Inverse treatment planning with adaptively evolving voxel-dependent penalty scheme

Efficient radiation treatment planning based on voxel importance

Objective.Radiation treatment planning (RTP) involves optimization over a large number of voxels, many of which carry limited information about the clinical problem. We propose an approach to reduce the large optimization problem by only using a representative subset of informative voxels. This way, we drastically improve planning efficiency while maintaining the plan quality.Approach.Within an initial probing step, we pre-solve an easier optimization problem involving a simplified objective from which we derive an importance score per voxel. This importance score is then turned into a sampling distribution, which allows us to subsample a small set of informative voxels using importance sampling. By solving a-now reduced-version of the original optimization problem using this subset, we effectively reduce the problem's size and computational demands while accounting for regions where satisfactory dose deliveries are challenging.Main results.In contrast to other stochastic (sub-)sampling methods, our technique only requires a single probing and sampling step to define a reduced optimization problem. This problem can be efficiently solved using established solvers without the need of modifying or adapting them. Empirical experiments on open benchmark data highlight substantially reduced optimization times, up to 50 times faster than the original ones, for intensity-modulated radiation therapy, all while upholding plan quality comparable to traditional methods.Significance.Our novel approach has the potential to significantly accelerate RTP by addressing its inherent computational challenges. We reduce the treatment planning time by reducing the size of the optimization problem rather than modifying and improving the optimization method. Our efforts are thus complementary to many previous developments.

Adaptive Region-Specific Loss for Improved Medical Image Segmentation

Defining the loss function is an important part of neural network design and critically determines the success of deep learning modeling. A significant shortcoming of the conventional loss functions is that they weight all regions in the input image volume equally, despite the fact that the system is known to be heterogeneous (i.e., some regions can achieve high prediction performance more easily than others). Here, we introduce a region-specific loss to lift the implicit assumption of homogeneous weighting for better learning. We divide the entire volume into multiple sub-regions, each with an individualized loss constructed for optimal local performance. Effectively, this scheme imposes higher weightings on the sub-regions that are more difficult to segment, and vice versa. Furthermore, the regional false positive and false negative errors are computed for each input image during a training step and the regional penalty is adjusted accordingly to enhance the overall accuracy of the prediction. Using different public and in-house medical image datasets, we demonstrate that the proposed regionally adaptive loss paradigm outperforms conventional methods in the multi-organ segmentations, without any modification to the neural network architecture or additional data preparation.

Automatic Inverse Treatment Planning of Gamma Knife Radiosurgery via Deep Reinforcement Learning

PURPOSE

Several inverse planning algorithms have been developed for Gamma Knife (GK) radiosurgery to determine a large number of plan parameters via solving an optimization problem, which typically consists of multiple objectives. The priorities among these objectives need to be repetitively adjusted to achieve a clinically good plan for each patient. This study aimed to achieve automatic and intelligent priority-tuning, by developing a deep reinforcement learning (DRL) based method to model the tuning behaviors of human planners.

METHODS

We built a priority-tuning policy network using deep convolutional neural networks. Its input was a vector composed of multiple plan metrics that were used in our institution for GK plan evaluation. The network can determine which tuning action to take, based on the observed quality of the intermediate plan. We trained the network using an end-to-end DRL framework to approximate the optimal action-value function. A scoring function was designed to measure the plan quality to calculate the received reward of a tuning action.

RESULTS

Vestibular schwannoma was chosen as the test bed in this study. The number of training, validation and testing cases were 5, 5, and 16, respectively. For these three datasets, the average scores of the initial plans obtained with a same initial priority set were 3.63 ± 1.34, 3.83 ± 0.86 and 4.20 ± 0.78, respectively, while can be improved to 5.28 ± 0.23, 4.97 ± 0.44 and 5.22 ± 0.26 through manual priority tuning by human expert planners. Our network achieved competitive results with 5.42 ± 0.11, 5.10 ± 0. 42, 5.28 ± 0.20, respectively.

CONCLUSIONS

Our network can generate GK plans of comparable or slightly higher quality comparing with the plans generated by human planners via manual priority tuning for vestibular schwannoma cases. The network can potentially be incorporated into the clinical workflow as a planning assistance to improve GK planning efficiency and help to reduce plan quality variation caused by inter-planner variability. We also hope that our method can reduce the workload of GK planners and allow them to spend more time on more challenging cases. This article is protected by copyright. All rights reserved.

