Search

Purposes: This study aimed to develop and validate algorithms for automating intensity modulated radiation therapy (IMRT) planning in breast cancer patients, with a focus on patient anatomical characteristics., Material and Methods: We retrospectively selected 400 breast cancer patients without lymph node involvement for automated treatment planning. Automation was achieved using the Eclipse Scripting Application Programming Interface (ESAPI) integrated into the Eclipse Treatment Planning System. We employed three beam insertion geometries and three optimization strategies, resulting in 3600 plans, each delivering a 40.05 Gy dose in 15 fractions. Gantry angles in the tangent fields were selected based on a criterion involving the minimum intersection area between the Planning Target Volume (PTV) and the ipsilateral lung in the Beam's Eye View projection. ESAPI was also used to gather patient anatomical data, serving as input for Random Forest models to select the optimal plan. The Random Forest classification considered both beam insertion geometry and optimization strategy. Dosimetric data were evaluated in accordance with the Radiation Therapy Oncology Group (RTOG) 1005 protocol., Results: Overall, all approaches generated high-quality plans, with approximately 94% meeting the acceptable dose criteria for organs at risk and/or target coverage as defined by RTOG guidelines. Average automated plan generation time ranged from 6 min and 37 s to 9 min and 22 s, with the mean time increasing with additional fields. The Random Forest approach did not successfully enable automatic planning strategy selection. Instead, our automated planning system allows users to choose from the tested geometry and strategy options., Conclusions: Although our attempt to correlate patient anatomical features with planning strategy using machine learning tools was unsuccessful, the resulting dosimetric outcomes proved satisfactory. Our algorithm consistently produced high-quality plans, offering significant time and efficiency advantages., (© 2024 The Authors. Journal of Applied Clinical Medical Physics is published by Wiley Periodicals, Inc. on behalf of The American Association of Physicists in Medicine.)

Purpose: Focal epilepsy is a neurological disease that can be surgically treated by removing area of the brain generating the seizures. The stereotactic electroencephalography (SEEG) procedure allows patient brain activity to be recorded in order to localize the onset of seizures through the placement of intracranial electrodes. The planning phase can be cumbersome and very time consuming, and no quantitative information is provided to neurosurgeons regarding the safety and efficacy of their trajectories. In this work, we present a novel architecture specifically designed to ease the SEEG trajectory planning using the 3D Slicer platform as a basis., Methods: Trajectories are automatically optimized following criteria like vessel distance and insertion angle. Multi-trajectory optimization and conflict resolution are optimized through a selective brute force approach based on a conflict graph construction. Additionally, electrode-specific optimization constraints can be defined, and an advanced verification module allows neurosurgeons to evaluate the feasibility of the trajectory., Results: A retrospective evaluation was performed using manually planned trajectories on 20 patients: the planning algorithm optimized and improved trajectories in 98% of cases. We were able to resolve and optimize the remaining 2% by applying electrode-specific constraints based on manual planning values. In addition, we found that the global parameters used discards 68% of the manual planned trajectories, even when they represent a safe clinical choice., Conclusions: Our approach improved manual planned trajectories in 98% of cases in terms of quantitative indexes, even when applying more conservative criteria with respect to actual clinical practice. The improved multi-trajectory strategy overcomes the previous work limitations and allows electrode optimization within a tolerable time span.

In inverse treatment planning of intensity-modulated radiation therapy (IMRT), the objective function is typically the sum of the weighted sub-scores, where the weights indicate the importance of the sub-scores. To obtain a high-quality treatment plan, the planner manually adjusts the objective weights using a trial-and-error procedure until an acceptable plan is reached. In this work, a new particle swarm optimization (PSO) method which can adjust the weighting factors automatically was investigated to overcome the requirement of manual adjustment, thereby reducing the workload of the human planner and contributing to the development of a fully automated planning process. The proposed optimization method consists of three steps. (i) First, a swarm of weighting factors (i.e., particles) is initialized randomly in the search space, where each particle corresponds to a global objective function. (ii) Then, a plan optimization solver is employed to obtain the optimal solution for each particle, and the values of the evaluation functions used to determine the particle's location and the population global location for the PSO are calculated based on these results. (iii) Next, the weighting factors are updated based on the particle's location and the population global location. Step (ii) is performed alternately with step (iii) until the termination condition is reached. In this method, the evaluation function is a combination of several key points on the dose volume histograms. Furthermore, a perturbation strategy - the crossover and mutation operator hybrid approach - is employed to enhance the population diversity, and two arguments are applied to the evaluation function to improve the flexibility of the algorithm. In this study, the proposed method was used to develop IMRT treatment plans involving five unequally spaced 6MV photon beams for 10 prostate cancer cases. The proposed optimization algorithm yielded high-quality plans for all of the cases, without human planner intervention. A comparison of the results with the optimized solution obtained using a similar optimization model but with human planner intervention revealed that the proposed algorithm produced optimized plans superior to that developed using the manual plan. The proposed algorithm can generate admissible solutions within reasonable computational times and can be used to develop fully automated IMRT treatment planning methods, thus reducing human planners' workloads during iterative processes., (Copyright © 2017 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.)

