L Yuan

Quantitative analysis of the factors which affect the interpatient organ-at-risk dose sparing variation in IMRT plans

PURPOSE The authors present an evidence-based approach to quantify the effects of an array of patient anatomical features of the planning target volumes (PTVs) and organs-at-risk (OARs) and their spatial relationships on the interpatient OAR dose sparing variation in intensity modulated radiation therapy (IMRT) plans by learning from a database of high-quality prior plans. METHODS The authors formulized the dependence of OAR dose volume histograms (DVHs) on patient anatomical factors into feature models which were learned from prior plans by a stepwise multiple regression method. IMRT plans for 64 prostate, 82 head-and-neck (HN) treatments were used to train the models. Two major groups of anatomical features were considered in this study: the volumetric information and the spatial information. The geometry of OARs relative to PTV is represented by the distance-to-target histogram, DTH. Important anatomical and dosimetric features were extracted from DTH and DVH by principal component analysis. The final models were tested by additional 24 prostate and 24 HN plans. RESULTS Significant patient anatomical factors contributing to OAR dose sparing in prostate and HN IMRT plans have been analyzed and identified. They are: the median distance between OAR and PTV, the portion of OAR volume within an OAR specific distance range, and the volumetric factors: the fraction of OAR volume which overlaps with PTV and the portion of OAR volume outside the primary treatment field. Overall, the determination coefficients R(2) for predicting the first principal component score (PCS1) of the OAR DVH by the above factors are above 0.68 for all the OARs and they are more than 0.53 for predicting the second principal component score (PCS2) of the OAR DVHs except brainstem and spinal cord. Thus, the above set of anatomical features combined has captured significant portions of the DVH variations for the OARs in prostate and HN plans. To test how well these features capture the interpatient organ dose sparing variations in general, the DVHs and specific dose-volume indices calculated from the regression models were compared with the actual DVHs and dose-volume indices from each patients plan in the validation dataset. The dose-volume indices compared were V99%, V85%, and V50% for bladder and rectum in prostate plans and parotids median dose in HN plans. The authors found that for the bladder and rectum models, 17 out of 24 plans (71%) were within 6% OAR volume error and 21 plans (85%) were within 10% error; For the parotids model, the median dose values for 30 parotids out of 48 (63%) were within 6% prescription dose error and the values in 40 parotids (83%) were within 10% error. CONCLUSIONS Quantitative analysis of patient anatomical features and their correlation with OAR dose sparing has identified a number of important factors that explain significant amount of interpatient DVH variations in OARs. These factors can be incorporated into evidence-based learning models as effective features to provide patient-specific OAR dose sparing goals.

Modeling the dosimetry of organ-at-risk in head and neck IMRT planning: An intertechnique and interinstitutional study

PURPOSE To build a statistical model to quantitatively correlate the anatomic features of structures and the corresponding dose-volume histogram (DVH) of head and neck (HN) Tomotherapy (Tomo) plans. To study if the model built upon one intensity modulated radiation therapy (IMRT) technique (such as conventional Linac) can be used to predict anticipated organs-at-risk (OAR) DVH of patients treated with a different IMRT technique (such as Tomo). To study if the model built upon the clinical experience of one institution can be used to aid IMRT planning for another institution. METHODS Forty-four Tomotherapy intensity modulate radiotherapy plans of HN cases (Tomo-IMRT) from Institution A were included in the study. A different patient group of 53 HN fixed gantry IMRT (FG-IMRT) plans was selected from Institution B. The analyzed OARs included the parotid, larynx, spinal cord, brainstem, and submandibular gland. Two major groups of anatomical features were considered: the volumetric information and the spatial information. The volume information includes the volume of target, OAR, and overlapped volume between target and OAR. The spatial information of OARs relative to PTVs was represented by the distance-to-target histogram (DTH). Important anatomical and dosimetric features were extracted from DTH and DVH by principal component analysis. Two regression models, one for Tomotherapy plan and one for IMRT plan, were built independently. The accuracy of intratreatment-modality model prediction was validated by a leave one out cross-validation method. The intertechnique and interinstitution validations were performed by using the FG-IMRT model to predict the OAR dosimetry of Tomo-IMRT plans. The dosimetry of OARs, under the same and different institutional preferences, was analyzed to examine the correlation between the model prediction and planning protocol. RESULTS Significant patient anatomical factors contributing to OAR dose sparing in HN Tomotherapy plans have been analyzed and identified. For all the OARs, the discrepancies of dose indices between the model predicted values and the actual plan values were within 2.1%. Similar results were obtained from the modeling of FG-IMRT plans. The parotid gland was spared in a comparable fashion during the treatment planning of two institutions. The model based on FG-IMRT plans was found to predict the median dose of the parotid of Tomotherapy plans quite well, with a mean error of 2.6%. Predictions from the FG-IMRT model suggested the median dose of the larynx, median dose of the brainstem and D2 of the brainstem could be reduced by 10.5%, 12.8%, and 20.4%, respectively, in the Tomo-IMRT plans. This was found to be correlated to the institutional differences in OAR constraint settings. Re-planning of six Tomotherapy patients confirmed the potential of optimization improvement predicted by the FG-IMRT model was correct. CONCLUSIONS The authors established a mathematical model to correlate the anatomical features and dosimetric indexes of OARs of HN patients in Tomotherapy plans. The model can be used for the setup of patient-specific OAR dose sparing goals and quality control of planning results.The institutional clinical experience was incorporated into the model which allows the model from one institution to generate a reference plan for another institution, or another IMRT technique.