A hierarchical deep reinforcement learning framework for intelligent automatic treatment planning of prostate cancer intensity modulated radiation therapy

Purpose.We have previously proposed an intelligent automatic treatment planning (IATP) framework that builds a virtual treatment planner network (VTPN) to operate a treatment planning system (TPS) to generate high-quality radiation therapy (RT) treatment plans. While the potential of IATP in automating RT treatment planning has been demonstrated, its poor scalability caused by an almost linear growth of network size with the number of treatment planning parameters (TPPs) is a bottleneck, preventing its application in complicate, but clinically relevant treatment planning problems. The decision-making behavior of the trained network is hard to understand. Motivated by the decision-making process of a human planner, this study proposes a hierarchical IATP framework.Methods and materials.The hierarchical VTPN (HieVTPN) consists of three networks, i.e. Structure-Net, Parameter-Net, and Action-Net. When interacting with a TPS, the networks are employed in a sequential order in each step to decide the structure to adjust, the TPP to adjust for the selected structure, and the specific adjustment manner for the parameter, respectively. We developed an end-to-end hierarchical deep reinforcement learning scheme to simultaneously train the three networks. We then evaluated the effectiveness of the proposed framework in the treatment planning problems for prostate cancer intensity modulated RT (IMRT) and stereotactic body RT (SBRT). We benchmarked the performance of our approach by comparing plans made by VTPN of a parallel architecture, and the human plans submitted for competition in the 2016 American Association of Medical Dosimetrist (AAMD)/Radiosurgery Society (RSS) Plan Study. We analyzed scalability of the network size with respect to the number of TPPs. Numerical experiments were also performed to understand the rationale of the decision-making behaviors of the trained HieVTPN.Results.Both HieVTPNs for prostate IMRT and SBRT were trained successfully using 10 training patient cases and 5 validation cases. For IMRT, HieVTPN was able to generate high-quality plans for 59 testing patient cases that were not included in training process, achieving an average plan score of 8.62 (±0.83), with 9 being the maximal score. The score was comparable to that of the VTPN, 8.45 (±0.48). For SBRT planning, HieVTPN achieved an average plan score of 139.07 on five testing patient cases compared to the score of 132.21 averaged over the human plans summited for competition in AAMD/RSS plan study. Different from VTPN with network size linearly scaling with the number of TPPs, the network size of HieVTPN is almost independent of the number of TPPs. It was also observed that the decision-making behaviors of HieVTPN were understandable and generally agreed with the human experience.Conclusions.With the scalability and explainability, the hierarchical IATP framework is more favorable than the previous framework in terms of handling treatment planning problems involving a large number of TPPs.

Improving efficiency of training a virtual treatment planner network via knowledge-guided deep reinforcement learning for intelligent automatic treatment planning of radiotherapy

PURPOSE

We previously proposed an intelligent automatic treatment planning framework for radiotherapy, in which a virtual treatment planner network (VTPN) is built using deep reinforcement learning (DRL) to operate a treatment planning system (TPS) by adjusting treatment planning parameters in it to generate high-quality plans. We demonstrated the potential feasibility of this idea in prostate cancer intensity-modulated radiation therapy (IMRT). Despite the success, the process to train a VTPN via the standard DRL approach with an ϵ-greedy algorithm was time-consuming. The required training time was expected to grow with the complexity of the treatment planning problem, preventing the development of VTPN for more complicated but clinically relevant scenarios. In this study, we proposed a novel knowledge-guided DRL (KgDRL) approach that incorporated knowledge from human planners to guide the training process to improve the efficiency of training a VTPN.

METHOD

Using prostate cancer IMRT as a test bed, we first summarized a number of rules in the actions of adjusting treatment planning parameters of our in-house TPS. During the training process of VTPN, in addition to randomly navigating the large state-action space, as in the standard DRL approach using the ϵ-greedy algorithm, we also sampled actions defined by the rules. The priority of sampling actions from rules decreased over the training process to encourage VTPN to explore new policy on parameter adjustment that were not covered by the rules. To test this idea, we trained a VTPN using KgDRL and compared its performance with another VTPN trained using the standard DRL approach. Both networks were trained using 10 training patient cases and five additional cases for validation, while another 59 cases were employed for the evaluation purpose.

RESULTS

It was found that both VTPNs trained via KgDRL and standard DRL spontaneously learned how to operate the in-house TPS to generate high-quality plans, achieving plan quality scores of 8.82 (±0.29) and 8.43 (±0.48), respectively. Both VTPNs outperformed treatment planning purely based on the rules, which had a plan score of 7.81 (±1.59). VTPN trained with eight episodes using KgDRL was able to perform similar to that trained using DRL with 100 epochs. The training time was reduced from more than a week to ~13 hrs.

