy was highest for simple dose comparison methods (Level 1: 97.7%, Level 2: 78.1%) while alternative methods scored lowest (Level 1: 91.6%, Level 2: 71.2%). Mean/stdev normalization particularly improved Level 2 classification. Higher image resolution improved error identification, although for SBRT lower image resolution was also sufficient.

Conclusions

The choice of dose comparison method has the largest impact on error identification for pre-treatment QA using DL, compared to image preprocessing. Model performance can improve by using simple dose comparison methods, mean/stdev normalization and high image resolution.

Improvement of IMRT QA prediction using imaging-based neural architecture search

PURPOSE

Machine learning has been used to predict the gamma passing rate of Intensity-modulated radiation therapy (IMRT) QA results. In this work, we applied a novel neural architecture search to automatically tune and search for the best deep neural networks instead of using hand-designed deep learning architectures.

METHOD AND MATERIALS

One hundred and eighty-two IMRT plans were created and delivered with portal dosimetry. A total of 1497 fields for multiple treatment sites were delivered and measured by portal imagers. Gamma criteria of 2%/2mm with a 5% threshold were used. Fluence maps calculated for each plan were used as inputs to a convolution neural network (CNN). Auto-Keras was implemented to search for the best CNN architecture for fluence image regression. The network morphism was adopted in the searching process, in which the base models were ResNet and DenseNet. The performance of this CNN approach was compared with tree-based machine learning models previously developed for this application, using the same data set.

RESULTS

The deep-learning-based approach had 98.3% of predictions within 3% of the measured 2%/2mm gamma passing rates with a maximum error of 3.1% and a mean absolute error of less than 1%. Our results show that this novel architecture search approach achieves comparable performance to the machine-learning-based approaches with handcrafted features.

CONCLUSIONS

We implemented a novel CNN model using imaging-based neural architecture for IMRT QA prediction. The imaging-based deep-learning method does not require manual extraction of relevant features and is able to automatically select the best network architecture. This article is protected by copyright. All rights reserved.

Prospective clinical validation of virtual patient-specific quality assurance of VMAT radiation therapy plans

PURPOSE

Performing measurement-based patient-specific quality assurance (PSQA) is recognized as a resource-intensive and time inefficient task in the radiotherapy treatment workflow. Paired with technological refinements in modern radiotherapy, research towards measurement-free PSQA has seen increased interest over the last five years. However, these efforts have not been clinically implemented or prospectively validated in the U.S. We propose a virtual QA (VQA) system and workflow to assess the safety and workload reduction of measurement-free PSQA.

METHODS

An XGBoost machine learning model was designed to predict PSQA outcomes of VMAT plans, represented as percent differences between the measured ion chamber point dose in a phantom and the corresponding planned dose. The final model was deployed within a web application to predict PSQA outcomes of clinical plans within an existing clinical workflow. The application also displays relevant feature importance and plan-specific distribution analyses relative to database plans for documentation and to aid physicist interpretation and evaluation. VQA predictions were prospectively validated over three months of measurements at our clinic to assess safety and efficiency gains.

RESULTS

Over three months, VQA predictions for 445 VMAT plans were prospectively validated at our institution. VQA predictions for these plans had a mean absolute error of 1.08 +/- 0.77%, with a maximum absolute error of 2.98%. Employing a 1% prediction threshold (i.e. plans predicted to have an absolute error of less than 1% would not require a measurement) would yield a 69.2% reduction in QA workload - saving 32.5 hours per month on average - with 81.5%/72.4%/0.81 sensitivity/specificity/AUC at a 3% clinical threshold and 100%/70%/0.93 sensitivity/specificity/AUC at a 4% clinical threshold.

CONCLUSION

This is the first prospective clinical implementation and validation of VQA in the U.S., which we observed to be efficient. Using a conservative threshold, VQA can substantially reduce the number of required measurements for PSQA, leading to more effective allocation of clinical resources.