Background: To investigate the capability of a not-yet commercially available fully automated lexicographic optimization (LO) planning algorithm, called mCycle (Elekta AB, Stockholm, Sweden), to further improve the plan quality of an already-validated Wish List (WL) pushing on the organs-at-risk (OAR) sparing without compromising target coverage and plan delivery accuracy. Material and Methods: Twenty-four mono-institutional consecutive cervical cancer Volumetric-Modulated Arc Therapy (VMAT) plans delivered between November 2019 and April 2022 (50 Gy/25 fractions) have been retrospectively selected. In mCycle the LO planning algorithm was combined with the a-priori multi-criterial optimization (MCO). Two versions of WL have been defined to reproduce manual plans (WL01), and to improve the OAR sparing without affecting minimum target coverage and plan delivery accuracy (WL02). Robust WLs have been tuned using a subset of 4 randomly selected patients. The remaining plans have been automatically re-planned by using the designed WLs. Manual plans (MP) and mCycle plans (mCP01 and mCP02) were compared in terms of dose distributions, complexity, delivery accuracy, and clinical acceptability. Two senior physicians independently performed a blind clinical evaluation, ranking the three competing plans. Furthermore, a previous defined global quality index has been used to gather into a single score the plan quality evaluation. Results: The WL tweaking requests 5 and 3 working days for the WL01 and the WL02, respectively. The re-planning took in both cases 3 working days. mCP01 best performed in terms of target coverage (PTV V95% (%): MP 98.0 [95.6–99.3], mCP01 99.2 [89.7–99.9], mCP02 96.9 [89.4–99.5]), while mCP02 showed a large OAR sparing improvement, especially in the rectum parameters (e.g., Rectum D50% (Gy): MP 41.7 [30.2–47.0], mCP01 40.3 [31.4–45.8], mCP02 32.6 [26.9–42.6]). An increase in plan complexity has been registered in mCPs without affecting plan delivery accuracy. In the blind comparisons, all automated plans were considered clinically acceptable, and mCPs were preferred over MP in 90% of cases. Globally, automated plans registered a plan quality score at least comparable to MP. Conclusions: This study showed the flexibility of the Lexicographic approach in creating more demanding Wish Lists able to potentially minimize toxicities in RT plans. [ABSTRACT FROM AUTHOR]

PurposeThe accuracy of dose prediction is essential for knowledge-based planning and automated planning techniques. We compare the dose prediction accuracy of 3 prediction methods including statistical voxel dose learning, spectral regression, and support vector regression based on limited patient training data.MethodsStatistical voxel dose learning, spectral regression, and support vector regression were used to predict the dose of noncoplanar intensity-modulated radiation therapy (4π) and volumetric-modulated arc therapy head and neck, 4π lung, and volumetric-modulated arc therapy prostate plans. Twenty cases of each site were used for k-fold cross-validation, with k = 4. Statistical voxel dose learning bins voxels according to their Euclidean distance to the planning target volume and uses the median to predict the dose of new voxels. Distance to the planning target volume, polynomial combinations of the distance components, planning target volume, and organ at risk volume were used as features for spectral regression and support vector regression. A total of 28 features were included. Principal component analysis was performed on the input features to test the effect of dimension reduction. For the coplanar volumetric-modulated arc therapy plans, separate models were trained for voxels within the same axial slice as planning target volume voxels and voxels outside the primary beam. The effect of training separate models for each organ at risk compared to all voxels collectively was also tested. The mean squared error was calculated to evaluate the voxel dose prediction accuracy.ResultsStatistical voxel dose learning using separate models for each organ at risk had the lowest root mean squared error for all sites and modalities: 3.91 Gy (head and neck 4π), 3.21 Gy (head and neck volumetric-modulated arc therapy), 2.49 Gy (lung 4π), and 2.35 Gy (prostate volumetric-modulated arc therapy). Compared to using the original features, principal component analysis reduced the 4π prediction error for head and neck spectral regression (-43.9%) and support vector regression (-42.8%) and lung support vector regression (-24.4%) predictions. Principal component analysis was more effective in using all/most of the possible principal components. Separate organ at risk models were more accurate than training on all organ at risk voxels in all cases.ConclusionCompared with more sophisticated parametric machine learning methods with dimension reduction, statistical voxel dose learning is more robust to patient variability and provides the most accurate dose prediction method.

