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.

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.

Effect of treatment planning system parameters on beam modulation complexity for treatment plans with single-layer multi-leaf collimator and dual-layer stacked multi-leaf collimator.

OBJECTIVE: High levels of beam modulation complexity (MC) and monitor units (MU) can compromise the plan deliverability of intensity-modulated radiotherapy treatments. Our study evaluates the effect of three treatment planning system (TPS) parameters on MC and MU using different multi-leaf collimator (MLC) architectures. METHODS: 192 volumetric modulated arc therapy plans were calculated using one virtual prostate phantom considering three main settings: (1) three TPS-parameters (Convergence; Aperture Shape Controller, ASC; and Dose Calculation Resolution, DCR) selected from Eclipse v15.6, (2) four levels of dose-sparing priority for organs at risk (OAR), and (3) two treatment units with same nominal conformity resolution and different MLC architectures (Halcyon-v2 dual-layer MLC, DL-MLC & TrueBeam single-layer MLC, SL-MLC). We use seven complexity metrics to evaluate the MC, including two new metrics for DL-MLC, assessed by their correlation with γ passing rate (GPR) analysis. RESULTS: DL-MLC plans demonstrated lower dose-sparing values than SL-MLC plans (p<0.05). TPS-parameters did not change significantly the complexity metrics for either MLC architectures. However, for SL-MLC, significant variations of MU, target volume dose-homogeneity, and dose spillage were associated with ASC and DCR (p<0.05). MU were found to be correlated (highly or moderately) with all complexity metrics (p<0.05) for both MLC plans. Additionally, our new complexity metrics presented a moderate correlation with GPR (r<0.65). An important correlation was demonstrated between MC (plan deliverability) and dose-sparing priority level for DL-MLC. CONCLUSIONS: TPS-parameters selected do not change MC for DL-MLC architecture, but they might have a potential use to control the MU, PTV homogeneity or dose spillage for SL-MLC. Our new DL-MLC complexity metrics presented important information to be considered in future pre-treatment quality assurance programs. Finally, the prominent dependence between plan deliverability and priority applied to OAR dose sparing for DL-MLC needs to be analyzed and considered as an additional predictor of GPRs in further studies. ADVANCES IN KNOWLEDGE: Dose-sparing priority might influence in modulation complexity of DL-MLC.

High-dose-rate brachytherapy delivered in two fractions as monotherapy for low-risk prostate cancer.

PURPOSE: High-dose-rate (HDR) brachytherapy has been accepted as an effective and safe method to treat prostate cancer. The aim of this study was to describe acute toxicity following HDR brachytherapy to the prostate, and to examine the association between dosimetric parameters and urinary toxicity in low-risk prostate cancer patients. MATERIAL AND METHODS: Patients with low-risk prostate cancer were given HDR brachytherapy as monotherapy in two 12.5 Gy fractions. Planning objectives for the planning target volume (PTV) were V100% ≥ 90% and V150% ≤ 35%. Planning objectives for organs at risk were V75% ≤ 1 cc for the bladder, rectum and perineum, and V125% ≤ 1 cc for the urethra. Toxicity was assessed three months after treatment using the Common Terminology Criteria for Adverse Events. RESULTS: Seventy-three patients were included in the analysis. Thirty-three patients (45%) reported having any type of toxicity in the three months following HDR brachytherapy. Most toxicity cases (26%) were grade 1 urinary toxicity. Mean coverage index was 0.89 and mean V100 was 88.85. Doses administered to the urethra were associated with urinary toxicity. Patients who received more than 111.3% of the prescribed dose in 1 cc of the urethra were four times more likely to have urinary toxicity compared to patients receiving less than 111.3% (OR = 4.71, 95% CI: 1.43-15.6; p = 0.011). CONCLUSIONS: High-dose-rate brachytherapy administered as monotherapy for prostate cancer proved to be a safe alternative treatment for patients with low-risk prostate cancer. Urinary toxicity was associated with the dose administered to 1 cc and 0.1 cc of the urethra and was remarkably inferior to the reported toxicity in similar studies.