CONCLUSION

The proposed KgDRL framework was effective in accelerating the training process of a VTPN by incorporating human knowledge, which will facilitate the development of VTPN for more complicated treatment planning scenarios.

Closing the Gap Between Deep Neural Network Modeling and Biomedical Decision-Making Metrics in Segmentation via Adaptive Loss Functions

Deep learning is becoming an indispensable tool for various tasks in science and engineering. A critical step in constructing a reliable deep learning model is the selection of a loss function, which measures the discrepancy between the network prediction and the ground truth. While a variety of loss functions have been proposed in the literature, a truly optimal loss function that maximally utilizes the capacity of neural networks for deep learning-based decision-making has yet to be established. Here, we devise a generalized loss function with functional parameters determined adaptively during model training to provide a versatile framework for optimal neural network-based decision-making in small target segmentation. The method is showcased by more accurate detection and segmentation of lung and liver cancer tumors as compared with the current state-of-the-art. The proposed formalism opens new opportunities for numerous practical applications such as disease diagnosis, treatment planning, and prognosis.

Fluence map optimisation for prostate cancer intensity modulated radiotherapy planning using iterative solution method

Here we projected a model-based IMRT treatment plan to produce the optimal radiation dosage by considering that the maximum amount of prescribed dose should be delivered to the target without affecting the surrounding healthy tissues especially the OARs. Fluence mapping is used for inverse planning. This suggested method can generate global minima for IMRT plans with reliable plan quality among diverse treatment planners and to provide better safety for significant parallel OARs in an effective way. The whole methodology is having the capability to handles various objectives and to generate effective treatment procedures as validated with illustrations on the CORT dataset. For the validation of our methodology, we have compared our result with the two other approaches for calculating the objectives based on dose-volume bounds and found that in our methodology dose across the prostate and lymph nodes is maximum and the time required for the convergence is minimum.

Operating a treatment planning system using a deep-reinforcement learning-based virtual treatment planner for prostate cancer intensity-modulated radiation therapy treatment planning

PURPOSE

In the treatment planning process of intensity-modulated radiation therapy (IMRT), a human planner operates the treatment planning system (TPS) to adjust treatment planning parameters, for example, dose volume histogram (DVH) constraints' locations and weights, to achieve a satisfactory plan for each patient. This process is usually time-consuming, and the plan quality depends on planer's experience and available planning time. In this study, we proposed to model the behaviors of human planners in treatment planning by a deep reinforcement learning (DRL)-based virtual treatment planner network (VTPN), such that it can operate the TPS in a human-like manner for treatment planning.

METHODS AND MATERIALS

Using prostate cancer IMRT as an example, we established the VTPN using a deep neural network developed. We considered an in-house optimization engine with a weighted quadratic objective function. Virtual treatment planner network was designed to observe an intermediate plan DVHs and decide the action to improve the plan by changing weights and threshold dose in the objective function. We trained the VTPN in an end-to-end DRL process in 10 patient cases. A plan score was used to measure plan quality. We demonstrated the feasibility and effectiveness of the trained VTPN in another 64 patient cases.

RESULTS

Virtual treatment planner network was trained to spontaneously learn how to adjust treatment planning parameters to generate high-quality treatment plans. In the 64 testing cases, with initialized parameters, quality score was 4.97 (±2.02), with 9.0 being the highest possible score. Using VTPN to perform treatment planning improved quality score to 8.44 (±0.48).

CONCLUSIONS

To our knowledge, this was the first time that intelligent treatment planning behaviors of human planner in external beam IMRT are autonomously encoded in an artificial intelligence system. The trained VTPN is capable of behaving in a human-like way to produce high-quality plans.

Intelligent inverse treatment planning via deep reinforcement learning, a proof-of-principle study in high dose-rate brachytherapy for cervical cancer