Solver competitions have been used in many areas of AI to assess the current state of the art and guide future research and development. AI planning is no exception, and the International Planning Competition (IPC) has been frequently run for nearly two decades. Due to the organisational and computational burden involved in running these competitions, solvers are generally compared using a single homogeneous hardware and software environment for all competitors. To what extent does the specific choice of hardware and software environment have an effect on solver performance, and is that effect distributed equally across the competing solvers? In this work, we use the competing planners and benchmark instance sets from the 2014 IPC to investigate these two questions. We recreate the 2014 IPC Optimal and Agile tracks on two distinct hardware environments and eight distinct software environments. We show that solver performance varies significantly based on the hardware and software environment, and that this variation is not equal for all planners. Furthermore, the observed variation is sufficient to change the competition rankings, including the top-ranked planners for some tracks. [ABSTRACT FROM AUTHOR]

Purpose/objectives: The complex planning and quality assurance required for spine SBRT are a barrier to implementation in time-sensitive or limited resource clinical situations. We developed and validated an automated inverse planning algorithm designed to streamline planning and allow rapid delivery of conformal single fraction spine SBRT using widely available technology. Materials/methods: The Rapid Spine (RaSp) automated script successfully generated single fraction SBRT plans for fourteen complex spinal lesions previously treated at a single high-volume institution. Automated RaSp plans were limited to 5 beams with a total of 15 segments (allowing calculation-based verification) and optimized based on RTOG 0631 objectives. Standard single fraction (16 Gy) stereotactic IMRT plans were generated for the same set of complex spinal lesions and used for comparison. A conservative 2 mm posterior isocenter shift was used to simulate minor set-up error. Results: Automated plans were generated in under 5 min from target definition and had a mean dose to the PTV of 1663 cGy (SD 131.5), a dose to 90 % of PTV (D90) of 1358 cGy (SD 111.0), and a maximum point dose (Dmax) to the PTV of 2055 cGy (SD 195.2) on average. IMRT plans took longer to generate but yielded more favorable dose escalation with a mean dose to the PTV of 1891 cGy (SD 117.6), D90 of 1731 cGy (SD 126.5), and Dmax of 2218 cGy (SD 195.7). A 2 mm posterior shift resulted in a 20 % (SD 10.5 %) increase in cord dose for IMRT plans and a 10 % (SD 5.3 %) increase for RaSp plans. The 2 mm perturbation caused 3 cord dose violations for the IMRT plans and 1 violation for corresponding RaSp plans. Conclusion: The Rapid Spine plan method yields timely and dosimetrically reasonable SBRT plans which meet RTOG 0631 objectives and are suitable for rapid yet robust pretreatment quality assurance followed by expedited treatment delivery. RaSp plans reduce the tradeoff between rapid treatment and optimal dosimetry in urgent cases and limited resource situations. [ABSTRACT FROM AUTHOR]

Automated planning is a well-established field of artificial intelligence (AI), with applications in route finding, robotics and operational research, among others. The task of developing a plan is often solved by finding a path in a graph representing the search domain; a robust plan consists of numerous paths that can be chosen if the execution of the best (optimal) one fails. While robust planning for a single entity is rather simple, development of a robust plan for multiple entities in a common environment can lead to combinatorial explosion. This paper proposes a novel hybrid approach, joining heuristic search and the wavefront algorithm to provide a plan featuring robustness in areas where it is needed, while maintaining a low level of computational complexity. [ABSTRACT FROM AUTHOR]

