Error detection for radiotherapy planning validation based on deep learning networks.

BACKGROUND: Quality assurance (QA) of patient-specific treatment plans for intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) necessitates prior validation. However, the standard methodology exhibits deficiencies and lacks sensitivity in the analysis of positional dose distribution data, leading to difficulties in accurately identifying reasons for plan verification failure. This issue complicates and impedes the efficiency of QA tasks. PURPOSE: The primary aim of this research is to utilize deep learning algorithms for the extraction of 3D dose distribution maps and the creation of a predictive model for error classification across multiple machine models, treatment methodologies, and tumor locations. METHOD: We devised five categories of validation plans (normal, gantry error, collimator error, couch error, and dose error), conforming to tolerance limits of different accuracy levels and employing 3D dose distribution data from a sample of 94 tumor patients. A CNN model was then constructed to predict the diverse error types, with predictions compared against the gamma pass rate (GPR) standard employing distinct thresholds (3%, 3 mm; 3%, 2 mm; 2%, 2 mm) to evaluate the model's performance. Furthermore, we appraised the model's robustness by assessing its functionality across diverse accelerators. RESULTS: The accuracy, precision, recall, and F1 scores of CNN model performance were 0.907, 0.925, 0.907, and 0.908, respectively. Meanwhile, the performance on another device is 0.900, 0.918, 0.900, and 0.898. In addition, compared to the GPR method, the CNN model achieved better results in predicting different types of errors. CONCLUSION: When juxtaposed with the GPR methodology, the CNN model exhibits superior predictive capability for classification in the validation of the radiation therapy plan on different devices. By using this model, the plan validation failures can be detected more rapidly and efficiently, minimizing the time required for QA tasks and serving as a valuable adjunct to overcome the constraints of the GPR method.

Physical dose validation of dynamic treatment for Gamma Knife radiosurgery.

BACKGROUND: Dynamic treatment in Gamma Knife (GK) radiosurgery systems delivers radiation continuously with couch movement, as opposed to stationary step-and-shoot treatment where radiation is paused when moving between isocenters. Previous studies have shown the potential for dynamic GK treatment to give faster treatment times and improved dose conformity and homogeneity. However, these studies focused only on computational simulations and lack physical validation. PURPOSE: This study aims conduct dynamic treatment dosimetric validation with physical experimental measurements. The experiments aim to (1) address assumptions made with computational studies, such as the validity of treating a continuous path as discretised points, (2) investigate uncertainties in translating computed plans to actual treatment, and (3) determine ideal treatment planning parameters, such as interval distance for the path discretization, collimator change limitations, and minimum isocenter treatment times. METHODS: This study uses a GK ICON treatment delivery machine, and a motion phantom custom-made to attach to the machine's mask adapter and move in 1D superior-inferior motion. Phantom positioning is first verified through comparisons against couch motion and computed doses. For dynamic treatment experiments, the phantom is moved through a program that first reads the desired treatment plan isocenters' position, time, and collimator sizes, then carries out the motion continuously while the treatment machine delivers radiation. Measurements are done with increasing levels of complexity: varying speed, varying collimator sizes, varying both speed and collimator sizes, then extends the same measurements to simulated 2D motion by combining phantom and couch motion. Dose comparisons between phantom motion radiation measurements and either couch motion measurements or dose calculations are analyzed with 2 mm/2% and 1 mm/2% gamma indices, using both local and global gamma index calculations. RESULTS: Phantom positional experiments show a high accuracy, with global gamma indices for all dose comparisons ≥ $\ge $ 99%. Discretization level to approximate continuous path as discrete points show the good dose matches with dose calculations when using 1 and 2-mm gaps. Complex 1D motion, including varying speed, collimator sizes, or both, as well as 2D motion with the same complexities, all show good dose matches with dose calculations: the scores are ≥ $\ge $ 92.0% for the strictest 1 mm/2% local gamma index calculation, ≥ $\ge $ 99.8% for 2 mm/2% local gamma index, and ≥ $\ge $ 97.0% for all global gamma indices. Five simulated 2D treatments with optimized plans scored highly as well, with all gamma index scores ≥ $\ge $ 95.3% when compared to stationary treatment, and scores ≥ $\ge $ 97.9% when compared to plan calculated dose. CONCLUSIONS: Dynamic treatment computational studies are validated, with dynamic treatment shown to be physically feasible and deliverable with high accuracy. A 2-mm discretization level in treatment planning is proposed as the best option for shorter dose calculation times while maintaining dose accuracy. Our experimental method enables dynamic treatment measurements using the existing clinical workflow, which may be replicated in other centers, and future studies may include 2D or 3D motion experiments, or planning studies to further quantify potential indication-specific benefits.