Inverse treatment planning in radiation therapy is formulated as solving optimization problems. The objective function and constraints consist of multiple terms designed for different clinical and practical considerations. Weighting factors of these terms are needed to define the optimization problem. While a treatment planning optimization engine can solve the optimization problem with given weights, adjusting the weights to yield a high-quality plan is typically performed by a human planner. Yet the weight-tuning task is labor intensive, time consuming, and it critically affects the final plan quality. An automatic weight-tuning approach is strongly desired. The procedure of weight adjustment to improve the plan quality is essentially a decision-making problem. Motivated by the tremendous success in deep learning for decision making with human-level intelligence, we propose a novel framework to adjust the weights in a human-like manner. This study used inverse treatment planning in high-dose-rate brachytherapy (HDRBT) for cervical cancer as an example. We developed a weight-tuning policy network (WTPN) that observes dose volume histograms of a plan and outputs an action to adjust organ weighting factors, similar to the behaviors of a human planner. We trained the WTPN via end-to-end deep reinforcement learning. Experience replay was performed with the epsilon greedy algorithm. After training was completed, we applied the trained WTPN to guide treatment planning of five testing patient cases. It was found that the trained WTPN successfully learnt the treatment planning goals and was able to guide the weight tuning process. On average, the quality score of plans generated under the WTPN's guidance was improved by ~8.5% compared to the initial plan with arbitrarily set weights, and by 10.7% compared to the plans generated by human planners. To our knowledge, this was the first time that a tool was developed to adjust organ weights for the treatment planning optimization problem in a human-like fashion based on intelligence learnt from a training process, which was different from existing strategies based on pre-defined rules. The study demonstrated potential feasibility to develop intelligent treatment planning approaches via deep reinforcement learning.

Application programming in C# environment with recorded user software interactions and its application in autopilot of VMAT/IMRT treatment planning

An autopilot scheme of volumetric‐modulated arc therapy (VMAT)/intensity‐modulated radiation therapy (IMRT) planning with the guidance of prior knowledge is established with recorded interactions between a planner and a commercial treatment planning system (TPS). Microsoft (MS) Visual Studio Coded UI is applied to record some common planner‐TPS interactions as subroutines. The TPS used in this study is a Windows‐based Eclipse system. The interactions of our application program with Eclipse TPS are realized through a series of subroutines obtained by prerecording the mouse clicks or keyboard strokes of a planner in operating the TPS. A strategy to autopilot Eclipse VMAT/IMRT plan selection process is developed as a specific example of the proposed “scripting” method. The autopiloted planning is navigated by a decision function constructed with a reference plan that has the same prescription and similar anatomy with the case at hand. The calculation proceeds by alternating between the Eclipse optimization and the outer‐loop optimization independent of the Eclipse. In the C# program, the dosimetric characteristics of a reference treatment plan are used to assess and modify the Eclipse planning parameters and to guide the search for a clinically sensible treatment plan. The approach is applied to plan a head and neck (HN) VMAT case and a prostate IMRT case. Our study demonstrated the feasibility of application programming method in C# environment with recorded interactions of planner‐TPS. The process mimics a planner's planning process and automatically provides clinically sensible treatment plans that would otherwise require a large amount of manual trial and error of a planner. The proposed technique enables us to harness a commercial TPS by application programming via the use of recorded human computer interactions and provides an effective tool to greatly facilitate the treatment planning process.

PACS number(s): 87.55.D‐, 87.55.kd, 87.55.de

Physically constrained voxel-based penalty adaptation for ultra-fast IMRT planning

Conventional treatment planning in intensity-modulated radiation therapy (IMRT) is a trial-and-error process that usually involves tedious tweaking of optimization parameters. Here, we present an algorithm that automates part of this process, in particular the adaptation of voxel-based penalties within normal tissue. Thereby, the proposed algorithm explicitly considers a priori known physical limitations of photon irradiation. The efficacy of the developed algorithm is assessed during treatment planning studies comprising 16 prostate and 5 head and neck cases. We study the eradication of hot spots in the normal tissue, effects on target coverage and target conformity, as well as selected dose volume points for organs at risk. The potential of the proposed method to generate class solutions for the two indications is investigated. Run-times of the algorithms are reported. Physically constrained voxel-based penalty adaptation is an adequate means to automatically detect and eradicate hot-spots during IMRT planning while maintaining target coverage and conformity. Negative effects on organs at risk are comparably small and restricted to lower doses. Using physically constrained voxel-based penalty adaptation, it was possible to improve the generation of class solutions for both indications. Considering the reported run-times of less than 20 s, physically constrained voxel-based penalty adaptation has the potential to reduce the clinical workload during planning and automated treatment plan generation in the long run, facilitating adaptive radiation treatments.

A DVH-guided IMRT optimization algorithm for automatic treatment planning and adaptive radiotherapy replanning

PURPOSE

To develop a novel algorithm that incorporates prior treatment knowledge into intensity modulated radiation therapy optimization to facilitate automatic treatment planning and adaptive radiotherapy (ART) replanning.