Standardized beam bouquets for lung IMRT planning

The selection of the incident angles of the treatment beams is a critical component of intensity modulated radiation therapy (IMRT) planning for lung cancer due to significant variations in tumor location, tumor size and patient anatomy. We investigate the feasibility of establishing a small set of standardized beam bouquets for planning. The set of beam bouquets were determined by learning the beam configuration features from 60 clinical lung IMRT plans designed by experienced planners. A k-medoids cluster analysis method was used to classify the beam configurations in the dataset. The appropriate number of clusters was determined by maximizing the value of average silhouette width of the classification. Once the number of clusters had been determined, the beam arrangements in each medoid of the clusters were designated as the standardized beam bouquet for the cluster. This standardized bouquet set was used to re-plan 20 cases randomly selected from the clinical database. The dosimetric quality of the plans using the beam bouquets was evaluated against the corresponding clinical plans by a paired t-test. The classification with six clusters has the largest average silhouette width value and hence would best represent the beam bouquet patterns in the dataset. The results shows that plans generated with a small number of standardized bouquets (e.g. 6) have comparable quality to that of clinical plans. These standardized beam configuration bouquets will potentially help improve plan efficiency and facilitate automated planning.

Incorporating single-side sparing in models for predicting parotid dose sparing in head and neck IMRT

PURPOSE Sparing of single-side parotid gland is a common practice in head-and-neck (HN) intensity modulated radiation therapy (IMRT) planning. It is a special case of dose sparing tradeoff between different organs-at-risk. The authors describe an improved mathematical model for predicting achievable dose sparing in parotid glands in HN IMRT planning that incorporates single-side sparing considerations based on patient anatomy and learning from prior plan data. METHODS Among 68 HN cases analyzed retrospectively, 35 cases had physician prescribed single-side parotid sparing preferences. The single-side sparing model was trained with cases which had single-side sparing preferences, while the standard model was trained with the remainder of cases. A receiver operating characteristics (ROC) analysis was performed to determine the best criterion that separates the two case groups using the physicians single-side sparing prescription as ground truth. The final predictive model (combined model) takes into account the single-side sparing by switching between the standard and single-side sparing models according to the single-side sparing criterion. The models were tested with 20 additional cases. The significance of the improvement of prediction accuracy by the combined model over the standard model was evaluated using the Wilcoxon rank-sum test. RESULTS Using the ROC analysis, the best single-side sparing criterion is (1) the predicted median dose of one parotid is higher than 24 Gy; and (2) that of the other is higher than 7 Gy. This criterion gives a true positive rate of 0.82 and a false positive rate of 0.19, respectively. For the bilateral sparing cases, the combined and the standard models performed equally well, with the median of the prediction errors for parotid median dose being 0.34 Gy by both models (p = 0.81). For the single-side sparing cases, the standard model overestimates the median dose by 7.8 Gy on average, while the predictions by the combined model differ from actual values by only 2.2 Gy (p = 0.005). Similarly, the sum of residues between the modeled and the actual plan DVHs is the same for the bilateral sparing cases by both models (p = 0.67), while the standard model predicts significantly higher DVHs than the combined model for the single-side sparing cases (p = 0.01). CONCLUSIONS The combined model for predicting parotid sparing that takes into account single-side sparing improves the prediction accuracy over the previous model.