Validation of a Fully Automated Hybrid Deep Learning Cardiac Substructure Segmentation Tool for Contouring and Dose Evaluation in Lung Cancer Radiotherapy.

BACKGROUND AND PURPOSE: Accurate and consistent delineation of cardiac substructures is challenging. The aim of this work was to validate a novel segmentation tool for automatic delineation of cardiac structures and subsequent dose evaluation, with potential application in clinical settings and large-scale radiation-related cardiotoxicity studies. MATERIALS AND METHODS: A recently developed hybrid method for automatic segmentation of 18 cardiac structures, combining deep learning, multi-atlas mapping and geometric segmentation of small challenging substructures, was independently validated on 30 lung cancer cases. These included anatomical and imaging variations, such as tumour abutting heart, lung collapse and metal artefacts. Automatic segmentations were compared with manual contours of the 18 structures using quantitative metrics, including Dice similarity coefficient (DSC), mean distance to agreement (MDA) and dose comparisons. RESULTS: A comparison of manual and automatic contours across all cases showed a median DSC of 0.75-0.93 and a median MDA of 2.09-3.34 mm for whole heart and chambers. The median MDA for great vessels, coronary arteries, cardiac valves, sinoatrial and atrioventricular conduction nodes was 3.01-8.54 mm. For the 27 cases treated with curative intent (planned target volume dose ≥50 Gy), the median dose difference was -1.12 to 0.57 Gy (absolute difference of 1.13-3.25%) for the mean dose to heart and chambers; and -2.25 to 4.45 Gy (absolute difference of 0.94-6.79%) for the mean dose to substructures. CONCLUSION: The novel hybrid automatic segmentation tool reported high accuracy and consistency over a validation set with challenging anatomical and imaging variations. This has promising applications in substructure dose calculations of large-scale datasets and for future studies on long-term cardiac toxicity.

Radiomics-based prediction of gamma pass rates for different intensity-modulated radiation therapy techniques for pelvic tumors / 中华放射医学与防护杂志

Objective:To explore the feasibility of a classification prediction model for gamma pass rates (GPRs) under different intensity-modulated radiation therapy techniques for pelvic tumors using a radiomics-based machine learning approach, and compare the classification performance of four integrated tree models.Methods:With a retrospective collection of 409 plans using different IMRT techniques, the three-dimensional dose validation results were adopted based on modality measurements, with a GPR criterion of 3%/2 mm and 10% dose threshold. Then prediction were built models by extracting radiomics features based on dose documentation. Four machine learning algorithms were used, namely random forest (RF), adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM). Their classification performance was evaluated by calculating sensitivity, specificity, F1 score, and AUC value. Results:The RF, AdaBoost, XGBoost, and LightGBM models had sensitivities of 0.96, 0.82, 0.93, and 0.89, specificities of 0.38, 0.54, 0.62, and 0.62, F1 scores of 0.86, 0.81, 0.88, and 0.86, and AUC values of 0.81, 0.77, 0.85, and 0.83, respectively. XGBoost model showed the highest sensitivity, specificity, F1 score, and AUC value, outperforming the other three models. Conclusions:To build a GPR classification prediction model using a radiomics-based machine learning approach is feasible for plans using different intensity-modulated radiotherapy techniques for pelvic tumors, providing a basis for future multi-institutional collaborative research on GPR prediction.

Evaluation of Inhomogeneity Correction Performed by Radiotherapy Treatment Planning System.