METHODS

The algorithm automatically creates a treatment plan guided by the DVH curves of a reference plan that contains information on the clinician-approved dose-volume trade-offs among different targets/organs and among different portions of a DVH curve for an organ. In ART, the reference plan is the initial plan for the same patient, while for automatic treatment planning the reference plan is selected from a library of clinically approved and delivered plans of previously treated patients with similar medical conditions and geometry. The proposed algorithm employs a voxel-based optimization model and navigates the large voxel-based Pareto surface. The voxel weights are iteratively adjusted to approach a plan that is similar to the reference plan in terms of the DVHs. If the reference plan is feasible but not Pareto optimal, the algorithm generates a Pareto optimal plan with the DVHs better than the reference ones. If the reference plan is too restricting for the new geometry, the algorithm generates a Pareto plan with DVHs close to the reference ones. In both cases, the new plans have similar DVH trade-offs as the reference plans.

RESULTS

The algorithm was tested using three patient cases and found to be able to automatically adjust the voxel-weighting factors in order to generate a Pareto plan with similar DVH trade-offs as the reference plan. The algorithm has also been implemented on a GPU for high efficiency.

CONCLUSIONS

A novel prior-knowledge-based optimization algorithm has been developed that automatically adjust the voxel weights and generate a clinical optimal plan at high efficiency. It is found that the new algorithm can significantly improve the plan quality and planning efficiency in ART replanning and automatic treatment planning.

A multicriteria framework with voxel-dependent parameters for radiotherapy treatment plan optimization

PURPOSE

To establish a new mathematical framework for radiotherapy treatment optimization with voxel-dependent optimization parameters.

METHODS

In the treatment plan optimization problem for radiotherapy, a clinically acceptable plan is usually generated by an optimization process with weighting factors or reference doses adjusted for a set of the objective functions associated to the organs. Recent discoveries indicate that adjusting parameters associated with each voxel may lead to better plan quality. However, it is still unclear regarding the mathematical reasons behind it. Furthermore, questions about the objective function selection and parameter adjustment to assure Pareto optimality as well as the relationship between the optimal solutions obtained from the organ-based and voxel-based models remain unanswered. To answer these questions, the authors establish in this work a new mathematical framework equipped with two theorems.

RESULTS

The new framework clarifies the different consequences of adjusting organ-dependent and voxel-dependent parameters for the treatment plan optimization of radiation therapy, as well as the impact of using different objective functions on plan qualities and Pareto surfaces. The main discoveries are threefold: (1) While in the organ-based model the selection of the objective function has an impact on the quality of the optimized plans, this is no longer an issue for the voxel-based model since the Pareto surface is independent of the objective function selection and the entire Pareto surface could be generated as long as the objective function satisfies certain mathematical conditions; (2) All Pareto solutions generated by the organ-based model with different objective functions are parts of a unique Pareto surface generated by the voxel-based model with any appropriate objective function; (3) A much larger Pareto surface is explored by adjusting voxel-dependent parameters than by adjusting organ-dependent parameters, possibly allowing for the generation of plans with better trade-offs among different clinical objectives.

CONCLUSIONS

The authors have developed a mathematical framework for radiotherapy treatment optimization using voxel-based parameters. The authors can improve the plan quality by adjusting voxel-based weighting factors and exploring the unique and large Pareto surface which include all the Pareto surfaces that can be generated by organ-based model using different objective functions.

Automatic treatment plan re-optimization for adaptive radiotherapy guided with the initial plan DVHs

Adaptive radiation therapy (ART) can reduce normal tissue toxicity and/or improve tumor control through treatment adaptations based on the current patient anatomy. Developing an efficient and effective re-planning algorithm is an important step toward the clinical realization of ART. For the re-planning process, manual trial-and-error approach to fine-tune planning parameters is time-consuming and is usually considered unpractical, especially for online ART. It is desirable to automate this step to yield a plan of acceptable quality with minimal interventions. In ART, prior information in the original plan is available, such as dose-volume histogram (DVH), which can be employed to facilitate the automatic re-planning process. The goal of this work is to develop an automatic re-planning algorithm to generate a plan with similar, or possibly better, DVH curves compared with the clinically delivered original plan. Specifically, our algorithm iterates the following two loops. An inner loop is the traditional fluence map optimization, in which we optimize a quadratic objective function penalizing the deviation of the dose received by each voxel from its prescribed or threshold dose with a set of fixed voxel weighting factors. In outer loop, the voxel weighting factors in the objective function are adjusted according to the deviation of the current DVH curves from those in the original plan. The process is repeated until the DVH curves are acceptable or maximum iteration step is reached. The whole algorithm is implemented on GPU for high efficiency. The feasibility of our algorithm has been demonstrated with three head-and-neck cancer IMRT cases, each having an initial planning CT scan and another treatment CT scan acquired in the middle of treatment course. Compared with the DVH curves in the original plan, the DVH curves in the resulting plan using our algorithm with 30 iterations are better for almost all structures. The re-optimization process takes about 30 s using our in-house optimization engine.