One of the main challenges in model-driven software engineering is to automate the resolution of design model inconsistencies. We propose to use the artificial intelligence technique of automated planning for the purpose of resolving such inconsistencies through the generation of one or more resolution plans. We implemented Badger, a regression planner in Prolog that generates such plans. We assess its scalability on the resolution of different types of structural inconsistencies in UML models using both generated models and reverse-engineered models of varying sizes, the largest ones containing more than 10,000 model elements. We illustrate the metamodel-independence of our approach by applying it to the resolution of code smells in a Java program. We discuss how the user can adapt the order in which resolution plans are presented by modifying the cost function of the planner algorithm. [ABSTRACT FROM AUTHOR]

Despite the advances made in the last decade in automated planning, no planner outperforms all the others in every known benchmark domain. This observation motivates the idea of selecting different planning algorithms for different domains. Moreover, the planners' performances are affected by the structure of the search space, which depends on the encoding of the considered domain. In many domains, the performance of a planner can be improved by exploiting additional knowledge, for instance, in the form of macro-operators or entanglements. In this paper we propose ASAP, an automatic Algorithm Selection Approach for Planning that: (i) for a given domain initially learns additional knowledge, in the form of macro-operators and entanglements, which is used for creating different encodings of the given planning domain and problems, and (ii) explores the 2 dimensional space of available algorithms, defined as encodings-planners couples, and then (iii) selects the most promising algorithm for optimising either the runtimes or the quality of the solution plans. [ABSTRACT FROM AUTHOR]

We describe a planning algorithm, NDP2, that finds strong-cyclic solutions to nondeterministic planning problems by using a classical planner to solve a sequence of classical planning problems. NDP2 is provably correct, and fixes several problems with prior work. We also describe two preprocessing algorithms that can provide a restricted version of the symbolic abstraction capabilities of the well-known MBP planner. The preprocessing algorithms accomplish this by rewriting the planning problems, hence do not require any modifications to NDP2 or its classical planner. In our experimental comparisons of NDP2 (using FF as the classical planner) to MBP in six different planning domains, each planner outperformed the other in some domains but not others. Which planner did better depended on three things: the amount of nondeterminism in the planning domain, domain characteristics that affected how well the abstraction techniques worked, and whether the domain contained unsolvable states. [ABSTRACT FROM AUTHOR]

Abstract: Increasing complexity of petro-chemical projects, fast-tracked engineering, and tighter schedules pose a challenge for pipe spool fabrication shops. To maintain competitiveness, it is necessary to improve the shop performance, i.e. fabrication cycle time. Pipe spool fabrication sequence is found to have significant impact on cycle time and presents an area with potential for improvement. Traditionally, fabrication sequences are determined by shop foremen in a heuristic manner and optimality is not guaranteed. This paper presents a dynamic programming (DP) algorithm to automatically identify the optimal fabrication sequences for pipe spools. Simulation experiments are conducted to test the effectiveness of the algorithm by comparing the cycle times resulting from the algorithm-generated sequences and human-planner-designed sequences, respectively. The results show that the DP algorithm reduces unnecessary position-welding (9 out of 20 position-welds) and results in a reduction in the total fabrication cycle time by a range of 4.8% to 12%. [Copyright &y& Elsevier]

Most of the great success of heuristic search as an approach to AI Planning is due to the right design of domain-independent heuristics. Although many heuristic planners perform reasonably well, the computational cost of computing the heuristic function in every search node is very high, causing the planner to scale poorly when increasing the size of the planning tasks. For tackling this problem, planners can incorporate additional domain-dependent heuristics in order to improve their performance. Learning-based planners try to automatically acquire these domain-dependent heuristics using previous solved problems. In this work, we present a case-based reasoning approach that learns abstracted state transitions that serve as domain control knowledge for improving the planning process. The recommendations from the retrieved cases are used as guidance for pruning or ordering nodes in different heuristic search algorithms applied to planning tasks. We show that the CBR guidance is appropriate for a considerable number of planning benchmarks. [ABSTRACT FROM AUTHOR]