From active shape model to active optical flow model: a shape-based approach to predicting voxel-level dose distributions in spine SBRT

Prediction of achievable dose distribution in spine stereotactic body radiation therapy (SBRT) can help in designing high-quality treatment plans to maximally protect spinal cords and to effectively control tumours. Dose distributions at spinal cords are primarily affected by the shapes of adjacent planning target volume (PTV) contours. In this work, we estimate such contour effects and predict dose distributions by exploring active optical flow model (AOFM) and active shape model (ASM). We first collect a sequence of dose sub-images and PTV contours near spinal cords from fifteen SBRT plans in the training dataset. The data collection is then classified into five groups according to the PTV locations in relation to spinal cords. In each group, we randomly choose a dose sub-image as the reference and register all other sub-images to the reference using an optical flow method. AOFM is then constructed by importing optical flow vectors and dose values into the principal component analysis (PCA). Similarly, we build ASM by using PCA on PTV contour points. The correlation between ASM and AOFM is estimated via a stepwise multiple regression model. When predicting dose distribution of a new case, the group is first determined based on the PTV contour. The prediction model of the selected group is used to estimate dose distributions by mapping the PTV contours from the ASM space to the AOFM space. This method was validated on fifteen SBRT plans in the testing dataset. Analysis of dose-volume histograms revealed that the important D2%, D5%, D10% and D0.1cc dosimetric parameters of spinal cords between the prediction and the clinical plans were 11.7 ± 1.7 Gy versus 11.8 ± 1.7 Gy (p = 0.95), 10.9 ± 1.7 Gy versus 11.1 ± 1.9 Gy (p = 0.8295), 10.2 ± 1.6 Gy versus 10.1 ± 1.7 (p = 0.9036) and 11.2 ± 2.0 Gy versus 11.1 ± 2.2 Gy (p = 0.5208), respectively. Here, the ‘cord’ is the spinal cord proper (not the thecal sac) extended 5 mm inferior and superior to the involved vertebral bodies, and the ‘PTV’ is the involved segment of the vertebral body expanded uniformly by 2 mm but excluding the spinal cord volume expanded by 2 mm (Ref. RTOG 0631). These results suggested that the AOFM-based approach is a promising tool for predicting accurate spinal cord dose in clinical practice. In this work, we demonstrated the feasibility of using AOFM and ASM models derived from previously treated patients to estimate the achievable dose distributions for new patients.

Impact of dose calculation accuracy during optimization on lung IMRT plan quality

The purpose of this study was to evaluate the effect of dose calculation accuracy and the use of an intermediate dose calculation step during the optimization of intensity‐modulated radiation therapy (IMRT) planning on the final plan quality for lung cancer patients. This study included replanning for 11 randomly selected free‐breathing lung IMRT plans. The original plans were optimized using a fast pencil beam convolution algorithm. After optimization, the final dose calculation was performed using the analytical anisotropic algorithm (AAA). The Varian Treatment Planning System (TPS) Eclipse v11, includes an option to perform intermediate dose calculation during optimization using the AAA. The new plans were created using this intermediate dose calculation during optimization with the same planning objectives and dose constraints as in the original plan. Differences in dosimetric parameters for the planning target volume (PTV) dose coverage, organs‐at‐risk (OARs) dose sparing, and the number of monitor units (MU) between the original and new plans were analyzed. Statistical significance was determined with a p‐value of less than 0.05. All plans were normalized to cover 95% of the PTV with the prescription dose. Compared with the original plans, the PTV in the new plans had on average a lower maximum dose (69.45 vs. 71.96 Gy, p=0.005), a better homogeneity index (HI) (0.08 vs. 0.12, p=0.002), and a better conformity index (CI) (0.69 vs. 0.59, p=0.003). In the new plans, lung sparing was increased as the volumes receiving 5, 10, and 30 Gy were reduced when compared to the original plans (40.39% vs. 42.73%, p=0.005; 28.93% vs. 30.40%, p=0.001; 14.11% vs. 14.84%, p=0.031). The volume receiving 20 Gy was not significantly lower (19.60% vs. 20.38%, p=0.052). Further, the mean dose to the lung was reduced in the new plans (11.55 vs. 12.12 Gy, p=0.024). For the esophagus, the mean dose, the maximum dose, and the volumes receiving 20 and 60 Gy were lower in the new plans than in the original plans (17.91 vs. 19.24 Gy, p=0.004; 57.32 vs. 59.81 Gy, p=0.020; 39.34% vs. 41.59%, p=0.097; 12.56% vs. 15.35%, p=0.101). For the heart, the mean dose, the maximum dose, and the volume receiving 40 Gy were also lower in new plans (11.07 vs. 12.04 Gy, p=0.007; 56.41 vs. 57.7 Gy, p=0.027; 7.16% vs. 9.37%, p=0.012). The maximum dose to the spinal cord in the new plans was significantly lower than in the original IMRT plans (29.1 vs. 31.39 Gy, p=0.014). Difference in MU between the IMRT plans was not significant (1216.90 vs. 1198.91, p=0.328). In comparison to the original plans, the number of iterations needed to meet the optimization objectives in the new plans was reduced by a factor of 2 (2–3 vs. 5–6 iterations). Further, optimization was 30% faster corresponding to an average time savings of 10–15 min for the reoptimized plans. Accuracy of the dose calculation algorithm during optimization has an impact on planning efficiency, as well as on the final plan dosimetric quality. For lung IMRT treatment planning, utilizing the intermediate dose calculation during optimization is feasible for dose homogeneity improvement of the PTV and for improvement of optimization efficiency. PACS numbers: 87.55.D‐, 87.55.de, 87.55.dk