BACKGROUND: Aim of this study is to evaluate the efficacy of inhomogeneity corrections calculated by radiotherapy treatment planning system (TPS) using various densities of materials. MATERIALS AND METHODS: Gammex Computed tomography electron density inserts (EDI's; 14 no's) were used to generate the CT to ED curve with high speed GE CT scanner by noting down the respective HU values of each rod. Treatment plans were generated in XiO TPS with three inhomogeneous phantoms (comprising combination of water, lung and bone equivalent slabs) with different field sizes and for EDI (8 no's) inserted in slots of acrylic tray and validation was carried out using 2D array detector with 20cm×20cm field size for 200 MU. Point dose and fluence measurements were carried with inhomogeneous phantoms combinations and EDI's (placed on the locally fabricated box filled with water medium). RESULTS: The mean percentage deviations with standard deviation of calculated point doses against measured ones obtained with 2D array detector at iso-center plane for all three inhomogeneous phantom combinations were found to be -1.13%±0.13%, -3.51%±0.14% and -0.63%±0.27% respectively. On point doses measured under each individual EDI, over all percentage deviation with standard deviation observed is -2.04% ± 1.1%. CONCLUSION: The described method can be implemented in any newly established radiotherapy department as a routine quality measure of TPS to verify its efficacy in performing of inhomogeneity calculation.

Quantitative comparison of pre-treatment predictive and post-treatment measured dosimetry for selective internal radiation therapy using cone-beam CT for tumor and liver perfusion territory definition.

BACKGROUND: Selective internal radiation therapy (SIRT) is a promising treatment for unresectable hepatic malignancies. Predictive dose calculation based on a simulation using 99mTc-labeled macro-aggregated albumin (99mTc-MAA) before the treatment is considered as a potential tool for patient-specific treatment planning. Post-treatment dose measurement is mainly performed to confirm the planned absorbed dose to the tumor and non-tumor liver volumes. This study compared the predicted and measured absorbed dose distributions. METHODS: Thirty-one patients (67 tumors) treated by SIRT with resin microspheres were analyzed. Predicted and delivered absorbed dose was calculated using 99mTc-MAA-SPECT and 90Y-TOF-PET imaging. The voxel-level dose distribution was derived using the local deposition model. Liver perfusion territories and tumors have been delineated on contrast-enhanced CBCT images, which have been acquired during the 99mTc-MAA work-up. Several dose-volume histogram (DVH) parameters together with the mean dose for liver perfusion territories and non-tumoral and tumoral compartments were evaluated. RESULTS: A strong correlation between the predicted and measured mean dose for non-tumoral volume was observed (r = 0.937). The ratio of measured and predicted mean dose to this volume has a first, second, and third interquartile range of 0.83, 1.05, and 1.25. The difference between the measured and predicted mean dose did not exceed 11 Gy. The correlation between predicted and measured mean dose to the tumor was moderate (r = 0.623) with a mean difference of - 9.3 Gy. The ratio of measured and predicted tumor mean dose had a median of 1.01 with the first and third interquartile ranges of 0.58 and 1.59, respectively. Our results suggest that 99mTc-MAA-based dosimetry could predict under or over dosing of the non-tumoral liver parenchyma for almost all cases. For more than two thirds of the tumors, a predictive absorbed dose correctly indicated either good tumor dose coverage or under-dosing of the tumor. CONCLUSION: Our results highlight the predictive value of 99mTc-MAA-based dose estimation to predict non-tumor liver irradiation, which can be applied to prescribe an optimized activity aiming at avoiding liver toxicity. Compared to non-tumoral tissue, a poorer agreement between predicted and measured absorbed dose is observed for tumors.

Verification by the FISH translocation assay of historic doses to Mayak workers from external gamma radiation.

The aim of this study was to apply the fluorescence in situ hybridization (FISH) translocation assay in combination with chromosome painting of peripheral blood lymphocytes for retrospective biological dosimetry of Mayak nuclear power plant workers exposed chronically to external gamma radiation. These data were compared with physical dose estimates based on monitoring with badge dosimeters throughout each person's working life. Chromosome translocation yields for 94 workers of the Mayak production association were measured in three laboratories: Southern Urals Biophysics Institute, Leiden University Medical Center and the former Health Protection Agency of the UK (hereinafter Public Health England). The results of the study demonstrated that the FISH-based translocation assay in workers with prolonged (chronic) occupational gamma-ray exposure was a reliable biological dosimeter even many years after radiation exposure. Cytogenetic estimates of red bone marrow doses from external gamma rays were reasonably consistent with dose measurements based on film badge readings successfully validated in dosimetry system "Doses-2005" by FISH, within the bounds of the associated uncertainties.