Abstract: The utility of including loops in plans has been long recognized by the planning community. Loops in a plan help increase both its applicability and the compactness of its representation. However, progress in finding such plans has been limited largely due to lack of methods for reasoning about the correctness and safety properties of loops of actions. We present novel algorithms for determining the applicability and progress made by a general class of loops of actions. These methods can be used for directing the search for plans with loops towards greater applicability while guaranteeing termination, as well as in post-processing of computed plans to precisely characterize their applicability. Experimental results demonstrate the efficiency of these algorithms. We also discuss the factors which can make the problem of determining applicability conditions for plans with loops incomputable. [Copyright &y& Elsevier]

Project management tasks, such as productivity monitoring and cost estimation, require data to be fused from multiple data sources. Data fusion approaches incorporated in the existing research studies within the construction management domain support a specific task or a decision (e.g., labor productivity monitoring or defect detection). Hence, most of the previously developed approaches do not necessarily support tasks other than the ones that they were intended for. This paper describes an automated planning approach as a general way to fuse data from multiple sources to support construction productivity monitoring tasks. A prototype system, which incorporates two planning algorithms, was developed to validate the generality of the approach on the basis of representative queries of construction engineers and managers identified in previous research studies. [ABSTRACT FROM AUTHOR]

Automated planning was proposed for military system-of-systems vulnerability analysis. Goal, capability, state and action were used as the conceptual framework for military system-of-systems description. Vulnerability analysis was reduced to problems concerning goal reachability and every-possible plan generation. A modification was made to the search strategy of graph planning algorithm. In the new algorithm, search in and/or graph is transformed to depth-first search in a tree with goals and actions as its nodes. The algorithm is capable of generating a complete set of courses of action in the sense of expelling aI1 redundancy actions. Correctness and completeness of the algorithm were proved. The algorithm is efficient when both the number of actions and the number of preconditions of actions are small. An analysis example shows that this method is feasible for system-of-systems vulnerability analysis. But further research should be conducted to limit search space of large scale problems by applying additional constraints. [ABSTRACT FROM AUTHOR]

Abstract: Constructing plans that can handle multiple problem instances is a longstanding open problem in AI. We present a framework for generalized planning that captures the notion of algorithm-like plans and unifies various approaches developed for addressing this problem. Using this framework, and building on the TVLA system for static analysis of programs, we develop a novel approach for computing generalizations of classical plans by identifying sequences of actions that will make measurable progress when placed in a loop. In a wide class of problems that we characterize formally in the paper, these methods allow us to find generalized plans with loops for solving problem instances of unbounded sizes and also to determine the correctness and applicability of the computed generalized plans. We demonstrate the scope and scalability of the proposed approach on a wide range of planning problems. [ABSTRACT FROM AUTHOR]

Abstract: Automated planning requires action models described using languages such as the Planning Domain Definition Language (PDDL) as input, but building action models from scratch is a very difficult and time-consuming task, even for experts. This is because it is difficult to formally describe all conditions and changes, reflected in the preconditions and effects of action models. In the past, there have been algorithms that can automatically learn simple action models from plan traces. However, there are many cases in the real world where we need more complicated expressions based on universal and existential quantifiers, as well as logical implications in action models to precisely describe the underlying mechanisms of the actions. Such complex action models cannot be learned using many previous algorithms. In this article, we present a novel algorithm called LAMP (Learning Action Models from Plan traces), to learn action models with quantifiers and logical implications from a set of observed plan traces with only partially observed intermediate state information. The LAMP algorithm generates candidate formulas that are passed to a Markov Logic Network (MLN) for selecting the most likely subsets of candidate formulas. The selected subset of formulas is then transformed into learned action models, which can then be tweaked by domain experts to arrive at the final models. We evaluate our approach in four planning domains to demonstrate that LAMP is effective in learning complex action models. We also analyze the human effort saved by using LAMP in helping to create action models through a user study. Finally, we apply LAMP to a real-world application domain for software requirement engineering to help the engineers acquire software requirements and show that LAMP can indeed help experts a great deal in real-world knowledge-engineering applications. [ABSTRACT FROM AUTHOR]

