Localization of deformable tumors from short-arc projections using Bayesian estimation.

PURPOSE: The authors present a stochastic framework for radiotherapy patient positioning directly utilizing radiographic projections. This framework is developed to be robust against anatomical nonrigid deformations and to cope with challenging imaging scenarios, involving only a few cone beam CT projections from short arcs. METHODS: Specifically, a Bayesian estimator (BE) is explicitly derived for the given scanning geometry. This estimator is compared to reference methods such as chamfer matching (CM) and the minimization of the median absolute error adapted as tools of robust image processing and statistics. In order to show the performance of the stochastic short-arc patient positioning method, a CIRS IMRT thorax phantom study is presented with movable markers and the utilization of an Elekta Synergy(®) XVI system. Furthermore, a clinical prostate CBCT scan of a Varian(®) On-Board Imager(®) system is utilized to investigate the robustness of the method for large variations of image quality (anterior-posterior vs lateral views). RESULTS: The results show that the BE shifts reduce the initial setup error of up to 3 cm down to 3 mm at maximum for an imaging arc as short as 10° while CM achieves residual errors of 7 mm at maximum only for arcs longer than 40°. Furthermore, the BE can compensate robustly for low image qualities using several low quality projections simultaneously. CONCLUSIONS: In conclusion, an estimation method for marker-based patient positioning for short imaging arcs is presented and shown to be robust and accurate for deformable anatomies.

An efficient inverse radiotherapy planning method for VMAT using quadratic programming optimization.

PURPOSE: The purpose of this study is to investigate the feasibility of an inverse planning optimization approach for the Volumetric Modulated Arc Therapy (VMAT) based on quadratic programming and the projection method. The performance of this method is evaluated against a reference commercial planning system (eclipse(TM) for rapidarc(TM)) for clinically relevant cases. METHODS: The inverse problem is posed in terms of a linear combination of basis functions representing arclet dose contributions and their respective linear coefficients as degrees of freedom. MLC motion is decomposed into basic motion patterns in an intuitive manner leading to a system of equations with a relatively small number of equations and unknowns. These equations are solved using quadratic programming under certain limiting physical conditions for the solution, such as the avoidance of negative dose during optimization and Monitor Unit reduction. The modeling by the projection method assures a unique treatment plan with beneficial properties, such as the explicit relation between organ weightings and the final dose distribution. Clinical cases studied include prostate and spine treatments. The optimized plans are evaluated by comparing isodose lines, DVH profiles for target and normal organs, and Monitor Units to those obtained by the clinical treatment planning system eclipse(TM). RESULTS: The resulting dose distributions for a prostate (with rectum and bladder as organs at risk), and for a spine case (with kidneys, liver, lung and heart as organs at risk) are presented. Overall, the results indicate that similar plan qualities for quadratic programming (QP) and rapidarc(TM) could be achieved at significantly more efficient computational and planning effort using QP. Additionally, results for the quasimodo phantom [Bohsung et al., "IMRT treatment planning: A comparative inter-system and inter-centre planning exercise of the estro quasimodo group," Radiother. Oncol. 76(3), 354-361 (2005)] are presented as an example for an extreme concave case. CONCLUSION: Quadratic programming is an alternative approach for inverse planning which generates clinically satisfying plans in comparison to the clinical system and constitutes an efficient optimization process characterized by uniqueness and reproducibility of the solution.

Stochastic formulation of patient positioning using linac-mounted cone beam imaging with prior knowledge.

PURPOSE: In this work, a novel stochastic framework for patient positioning based on linac-mounted CB projections is introduced. Based on this formulation, the most probable shifts and rotations of the patient are estimated, incorporating interfractional deformations of patient anatomy and other uncertainties associated with patient setup. METHODS: The target position is assumed to be defined by and is stochastically determined from positions of various features such as anatomical landmarks or markers in CB projections, i.e., radiographs acquired with a CB-CT system. The patient positioning problem of finding the target location from CB projections is posed as an inverse problem with prior knowledge and is solved using a Bayesian maximum a posteriori (MAP) approach. The prior knowledge is three-fold and includes the accuracy of an initial patient setup (such as in-room laser and skin marks), the plasticity of the body (relative shifts between target and features), and the feature detection error in CB projections (which may vary depending on specific detection algorithm and feature type). For this purpose, MAP estimators are derived and a procedure of using them in clinical practice is outlined. Furthermore, a rule of thumb is theoretically derived, relating basic parameters of the prior knowledge (initial setup accuracy, plasticity of the body, and number of features) and the parameters of CB data acquisition (number of projections and accuracy of feature detection) to the expected estimation accuracy. RESULTS: MAP estimation can be applied to arbitrary features and detection algorithms. However, to experimentally demonstrate its applicability and to perform the validation of the algorithm, a water-equivalent, deformable phantom with features represented by six 1 mm chrome balls were utilized. These features were detected in the cone beam projections (XVI, Elekta Synergy) by a local threshold method for demonstration purposes only. The accuracy of estimation (strongly varying for different plasticity parameters of the body) agreed with the rule of thumb formula. Moreover, based on this rule of thumb formula, about 20 projections for 6 detectable features seem to be sufficient for a target estimation accuracy of 0.2 cm, even for relatively large feature detection errors with standard deviation of 0.5 cm and spatial displacements of the features with standard deviation of 0.5 cm. CONCLUSIONS: The authors have introduced a general MAP-based patient setup algorithm accounting for different sources of uncertainties, which are utilized as the prior knowledge in a transparent way. This new framework can be further utilized for different clinical sites, as well as theoretical developments in the field of patient positioning for radiotherapy.

An alternative VMAT with prior knowledge about the type of leaf motion utilizing projection method for concave targets.

The authors present an alternative approach to inverse planning optimization and apply it to volumetric modulated are therapy (VMAT) in one rotation with a prior knowledge about the type of leaf motions. The optimization is based on the projection theorem in inner product spaces. MLC motion is directly considered in the optimization, thus avoiding leaf segmentation characteristic of IMRT optimization. In this work they realize the method for concave irregular targets encompassing an organ at risk leading to a repetitive MLC motion pattern. Applying the projection theorem leads to a noniterative optimization method and reduces to solving few systems of linear equations with small numbers of dimensions. The solution of the inverse problem is unique, and false minima are naturally excluded. They divided the full rotation into about 50 short arc segments and for each segment decomposed dose into separate contributions related to stages of MLC motion. This results generally in an inverse problem with just four free parameters per arc segment. Practically three degrees of freedom will be used for the purpose of a constant angular speed of the gantry. Therefore the total number of degrees of freedom for a 3D problem is about 3 x 50 x number of collimator leaf pairs for irradiating the whole target volume in one rotation. Two 2D and one 3D concave target volumes are applied for a slice by slice optimization. A 6 MV photon beam model is used, including realistic scattering and attenuation, and a maximal leaf velocity of 3 cm/s is regarded. The resulting dose distributions cover the PTVs very well and have maxima at about 108% of dose in the PTVs. The OAR is spared very strong in all cases. As a result of optimization, the MLC apertures are repetitively opening and closing and can be interpreted in an intuitive way. Applying the projection method for this knowledge-based VMAT delivery scheme for concave target volumes is an alternative technique for dose optimization. There are several properties, such as uniqueness of MLC motions and their continuous dependence on geometry and prescribed dose, that make this approach interesting to inverse planning. This method is still in an investigational stage, but promising results are presented. In future work it will be extended directly (without conceptual changes) in several directions to be more clinically applicable.