Prediction of multi-criteria optimization (MCO) parameter efficiency in volumetric modulated arc therapy (VMAT) treatment planning using machine learning (ML).

PURPOSE: To predict the impact of optimization parameter changes on dosimetric plan quality criteria in multi-criteria optimized volumetric-modulated-arc therapy (VMAT) planning prior to optimization using machine learning (ML). METHODS: A data base comprising a total of 21,266 VMAT treatment plans for 44 cranial and 18 spinal patient geometries was generated. The underlying optimization algorithm is governed by three highly composite parameters which model a combination of important aspects of the solution. Patient geometries were parametrized via volume- and shape properties of the voxel objects and overlap-volume histograms (OVH) of the planning-target-volume (PTV) and a relevant organ-at-risk (OAR). The impact of changes in one of the three optimization parameters on the maximally achievable value range of five dosimetric properties of the resulting dose distributions was studied. To predict the extent of this impact based on patient geometry, treatment site, and current parameter settings prior to optimization, three different ML-models were trained and tested. Precision-recall curves, as well as the area-under-curve (AUC) of the resulting receiver-operator-characteristic (ROC) curves were analyzed for model assessment. RESULTS: Successful identification of parameter regions resulting in a high variability of dosimetric plan properties depended on the choice of geometry features, the treatment indication and the plan property under investigation. AUC values between 0.82 and 0.99 could be achieved. The best average-precision (AP) values obtained from the corresponding precision/recall curves ranged from 0.71 to 0.99. CONCLUSIONS: Machine learning models trained on a database of pre-optimized treatment plans can help finding relevant optimization parameter ranges prior to optimization.

Approximation of dose quality indicator values in multi-criteria optimized (MCO) volumetric modulated arc therapy (VMAT) treatment planning using trilinear dose interpolation.

PURPOSE: To approximate dose-volume histogram (DVH) based quality indicators in volumetric modulated arc therapy (VMAT) planning using multi-criteria optimization (MCO) with a low number of composite optimization parameters. METHODS: The solution space for VMAT optimization with a low number of composite optimization parameters is approximated by trilinear dose inter- polation and prediction of dose-volume-histogram (DVH) based plan quality indicator values. To assess the approximation quality a diverse dataset of 44 cranial and 18 spine patient geometries was chosen. Optimization results are governed by three composite parameters focusing on target-organ-at-risk- (OAR)-trade-off, overall healthy tissue sparing, and delivery/quality assurance complexity. 21,266 optimized dose distributions were pre-calculated and the numerical values for a choice of 10 DVH points, referred to as plan quality indicators, were stored to serve as ground truth. Using a subset of 8 and 27 pre-calculated optimization results, dose distributions for unknown parameter values were approximated by trilinear interpolation. The resulting quality indicator values were compared to the previously obtained exact solutions. RESULTS: The magnitude of the deviation between exact and approximated values varied largely with respect to patient geometry and the criterion under investigation. Approximation with 27 pre-calculated results yielded lower deviations than approximation with 8 results, at the cost of a higher pre-calculation workload. CONCLUSIONS: Solution space approximation via trilinear dose interpolation in VMAT treatment planning governed by composite optimization parameters is possible without further knowledge of the internal implementation of the underlying optimizer. Maximum average deviations between approxi- mation and actual values of characteristic dose quality indicators below 1% (cranial) and 8% (spine) allow for a quick qualitative assessment of the possible solution landscape.

A dose error evaluation study for 4D dose calculations.

Previous studies have shown that respiration induced motion is not negligible for Stereotactic Body Radiation Therapy. The intrafractional breathing induced motion influences the delivered dose distribution on the underlying patient geometry such as the lung or the abdomen. If a static geometry is used, a planning process for these indications does not represent the entire dynamic process. The quality of a full 4D dose calculation approach depends on the dose coordinate transformation process between deformable geometries. This article provides an evaluation study that introduces an advanced method to verify the quality of numerical dose transformation generated by four different algorithms.The used transformation metric value is based on the deviation of the dose mass histogram (DMH) and the mean dose throughout dose transformation. The study compares the results of four algorithms. In general, two elementary approaches are used: dose mapping and energy transformation. Dose interpolation (DIM) and an advanced concept, so called divergent dose mapping model (dDMM), are used for dose mapping. The algorithms are compared to the basic energy transformation model (bETM) and the energy mass congruent mapping (EMCM). For evaluation 900 small sample regions of interest (ROI) are generated inside an exemplary lung geometry (4DCT). A homogeneous fluence distribution is assumed for dose calculation inside the ROIs. The dose transformations are performed with the four different algorithms.The study investigates the DMH-metric and the mean dose metric for different scenarios (voxel sizes: 8 mm, 4 mm, 2 mm, 1 mm; 9 different breathing phases). dDMM achieves the best transformation accuracy in all measured test cases with 3-5% lower errors than the other models. The results of dDMM are reasonable and most efficient in this study, although the model is simple and easy to implement. The EMCM model also achieved suitable results, but the approach requires a more complex programming structure. The study discloses disadvantages for the bETM and for the DIM. DIM yielded insufficient results for large voxel sizes, while bETM is prone to errors for small voxel sizes.

A methodology for verification of radiotherapy dose calculation.

OBJECT: A methodology for dosimetric verification of radiation therapy plans was developed and implemented. Dosimetric accuracy of clinically active intensity-modulated radiotherapy (IMRT) and intensity-modulated radiosurgery (IMRS) programs was assessed using this methodology. METHODS: The methodology included several dosimetric tasks that were performed to assess the dosimetric accuracy of a treatment plan. Absolute dosimetry of the composite plan was performed using an ionization chamber. Film dosimetry was performed for each individual field and for the multifield composite plan. Calculated dose distributions and film measurements were compared using software developed for the specific tasks. Two-dimensional maps of gamma index, dose difference, and distance-to-agreement were calculated and displayed. To date, good agreement between measurements and calculations has been observed in 160 clinical IMRT and IMRS plans. The largest observed absolute dose disagreement was -4.79%. The mean absolute dose difference was 0.26%, with a standard deviation of 1.75%. The authors specify a 3% dose difference and 3-mm distance as the scaling acceptability criteria for the gamma index calculations of the film measurement analysis. The planning and delivery system in clinical use has proven consistently to satisfy these criteria. CONCLUSIONS: The dosimetric verification methods and the software tools developed were both quantitative and clinically practical. The measurements and the analysis demonstrated that the IMRT and IMRS planning and delivery system in use was sufficiently accurate for highly conformal treatments.