Deep Learning for Patient-Specific Quality Assurance: Predicting Gamma Passing Rates for IMRT Based on Delivery Fluence Informed by log Files

Objectives: In this study, we propose a deep learning-based approach to predict Intensity-modulated radiation therapy (IMRT) quality assurance (QA) gamma passing rates using delivery fluence informed by log files. Methods: A total of 112 IMRT plans for chest cancers were planned and measured by portal dosimetry equipped on TrueBeam linac. The convolutional neural network (CNN) based learning model was trained using delivery fluence as inputs and gamma passing rates (GPRs) of 4 different criteria (3%/3 mm, 2%/3 mm, 3%/2 mm, and 2%/2 mm) as outputs. Model performance for both validation and test sets was assessed using mean absolute error (MAE), mean squared error (MSE), root MSE (RMSE), Spearman rank correlation coefficients (Sr), and Determination coefficient (R²) between the measured and predicted GPR values. Results: In the test set, the MAE of the prediction model were 0.402, 0.511, 1.724, and 2.530, the MSE were 0.640, 0.986, 6.654, and 9.508, the RMSE were 0.800, 0.993, 2.580, and 3.083, the Sr were 0.643, 0.684, 0.821, and 0.824 (P < .001) and the R² were 0.4110, 0.4666, 0.6677, and 0.6769 for 3%/3 mm, 3%/2 mm, 2%/3 mm, and 2%/2 mm, respectively. The MAE and RMSE of the prediction model decreased with stricter gamma criteria while the Sr and R² between measured and predicted GPR values increased. Conclusions: The CNN prediction model based on delivery fluence informed by log files could accurately predict IMRT QA passing rates for different gamma criteria. It could reduce QA workload and improve efficiency in pretreatment QA. Our results suggest that the CNN prediction model based on delivery fluence informed by log files may be a promising tool for the gamma evaluation of IMRT QA.

Deep Multimodal Neural Network Based on Data-Feature Fusion for Patient-Specific Quality Assurance

Patient-specific quality assurance (QA) for Volumetric Modulated Arc Therapy (VMAT) plans is routinely performed in the clinical. However, it is labor-intensive and time-consuming for medical physicists. QA prediction models can address these shortcomings and improve efficiency. Current approaches mainly focus on single cancer and single modality data. They are not applicable to clinical practice. To assess the accuracy of QA results for VMAT plans, this paper presents a new model that learns complementary features from the multi-modal data to predict the gamma passing rate (GPR). According to the characteristics of VMAT plans, a feature-data fusion approach is designed to fuse the features of imaging and non-imaging information in the model. In this study, 690 VMAT plans are collected encompassing more than ten diseases. The model can accurately predict the most VMAT plans at all three gamma criteria: 2%/2 mm, 3%/2 mm and 3%/3 mm. The mean absolute error between the predicted and measured GPR is 2.17%, 1.16% and 0.71%, respectively. The maximum deviation between the predicted and measured GPR is 3.46%, 4.6%, 8.56%, respectively. The proposed model is effective, and the features of the two modalities significantly influence QA results.

Improving the Quality of Care in Radiation Oncology using Artificial Intelligence

Radiation therapy is a complex process involving multiple professionals and steps from simulation to treatment planning to delivery, and these procedures are prone to error. Additionally, the imaging and treatment delivery equipment in radiotherapy is highly complex and interconnected and represents another risk point in the quality of care. Numerous quality assurance tasks are carried out to ensure quality and to detect and prevent potential errors in the process of care. Recent developments in artificial intelligence provide potential tools to the radiation oncology community to improve the efficiency and performance of quality assurance efforts. Targets for artificial intelligence enhancement include the quality assurance of treatment plans, target and tissue structure delineation used in the plans, delivery of the plans and the radiotherapy delivery equipment itself. Here we review recent developments of artificial intelligence applications that aim to improve quality assurance processes in radiation therapy and discuss some of the challenges and limitations that require further development work to realise the potential of artificial intelligence for quality assurance.

Evaluation of prediction and classification performances in different machine learning models for patient-specific quality assurance of head-and-neck VMAT plans

PURPOSE

The purpose of this study is to evaluate the prediction and classification performances of the gamma passing rate (GPR) for different machine learning models and to select the best model for achieving machine learning-based patient-specific quality assurance (PSQA).

METHODS

The measurement verification of 356 head-and-neck volumetric modulated arc therapy plans was performed using a diode array phantom (Delta4 Phantom), and GPR values at 2%/2 mm with global normalization and 3%/2 mm with local normalization were calculated. Machine learning models, including ridge regression (RIDGE), random forest (RF), support vector regression (SVR), and stacked generalization (STACKING), were used to predict the GPR. Each machine learning model was trained using 260 plans, and the prediction accuracy was evaluated using the remaining 96 plans. The prediction error between the measured and predicted GPR was evaluated. For the classification evaluation, the lower control limit for the measured GPR and lower control limit for predicted GPR (LCL_p ) was defined to identify whether the GPR values represent a "pass" or a "fail." LCL_p values with 99% and 99.9% confidence levels were calculated as the upper prediction limits for the GPR estimated from the linear regression between the measured and predicted GPR.