Fluence map optimization (FMO) with dose-volume constraints in IMRT using the geometric distance sorting method

A new heuristic algorithm based on the so-called geometric distance sorting technique is proposed for solving the fluence map optimization with dose-volume constraints which is one of the most essential tasks for inverse planning in IMRT. The framework of the proposed method is basically an iterative process which begins with a simple linear constrained quadratic optimization model without considering any dose-volume constraints, and then the dose constraints for the voxels violating the dose-volume constraints are gradually added into the quadratic optimization model step by step until all the dose-volume constraints are satisfied. In each iteration step, an interior point method is adopted to solve each new linear constrained quadratic programming. For choosing the proper candidate voxels for the current dose constraint adding, a so-called geometric distance defined in the transformed standard quadratic form of the fluence map optimization model was used to guide the selection of the voxels. The new geometric distance sorting technique can mostly reduce the unexpected increase of the objective function value caused inevitably by the constraint adding. It can be regarded as an upgrading to the traditional dose sorting technique. The geometry explanation for the proposed method is also given and a proposition is proved to support our heuristic idea. In addition, a smart constraint adding/deleting strategy is designed to ensure a stable iteration convergence. The new algorithm is tested on four cases including head-neck, a prostate, a lung and an oropharyngeal, and compared with the algorithm based on the traditional dose sorting technique. Experimental results showed that the proposed method is more suitable for guiding the selection of new constraints than the traditional dose sorting method, especially for the cases whose target regions are in non-convex shapes. It is a more efficient optimization technique to some extent for choosing constraints than the dose sorting method. By integrating a smart constraint adding/deleting scheme within the iteration framework, the new technique builds up an improved algorithm for solving the fluence map optimization with dose-volume constraints.

Toward truly optimal IMRT dose distribution: inverse planning with voxel-specific penalty

PURPOSE

To establish an inverse planning framework with adjustable voxel penalty for more conformal IMRT dose distribution as well as improved interactive controllability over the regional dose distribution of the resultant plan.

MATERIALS AND METHOD

In the proposed coarse-to-fine planning scheme, a conventional inverse planning with organ specific parameters is first performed. The voxel penalty scheme is then "switched on" by allowing the prescription dose to change on an individual voxel scale according to the deviation of the actual voxel dose from the ideally desired dose. The rationale here is intuitive: when the dose at a voxel does not meet its ideal dose, it simply implies that this voxel is not competitive enough when compared with the ones that have met their planning goal. In this case, increasing the penalty of the voxel by varying the prescription can boost its competitiveness and thus improve its dose. After the prescription adjustment, the plan is re-optimized. The dose adjustment/re-optimization procedure is repeated until the resultant dose distribution cannot be improved anymore. The prescription adjustment on a finer scale can be accomplished either automatically or manually. In the latter case, the regions/voxels where a dose improvement is needed are selected visually, unlike in the automatic case where the selection is done purely based on the difference of the actual dose at a given voxel and its ideal prescription. The performance of the proposed method is evaluated using a head and neck and a prostate case.

RESULTS

An inverse planning framework with the voxel-specific penalty is established. By adjusting voxel prescriptions iteratively to boost the region where large mismatch between the actual calculated and desired doses occurs, substantial improvements can be achieved in the final dose distribution. The proposed method is applied to a head and neck case and a prostate case. For the former case, a significant reduction in the maximum dose to the brainstem is achieved while the PTV dose coverage is greatly improved. The doses to other organs at risk are also reduced, ranging from 10% to 30%. For the prostate case, the use of the voxel penalty scheme also results in vast improvements to the final dose distribution. The PTV experiences improved dose uniformity and the mean dose to the rectum and bladder is reduced by as much as 15%.

CONCLUSION

Introduction of the spatially non-uniform and adjustable prescription provides room for further improvements of currently achievable dose distributions and equips the planner with an effective tool to modify IMRT dose distributions interactively. The technique is easily implementable in any existing inverse planning platform, which should facilitate clinical IMRT planning process and, in future, off-line/on-line adaptive IMRT.