Abstract: Although new AI planning algorithms and techniques are being developed and improved rapidly, there is a lack of efficient and easy to use systems able to incorporate and utilize them. Furthermore, while visual representation facilitates design, maintenance and comprehension of planning domains and problems, very few systems incorporate it. This paper presents VLEPPO, an integrated system aiming at visually modeling planning domains and problems through a convenient graphical interface, while maintaining compatibility with the Planning Domain Definition Language (PDDL), with import and export features. Solutions to planning problems can be obtained by invoking different planners employing the web services technology. The demonstration of the system is performed through a case study involving web service composition viewed as a planning problem. [Copyright &y& Elsevier]

Traditional planning assumes reachability goals and/or full observability. In this paper, we propose a novel solution for safety and reachability planning with partial observability. Given a planning domain, a safety property, and a reachability goal, we automatically learn a safe permissive plan to guide the planning domain so that the safety property is not violated and that can force the planning domain to eventually reach states that satisfy the reachability goal, regardless of how the planning domain behaves. Our technique is based on the active learning of regular languages and symbolic model checking. The planning method first learns a safe plan using the L* algorithm, which is an efficient active learning algorithm for regular languages. We then check whether the safe plan learned is also permissive by Alternating-time Temporal Logic (ATL) model checking. If the plan is permissive, it is indeed a safe permissive plan. Otherwise, we identify and add a safe string to converge a safe permissive plan. We describe an implementation of the proposed technique and demonstrate that our tool can efficiently construct safe permissive plans for four sets of examples. [ABSTRACT FROM AUTHOR]

Beam search is commonly used to help maintain tractability in large search spaces at the expense of completeness and optimality. Here we study supervised learning of linear ranking functions for controlling beam search. The goal is to learn ranking functions that allow for beam search to perform nearly as well as unconstrained search, and hence gain computational efficiency without seriously sacrificing optimality. In this paper, we develop theoretical aspects of this learning problem and investigate the application of this framework to learning in the context of automated planning. We first study the computational complexity of the learning problem, showing that even for exponentially large search spaces the general consistency problem is in NP. We also identify tractable and intractable subclasses of the learning problem, giving insight into the problem structure. Next, we analyze the convergence of recently proposed and modified online learning algorithms, where we introduce several notions of problem margin that imply convergence for the various algorithms. Finally, we present empirical results in automated planning, where ranking functions are learned to guide beam search in a number of benchmark planning domains. The results show that our approach is often able to outperform an existing state-of-the-art planning heuristic as well as a recent approach to learning such heuristics. [ABSTRACT FROM AUTHOR]

Abstract: AI planning requires the definition of action models using a formal action and plan description language, such as the standard Planning Domain Definition Language (PDDL), as input. However, building action models from scratch is a difficult and time-consuming task, even for experts. In this paper, we develop an algorithm called ARMS (action-relation modelling system) for automatically discovering action models from a set of successful observed plans. Unlike the previous work in action-model learning, we do not assume complete knowledge of states in the middle of observed plans. In fact, our approach works when no or partial intermediate states are given. These example plans are obtained by an observation agent who does not know the logical encoding of the actions and the full state information between the actions. In a real world application, the cost is prohibitively high in labelling the training examples by manually annotating every state in a plan example from snapshots of an environment. To learn action models, ARMS gathers knowledge on the statistical distribution of frequent sets of actions in the example plans. It then builds a weighted propositional satisfiability (weighted MAX-SAT) problem and solves it using a MAX-SAT solver. We lay the theoretical foundations of the learning problem and evaluate the effectiveness of ARMS empirically. [Copyright &y& Elsevier]

Although in the last decade the performance of domain-independent planners has significantly improved, there is no planner that outperforms all the others in every benchmark domains. In many domains, the planning performance can be improved by automatically deriving and exploiting knowledge about the domain and problem structure that is not explicitly given in the input formalization, and that can be used for optimizing the planner behavior.This thesis proposes three innovative techniques for deriving and using additional domain and problem knowledge through automatic algorithm configuration and machine learning techniques: a prediction model of the planning horizons, a method for synthesizing domain-optimized planners from a highly parameterized generic planner, and a system for the automatic configuration of a portfolio planner. [ABSTRACT FROM AUTHOR]

Generalized planning problems capture planning problems with uncertainties in object quantities and properties. While the generalized planning problem is incomputable in general, this thesis presents a new formalization of the problem, a study of the limits of computability of generalized plans and algorithms for solving a broad class of generalized planning problems. [ABSTRACT FROM AUTHOR]