RESULTS

There was an overestimation trend of the low measured GPR. The maximum prediction errors for RIDGE, RF, SVR, and STACKING were 3.2%, 2.9%, 2.3%, and 2.2% at the global 2%/2 mm and 6.3%, 6.6%, 6.1%, and 5.5% at the local 3%/2 mm, respectively. In the global 2%/2 mm, the sensitivity was 100% for all the machine learning models except RIDGE when using 99% LCL_p . The specificity was 76.1% for RIDGE, RF, and SVR and 66.3% for STACKING; however, the specificity decreased dramatically when 99.9% LCL_p was used. In the local 3%/2 mm, however, only STACKING showed 100% sensitivity when using 99% LCL_p . The decrease in the specificity using 99.9% LCL_p was smaller than that in the global 2%/2 mm, and the specificity for RIDGE, RF, SVR, and STACKING was 61.3%, 61.3%, 72.0%, and 66.8%, respectively.

CONCLUSIONS

STACKING had better prediction accuracy for low GPR values than other machine learning models. Applying LCL_p to a regression model enabled the consistent evaluation of quantitative and qualitative GPR predictions. Adjusting the confidence level of the LCL_p helped improve the balance between the sensitivity and specificity. We suggest that STACKING can assist the safe and efficient operation of PSQA.

Improvement Using Planomics Features on Prediction and Classification of Patient-Specific Quality Assurance Using Head and Neck Volumetric Modulated Arc Therapy Plan

Purpose: This study aimed to evaluate the utility of a new plan feature (planomics feature) for predicting the results of patient-specific quality assurance using the head and neck (H&N) volumetric modulated arc therapy (VMAT) plan. Methods: One hundred and thirty-one H&N VMAT plans in our institution from 2019 to 2021 were retrospectively collected. Dosimetric verification for all plans was carried out using the portal dosimetry system integrated into the Eclipse treatment planning system based on the electronic portal imaging devices. Gamma passing rates (GPR) were analyzed using three gamma indices of 3%/3 mm, 3%/2 mm, and 2%/2 mm with a 10% dose threshold. Forty-eight conventional features affecting the dose delivery accuracy were used in the study, and 2,476 planomics features were extracted based on the radiotherapy plan file. Three prediction and classification models using conventional features (CF), planomics features (PF), and hybrid features (HF) combining two sets of features were constructed by the gradient boosting regressor (GBR) and Ridge classifier for each GPR of 3%/3 mm, 3%/2 mm, and 2%/2 mm, respectively. The absolute prediction error (APE) and the area under the curve (AUC) were adopted for assessing the performance of prediction and classification models. Results: In the GPR prediction, the average APE of the models using CF, PF, and HF was 1.3 ± 1.2%/3.6 ± 3.0%, 1.7 ± 1.5%/3.8 ± 3.5%, and 1.1 ± 1.0%/4.1 ± 3.1% for 2%/2 mm; 0.7 ± 0.6%/2.0 ± 2.0%, 1.0±1.1%/2.2 ± 1.8%, and 0.6 ± 0.6%/2.2 ± 1.9% for 3%/2 mm; and 0.4 ± 0.3%/1.2 ± 1.2%, 0.4±0.5%/1.3 ± 1.0%, and 0.3±0.3%/1.2 ± 1.1% for 3%/3 mm, respectively. In the regression prediction, three models give a similar modeling performance for predicting the GPR. The classification results were 0.67 ± 0.03/0.66 ± 0.07, 0.77 ± 0.03/0.73 ± 0.06, and 0.78 ± 0.02/0.75 ± 0.04 for 3%/3 mm, respectively. For 3%/2 mm, the AUCs of the training and testing cohorts were 0.64 ± 0.03/0.62 ± 0.07, 0.70 ± 0.03/0.67 ± 0.06, and 0.75 ± 0.03/0.71 ± 0.07, respectively, and for 2%/2 mm, the average AUCs of the training and testing cohorts were 0.72 ± 0.03/0.72 ± 0.06, 0.78 ± 0.04/0.73 ± 0.07, and 0.81 ± 0.03/0.75 ± 0.06, respectively. In the classification, the PF model has a better classification performance than the CF model. Moreover, the HF model provides the best result among the three classifications models. Conclusions: The planomics features can be used for predicting and classifying the GPR results and for improving the model performance after combining the conventional features for the GPR classification.