The equivalence of multi-criteria methods for radiotherapy plan optimization

Several methods can be used to achieve multi-criteria optimization of radiation therapy treatment planning, which strive for Pareto-optimality. The property of the solution being Pareto optimal is desired, because it guarantees that no criteria can be improved without deteriorating another criteria. The most widely used methods are the weighted-sum method, in which the different treatment objectives are weighted, and constrained optimization methods, in which treatment goals are set and the algorithm has to find the best plan fulfilling these goals. The constrained method used in this paper, the 2p element of c (2-phase element-constraint) method is based on the element-constraint method, which generates Pareto-optimal solutions. Both approaches are uniquely related to each other. In this paper, we will show that it is possible to switch from the constrained method to the weighted-sum method by using the Lagrange multipliers from the constrained optimization problem, and vice versa by setting the appropriate constraints. In general, the theory presented in this paper can be useful in cases where a new situation is slightly different from the original situation, e.g. in online treatment planning, with deformations of the volumes of interest, or in automated treatment planning, where changes to the automated plan have to be made. An example of the latter is given where the planner is not satisfied with the result from the constrained method and wishes to decrease the dose in a structure. By using the Lagrange multipliers, a weighted-sum optimization problem is constructed, which generates a Pareto-optimal solution in the neighbourhood of the original plan, but fulfills the new treatment objectives.

A novel approach to multi-criteria inverse planning for IMRT

Treatment plan optimization is a multi-criteria process. Optimizing solely on one objective or on a sum of a priori weighted objectives may result in inferior treatment plans. Manually adjusting weights or constraints in a trial and error procedure is time consuming. In this paper we introduce a novel multi-criteria optimization approach to automatically optimize treatment constraints (dose-volume and maximum-dose). The algorithm tries to meet these constraints as well as possible, but in the case of conflicts it relaxes lower priority constraints so that higher priority constraints can be met. Afterwards, all constraints are tightened, starting with the highest priority constraints. Applied constraint priority lists can be used as class solutions for patients with similar tumour types. The presented algorithm does iteratively apply an underlying algorithm for beam profile optimization, based on a quadratic objective function with voxel-dependent importance factors. These voxel-dependent importance factors are automatically adjusted to reduce dose-volume and maximum-dose constraint violations.

Comparison of intensity modulated x-ray therapy and intensity modulated proton therapy for selective subvolume boosting: a phantom study

Selective subvolume boosting can theoretically improve tumour control probability while maintaining normal tissue complication probabilities similar to those of uniform dose distributions. In this work the abilities of intensity-modulated x-ray therapy (IMXT) and intensity-modulated proton therapy (IMPT) to deliver boosts to multiple subvolumes of varying size and proximities are compared in a thorough phantom study. IMXT plans were created using the step-and-shoot (IMXT-SAS) and helical tomotherapy (IMXT-HT) methods. IMPT plans were created with the spot scanning (IMPT-SS) and distal gradient tracking (IMPT-DGT) methods. IMPT-DGT is a generalization of the distal edge tracking method designed to reduce the number of proton beam spots required to deliver non-uniform dose distributions relative to IMPT-SS. The IMPT methods were delivered over both 180 degrees and 360 degrees arcs. The IMXT-SAS and IMPT-SS methods optimally satisfied the non-uniform dose prescriptions the least and the most, respectively. The IMPT delivery methods reduced the normal tissue integral dose by a factor of about 2 relative to the IMXT delivery methods, regardless of the delivery arc. The IMPT-DGT method reduced the number of proton beam spots by a factor of about 3 relative to the IMPT-SS method.

Formulating adaptive radiation therapy (ART) treatment planning into a closed-loop control framework

While ART has been studied for years, the specific quantitative implementation details have not. In order for this new scheme of radiation therapy (RT) to reach its potential, an effective ART treatment planning strategy capable of taking into account the dose delivery history and the patient's on-treatment geometric model must be in place. This paper performs a theoretical study of dynamic closed-loop control algorithms for ART and compares their utility with data from phantom and clinical cases. We developed two classes of algorithms: those Adapting to Changing Geometry and those Adapting to Geometry and Delivered Dose. The former class takes into account organ deformations found just before treatment. The latter class optimizes the dose distribution accumulated over the entire course of treatment by adapting at each fraction, not only to the information just before treatment about organ deformations but also to the dose delivery history. We showcase two algorithms in the class of those Adapting to Geometry and Delivered Dose. A comparison of the approaches indicates that certain closed-loop ART algorithms may significantly improve the current practice. We anticipate that improvements in imaging, dose verification and reporting will further increase the importance of adaptive algorithms.