Outlier identification in radiation therapy knowledge‐based planning: A study of pelvic cases

Purpose: The purpose of this study was to apply statistical metrics to identify outliers and to investigate the impact of outliers on knowledge‐based planning in radiation therapy of pelvic cases. We also aimed to develop a systematic workflow for identifying and analyzing geometric and dosimetric outliers. Methods: Four groups (G1–G4) of pelvic plans were sampled in this study. These include the following three groups of clinical IMRT cases: G1 (37 prostate cases), G2 (37 prostate plus lymph node cases) and G3 (37 prostate bed cases). Cases in G4 were planned in accordance with dynamic‐arc radiation therapy procedure and include 10 prostate cases in addition to those from G1. The workflow was separated into two parts: 1. identifying geometric outliers, assessing outlier impact, and outlier cleaning; 2. identifying dosimetric outliers, assessing outlier impact, and outlier cleaning. G2 and G3 were used to analyze the effects of geometric outliers (first experiment outlined below) while G1 and G4 were used to analyze the effects of dosimetric outliers (second experiment outlined below).A baseline model was trained by regarding all G2 cases as inliers. G3 cases were then individually added to the baseline model as geometric outliers. The impact on the model was assessed by comparing leverages of inliers (G2) and outliers (G3). A receiver‐operating‐characteristic (ROC) analysis was performed to determine the optimal threshold. The experiment was repeated by training the baseline model with all G3 cases as inliers and perturbing the model with G2 cases as outliers.A separate baseline model was trained with 32 G1 cases. Each G4 case (dosimetric outlier) was subsequently added to perturb the model. Predictions of dose‐volume histograms (DVHs) were made using these perturbed models for the remaining 5 G1 cases. A Weighted Sum of Absolute Residuals (WSAR) was used to evaluate the impact of the dosimetric outliers. Results: The leverage of inliers and outliers was significantly different. The Area‐Under‐Curve (AUC) for differentiating G2 (outliers) from G3 (inliers) was 0.98 (threshold: 0.27) for the bladder and 0.81 (threshold: 0.11) for the rectum. For differentiating G3 (outlier) from G2 (inlier), the AUC (threshold) was 0.86 (0.11) for the bladder and 0.71 (0.11) for the rectum. Significant increase in WSAR was observed in the model with 3 dosimetric outliers for the bladder (P < 0.005 with Bonferroni correction), and in the model with only 1 dosimetric outlier for the rectum (P < 0.005). Conclusions: We established a systematic workflow for identifying and analyzing geometric and dosimetric outliers, and investigated statistical metrics for outlier detection. Results validated the necessity for outlier detection and clean‐up to enhance model quality in clinical practice.

MO‐D‐BRB‐09: Treatment Delivery QA for Online Adaptive Radiation Therapy Based on Dynamic Machine Information (DMI): A Feasibility Study

