Objective function based ranking method for selection of optimal beam angles in IMRT

An uncertainty-incorporated method for fast beam angle selection in intensity-modulated proton therapy

Aim

We propose a novel metric called ψ - score to rank the Intensity Modulated Proton Therapy (IMPT) beams in the order of their optimality and robustness. The beams ranked based on this metric were accordingly chosen for IMPT optimization. The objective of this work is to study the effectiveness of the proposed method in various clinical cases.

Methods and Materials

We have used Pinnacle TPS (Philips Medical System V 16.2) for performing the optimization. To validate our approach, we have applied it in four clinical cases: Lung, Pancreas, Prostate+Node and Prostate. Basically, for all clinical cases, four set of plans were created using Multi field optimization (MFO) and Robust Optimization (RO) with same clinical objectives, namely (1) Conventional angle plan without Robust Optimization (CA Plan), (2) Suitable angle Plan without Robust Optimization (SA Plan), (3) Conventional angle plan with Robust Optimization (CA-RO Plan), (4) Suitable angle Plan with Robust Optimization (SA-RO Plan). Initial plan was generated with 20 equiangular beams starting from the gantry angle of 0°. In the corresponding SA Plan and SA-RO Plan, the beam angles were obtained using the guidance provided by ψ - score.

Results

All CA plans were compared against the SA plans in terms of Dose distribution, Dose volume histogram (DVH) and percentage of dose difference. The results obtained from the clinical cases indicate that the plan quality is considerably improved without significantly compromising the robustness when the beam angles are optimized using the proposed method. It takes approximately 10-15 min to find the suitable beam angles without Robust Optimization (RO), while it takes approximately 20-30 min to find the suitable beam angles with RO. However, the inclusion of RO in BAO did not result in a change in the final beam angles for anatomies other than lung.

Conclusion

The results obtained in different anatomic sites demonstrate the usefulness of our approach in improving the plan quality by determining optimal beam angles in IMPT.

A reinforcement learning application of a guided Monte Carlo Tree Search algorithm for beam orientation selection in radiation therapy

Current beam orientation optimization algorithms for radiotherapy, such as column generation (CG), are typically heuristic or greedy in nature because of the size of the combinatorial problem, which leads to suboptimal solutions. We propose a reinforcement learning strategy using a Monte Carlo Tree Search (MCTS) that can find a better beam orientation set in less time than CG. We utilize a reinforcement learning structure involving a supervised learning network to guide the MCTS and to explore the decision space of beam orientation selection problems. We previously trained a deep neural network (DNN) that takes in the patient anatomy, organ weights, and current beams, then approximates beam fitness values to indicate the next best beam to add. Here, we use this DNN to probabilistically guide the traversal of the branches of the Monte Carlo decision tree to add a new beam to the plan. To assess the feasibility of the algorithm, we used a test set of 13 prostate cancer patients, distinct from the 57 patients originally used to train and validate the DNN, to solve five-beam plans. To show the strength of the guided MCTS (GTS) compared to other search methods, we also provided the performances of Guided Search, Uniform Tree Search and Random Search algorithms. On average, GTS outperformed all the other methods. It found a better solution than CG in 237 s on average, compared to 360 s for CG, and outperformed all other methods in finding a solution with a lower objective function value in less than 1000 s. Using our GTS method, we could maintain planning target volume (PTV) coverage within 1% error similar to CG, while reducing the organ-at-risk mean dose for body, rectum, left and right femoral heads; the mean dose to bladder was 1% higher with GTS than with CG.

An empirical method for splitting arcs in VMAT

PURPOSE

We present a new approach to determine the optimal arc split for VMAT beams which is an extension of our recently published algorithm for selecting optimal beam angles in Intensity Modulated Radiation Therapy (IMRT) MATERIAL AND METHODS: The proposed approach uses an objective function based scoring method called "ψ - score" to determine optimal arc splitting strategy. To validate our approach, we applied it in different clinical cases: Abdomen-Para aortic node, Lung, Pancreas and Prostate. Basically, for all clinical cases, two set of plans were created, namely VMAT plan and VMAT_S plan using Pinnacle³ (V16.2, Philips Medical Systems (Cleveland), Inc.). In the VMAT plans, full arc (360°) with 4-degree gantry spacing was used during optimization to compute the "ψ - score". Subsequently the avoidable arc portions were identified and removed using the ψ - score plot followed by the final optimization (VMAT_S).

RESULTS

Equivalent or better OAR sparing, and similar target coverage were achieved in VMAT_S plans compared to VMAT plans. VMAT_S reduced the number of control points and monitor units by 24.2% and 12.9% respectively. On the average, beam on time was reduced by 21.9% and low dose volume (5 Gy isodose volume) to healthy tissues was reduced by 4.9% in VMAT_S compared to VMAT plans.

CONCLUSION

The results demonstrated that the proposed method is useful for reducing the monitor units, beam on time and low dose volume without significantly compromising plan quality and most useful for non-centrically located targets.

A self-adaptive prescription dose optimization algorithm for radiotherapy

Purpose

The aim of this study is to investigate an implementation method and the results of a voxel-based self-adaptive prescription dose optimization algorithm for intensity-modulated radiotherapy.

Materials and methods

The self-adaptive prescription dose optimization algorithm used a quadratic objective function, and the optimization engine was implemented using the molecular dynamics. In the iterative optimization process, the optimization prescription dose changed with the relationship between the initial prescription dose and the calculated dose. If the calculated dose satisfied the initial prescription dose, the optimization prescription dose was equal to the calculated dose; otherwise, the optimization prescription dose was equal to the initial prescription dose. We assessed the performance of the self-adaptive prescription dose optimization algorithm with two cases: a mock head and neck case and a breast case. Isodose lines, dose–volume histogram, and dosimetric parameters were compared between the conventional molecular dynamics optimization algorithm and the self-adaptive prescription dose optimization algorithm.

Results

The self-adaptive prescription dose optimization algorithm produces the different optimization results compared with the conventional molecular dynamics optimization algorithm. For the mock head and neck case, the planning target volume (PTV) dose uniformity improves, and the dose to organs at risk is reduced, ranging from 1 to 4%. For the breast case, the use of self-adaptive prescription dose optimization algorithm also leads to improvements in the dose distribution, with the dose to organs at risk almost unchanged.

Conclusion

The self-adaptive prescription dose optimization algorithm can generate an ideal clinical plan more effectively, and it could be integrated into a treatment planning system after more cases are studied.