IMRT treatment planning based on prioritizing prescription goals

Determining the 'best' optimization parameters in IMRT planning is typically a time-consuming trial-and-error process with no unambiguous termination point. Recently we and others proposed a goal-programming approach which better captures the desired prioritization of dosimetric goals. Here, individual prescription goals are addressed stepwise in their order of priority. In the first step, only the highest order goals are considered (target coverage and dose-limiting normal structures). In subsequent steps, the achievements of the previous steps are turned into hard constraints and lower priority goals are optimized, in turn, subject to higher priority constraints. So-called 'slip' factors were introduced to allow for slight, clinically acceptable violations of the constraints. Focusing on head and neck cases, we present several examples for this planning technique. The main advantages of the new optimization method are (i) its ability to generate plans that meet the clinical goals, as well as possible, without tuning any weighting factors or dose-volume constraints, and (ii) the ability to conveniently include more terms such as fluence map smoothness. Lower level goals can be optimized to the achievable limit without compromising higher order goals. The prioritized prescription-goal planning method allows for a more intuitive and human-time-efficient way of dealing with conflicting goals compared to the conventional trial-and-error method of varying weighting factors and dose-volume constraints.

Overview of image-guided radiation therapy

Radiation therapy has gone through a series of revolutions in the last few decades and it is now possible to produce highly conformal radiation dose distribution by using techniques such as intensity-modulated radiation therapy (IMRT). The improved dose conformity and steep dose gradients have necessitated enhanced patient localization and beam targeting techniques for radiotherapy treatments. Components affecting the reproducibility of target position during and between subsequent fractions of radiation therapy include the displacement of internal organs between fractions and internal organ motion within a fraction. Image-guided radiation therapy (IGRT) uses advanced imaging technology to better define the tumor target and is the key to reducing and ultimately eliminating the uncertainties. The purpose of this article is to summarize recent advancements in IGRT and discussed various practical issues related to the implementation of the new imaging techniques available to radiation oncology community. We introduce various new IGRT concepts and approaches, and hope to provide the reader with a comprehensive understanding of the emerging clinical IGRT technologies. Some important research topics will also be addressed.

Multiobjective optimization with a modified simulated annealing algorithm for external beam radiotherapy treatment planning

Inverse planning in external beam radiotherapy often requires a scalar objective function that incorporates importance factors to mimic the planner's preferences between conflicting objectives. Defining those importance factors is not straightforward, and frequently leads to an iterative process in which the importance factors become variables of the optimization problem. In order to avoid this drawback of inverse planning, optimization using algorithms more suited to multiobjective optimization, such as evolutionary algorithms, has been suggested. However, much inverse planning software, including one based on simulated annealing developed at our institution, does not include multiobjective-oriented algorithms. This work investigates the performance of a modified simulated annealing algorithm used to drive aperture-based intensity-modulated radiotherapy inverse planning software in a multiobjective optimization framework. For a few test cases involving gastric cancer patients, the use of this new algorithm leads to an increase in optimization speed of a little more than a factor of 2 over a conventional simulated annealing algorithm, while giving a close approximation of the solutions produced by a standard simulated annealing. A simple graphical user interface designed to facilitate the decision-making process that follows an optimization is also presented.

Fast, multiple optimizations of quadratic dose objective functions in IMRT

Inverse treatment planning for intensity-modulated radiotherapy may include time consuming, multiple minimizations of an objective function. In this paper, methods are presented to speed up the process of (repeated) minimization of the well-known quadratic dose objective function, extended with a smoothing term that ensures generation of clinically acceptable beam profiles. In between two subsequent optimizations, the voxel-dependent importance factors of the quadratic terms will generally be adjusted, based on an intermediate plan evaluation. The objective function has been written in matrix-vector format, facilitating the use of a recently published, fast quadratic minimization algorithm, instead of commonly applied gradient-based methods. This format also reduces the calculation time in between subsequent minimizations, related to adjustment of the voxel-dependent importance factors. Sparse matrices are used to limit the required amount of computer memory. For three patients, comparisons have been made with a gradient method. Mean speed improvements of up to a factor of 37 have been achieved.

The IMRT information process—mastering the degrees of freedom in external beam therapy

The techniques and procedures for intensity-modulated radiation therapy (IMRT) are reviewed in the context of the information process central to treatment planning and delivery of IMRT. A presentation is given of the evolution of the information based radiotherapy workflow and dose delivery techniques, as well as the volume and planning concepts for relating the dose information to image based patient representations. The formulation of the dose shaping process as an optimization problem is described. The different steps in the calculation flow for determination of machine parameters for dose delivery are described starting from the formulation of optimization objectives over dose calculation to optimization procedures. Finally, the main elements of the quality assurance procedure necessary for implementing IMRT clinically are reviewed.