PURPOSE To implement a quality assurance (QA) system for the treatment delivery of online adaptive radiation therapy utilizing Dynamic Machine Information (DMI). MATERIALS AND METHODS DMI provides the expected/actual MLC leaf-positions, delivered MU, and beam-on status every 50ms during delivery. In this study a stream of DMI inputs is simulated by playing back Dynalog information recorded while delivering a test fluence map (FM). Based on these DMI inputs, the QA system performs three levels of monitoring/verification on the plan delivery process: (1) Following each input, actual and expected FMs delivered up to the current MLC position is dynamically updated using corresponding MLC positions in the DMI. The magnitude and frequency of pixel-by-pixel fluence differences between these two FMs are calculated and visualized in histograms.(2) At each control point, actual MLC positions are verified against the treatment plan for potential errors in data transfer between the treatment planning system (TPS) and the MLC controller.(3) Both (1) and (2) can signal beam-hold with a user-specified error tolerance.(4) After treatment, delivered dose is reconstructed in TPS based on DMI data during delivery, and compared to planned dose. RESULTS (1) Efficiency: Average latency from DMI input to the completion of fluence difference calculation is <1ms.(2) Efficacy: For test FM, transient error in leaf positions is (-0.07±0.28)mm; cumulative errors in delivered fluence is (0.003±0.183)% of the maximal fluence. The system can also identify data transfer errors between TPS and MLC controller. Off-line dose reconstruction and evaluation show <0.5% dosimetric discrepancy from planned dose distribution for the test FM. CONCLUSION This QA system is capable of identifying MLC position/fluence errors in near real-time, and assessing dosimetric impact of the treatment delivery process. It is thus a valuable tool for clinical implementation of online adaptive radiation therapy. (Research partially supported by Varian) Research partially supported by Varian Medical Systems.

SU‐E‐T‐537: Local Multi‐Criteria Optimization for Clinical Tradeoff Decision Guidance in RT Planning

Purpose: We present a novel method for generating a local segment of Pareto surface around the best achievable plan predicted by a knowledge model. This local-MCO will provide an efficient method to enable clinically viable tradeoff decisions in IMRT planning tailored to the patient’s specific needs. Methods: Multi-criteria optimization (MCO) methods provide physicians and planners the ability to explore tradeoff options in RT planning. However, generic MCO methods are often time consuming because a global Pareto surface (PS) need to be explored if patient-specific clinical conditions and planners’ planning experiences are not taken into consideration. We have developed a local-MCO approach, which incorporates the knowledge model prediction based on the individual patient’s features and planning experience into the generation of a local PS around a predicted plan. In the proposed method, the starting points of clinical relevant organ sparing objectives are predicted by the knowledge models, and then a local PS is searched near the model predictions. As an initial assessment, the local PS is compared with the global PS generated for two prostate cancer cases. The mean minimum distance from each plan on the local PS to the global PS and the range of clinical acceptable dosimetric parameters covered by the local and global PS are calculated. Results: The local PS agrees well with the global PS. The mean minimum distance between the local and global PS in the PTV-bladder-rectum dose objective space are about 3% and 1% of prescription dose for the two plans, respectively. Although the local PS is only a small portion of the global PS, they cover most of the clinically relevant dose range. Conclusion: The local MCO results in a smaller but clinically more relevant PS. It is an efficient method to provide physicians with guidance of patient-specific trade-off options based on practice experience. Partially supported by NIH/NCI under grant #R21CA161389 and a master research grant by Varian Medical Systems.

SU-E-T-229: Machine Learning Methods for Knowledge Based Treatment Planning of Prostate Cancer

PURPOSE To evaluate the accuracy of the dose prediction models constructed with machine learning techniques, Support Vector Machine (SVM) and Artificial Neural Network (ANN) for the prediction of dose volume histogram (DVH) of organs-at-risk (OAR) in IMRT, compared to the model constructed by stepwise multiple regression (MR), and to investigate the number of prior plans required for the models to produce reliable predictions. METHODS IMRT plans from 102 prostate cases were randomly divided into two datasets for training and testing, respectively. The testing dataset contains a fixed number of 20 cases, while the number of cases in the training dataset varied from 5 to 80. Models were constructed with SVM, ANN, or MR to formulate the dependence of the OAR DVH on patient anatomical features including the Distance to Target Histogram (DTH), PTV and OAR volumes and their overlap, among other volumetric or spatial information. The D50 (Dose value at 50% volume) and the mean square of difference between D50 of clinical and predicted DVH were calculated for each modeling technique at each specific training dataset number. RESULTS The mean square of difference of D50 between clinical and predicted DVH decreases with the number of cases in the training dataset, and reaches stable beyond 30 for MR. With the 80 case training dataset, for the bladder model, the SVM predicted 70% D50 values within 10% error and the ANN predicted 85%, compared to 85% with multiple regression. For the rectum model, the numbers are SVM 80%, ANN 70%, and MR 85%. CONCLUSION The machine learning techniques SVM and ANN are comparable to MR for producing OAR DVH prediction of the prostate cancer. The minimal number of training cases is around 30. Partially supported by NIH/NCI under grant #R21CA161389 and a master research grant by Varian Medical System.