Applications of machine and deep learning to patient-specific IMRT/VMAT quality assurance

In order to deliver accurate and safe treatment to cancer patients in radiation therapy using advanced techniques such as intensity modulated radiation therapy (IMRT) and volumetric-arc radiation therapy (VMAT), patient specific quality assurance (QA) should be performed before treatment. IMRT/VMAT dose measurements in a phantom using various devices have been clinically adopted as standard method for QA. This approach allows the verification of the accuracy of the dose calculation, data transfer, and the delivery system. However, patient-specific QA procedures are expensive and require significant time and effort by the physicists. Over the past 5 years, machine learning (ML) and deep learning (DL) algorithms for predictions of IMRT/VMAT QA outcome have been investigated. Various ML and DL models have shown promising prediction accuracy and a high potential as time-efficient virtual QA tool. In this paper, we review the ML and DL based models that were developed for patient specific IMRT and VMAT QA outcome predictions from algorithmic and clinical applicability perspectives. We focus on comparing the algorithms, the dataset sizes, the input parameters and features, the QA outcome prediction approaches, the validation, the performance, the clinical applicability, and the potential clinical impact. In addition, we discuss the present challenges as well as the future directions in the implementation of these models. To the best of our knowledge, this is the first review on the application of ML and DL based models in IMRT/VMAT QA predictions.

Virtual Patient-Specific Quality Assurance of IMRT Using UNet++: Classification, Gamma Passing Rates Prediction, and Dose Difference Prediction

The dose verification in radiotherapy quality assurance (QA) is time-consuming and places a heavy workload on medical physicists. To provide a clinical tool to perform patient specific QA accurately, the UNet++ is investigated to classify failed or pass fields (the GPR lower than 85% is considered “failed” while the GPR higher than 85% is considered “pass”), predict gamma passing rates (GPR) for different gamma criteria, and predict dose difference from virtual patient-specific quality assurance in radiotherapy. UNet++ was trained and validated with 473 fields and tested with 95 fields. All plans used Portal Dosimetry for dose verification pre-treatment. Planar dose distribution of each field was used as the input for UNet++, with QA classification results, gamma passing rates of different gamma criteria, and dose difference were used as the output. In the test set, the accuracy of the classification model was 95.79%. The mean absolute error (MAE) were 0.82, 0.88, 2.11, 2.52, and the root mean squared error (RMSE) were 1.38, 1.57, 3.33, 3.72 for 3%/3mm, 3%/2 mm, 2%/3 mm, 2%/2 mm, respectively. The trend and position of the predicted dose difference were consistent with the measured dose difference. In conclusion, the Virtual QA based on UNet++ can be used to classify the field passed or not, predict gamma pass rate for different gamma criteria, and predict dose difference. The results show that UNet++ based Virtual QA is promising in quality assurance for radiotherapy.

Artificial intelligence for quality assurance in radiotherapy

In radiotherapy, patient-specific quality assurance is very time-consuming and causes machine downtime. It consists of testing (using measurement with a phantom and detector) if a modulated plan is correctly delivered by a treatment unit. Artificial intelligence and in particular machine learning algorithms were mentioned in recent reports as promising solutions to reduce or eliminate the patient-specific quality assurance workload. Several teams successfully experienced a virtual patient-specific quality assurance by training a machine learning tool to predict the results. Training data are generally composed of previous treatment plans and associated patient-specific quality assurance results. However, other training data types were recently introduced such as actual positions and velocities of multileaf collimators, metrics of the plan's complexity, and gravity vectors. Different types of machine learning algorithms were investigated (Poisson regression algorithms, convolutional neural networks, support vector classifiers) with sometimes promising results. These tools are being used for treatment units' quality assurance as well, in particular to analyse the results of imaging devices. Most of these reports were feasibility studies. Using machine learning in clinical routines as a tool that could fully replace quality assurance tests conducted by physics teams has yet to be implemented.

Commissioning and clinical implementation of an Autoencoder based Classification-Regression model for VMAT patient-specific QA in a multi-institution scenario

BACKGROUND AND PURPOSE

To commission and implement an Autoencoder based Classification-Regression (ACLR) model for VMAT patient-specific quality assurance (PSQA) in a multi-institution scenario.

MATERIALS AND METHODS

1835 VMAT plans from seven institutions were collected for the ACLR model commissioning and multi-institutional validation. We established three scenarios to validate the gamma passing rates (GPRs) prediction and classification accuracy with the ACLR model for different delivery equipment, QA devices, and treatment planning systems (TPS). The prediction performance of the ACLR model was evaluated using mean absolute error (MAE) and root mean square error (RMSE). The classification performance was evaluated using sensitivity and specificity. An independent end-to-end test (E2E) and routine QA of the ACLR model were performed to validate the clinical use of the model.

RESULTS

For multi-institution validations, the MAEs were 1.30-2.80% and 2.42-4.60% at 3%/3 mm and 3%/2 mm, respectively, and RMSEs were 1.55-2.98% and 2.83-4.95% at 3%/3 mm and 3%/2 mm, respectively, with different delivery equipment, QA devices, and TPS, while the sensitivity was 90% and specificity was 70.1% at 3%/2 mm. For the E2E, the deviations between the predicted and measured results were within 3%, and the model passed the consistency check for clinical implementation. The predicted results of the model were the same in daily QA, while the deviations between the repeated monthly measured GPRs were all within 2%.

CONCLUSIONS

The performance of the ACLR model in multi-institution scenarios was validated on a large scale. Routine QA of the ACLR model was established and the model could be used for VMAT PSQA clinically.

Error detection model developed using a multi-task convolutional neural network in patient-specific quality assurance for volumetric-modulated arc therapy

PURPOSE

In patient-specific quality assurance (QA) for static beam intensity-modulated radiation therapy (IMRT), machine-learning-based dose analysis methods have been developed to identify the cause of an error as an alternative to gamma analysis. Although these new methods have revealed that the cause of the error can be identified by analyzing the dose distribution obtained from the two-dimensional detector, they have not been extended to the analysis of volumetric-modulated arc therapy (VMAT) QA. In this study, we propose a deep learning approach to detect various types of errors in patient-specific VMAT QA.

METHODS

A total of 161 beams from 104 prostate VMAT plans were analyzed. All beams were measured using a cylindrical detector (Delta4; ScandiDos, Uppsala, Sweden), and predicted dose distributions in a cylindrical phantom were calculated using a treatment planning system (TPS). In addition to the error-free plan, we simulated 12 types of errors: two types of multileaf collimator positional errors (systematic or random leaf offset of 2 mm), two types of monitor unit (MU) scaling errors (±3%), two types of gantry rotation errors (±2° in clockwise and counterclockwise direction), and six types of phantom setup errors (±1 mm in lateral, longitudinal, and vertical directions). The error-introduced predicted dose distributions were created by editing the calculated dose distributions using a TPS with in-house software. Those 13 types of dose difference maps, consisting of an error-free map and 12 error maps, were created from the measured and predicted dose distributions and were used to train the convolutional neural network (CNN) model. Our model was a multi-task model that individually detected each of the 12 types of errors. Two datasets, Test sets 1 and 2, were prepared to evaluate the performance of the model. Test set 1 consisted of 13 types of dose maps used for training, whereas Test set 2 included the dose maps with 25 types of errors in addition to the error-free dose map. The dose map, which introduced 25 types of errors, was generated by combining two of the 12 types of simulated errors. For comparison with the performance of our model, gamma analysis was performed for Test sets 1 and 2 with the criteria set to 3%/2 mm and 2%/1 mm (dose difference/distance to agreement).

RESULTS

For Test set 1, the overall accuracy of our CNN model, gamma analysis with the criteria set to 3%/2 mm, and gamma analysis with the criteria set to 2%/1 mm was 0.92, 0.19, and 0.81, respectively. Similarly, for Test set 2, the overall accuracy was 0.44, 0.42, and 0.95, respectively. Our model outperformed gamma analysis in the classification of dose maps containing a single type error, and the performance of our model was inferior in the classification of dose maps containing compound errors.

CONCLUSIONS

A multi-task CNN model for detecting errors in patient-specific VMAT QA using a cylindrical measuring device was constructed, and its performance was evaluated. Our results demonstrate that our model was effective in identifying the error type in the dose map for VMAT QA.