A Collimator Setting Optimization Algorithm for Dual-Arc Volumetric Modulated Arc Therapy in Pancreas Stereotactic Body Radiation Therapy

Abstract

Purpose: To optimize collimator setting to improve dosimetric quality of pancreas volumetric modulated arc therapy plan for stereotactic body radiation therapy.

Materials and methods: Fifty-five volumetric modulated arc therapy cases in stereotactic body radiation therapy of pancreas were retrospectively included in this study with internal review board approval. Different from the routine practice of initializing collimator settings with a template, the proposed algorithm simultaneously optimizes the collimator angles and jaw positions that are customized to the patient geometry. Specifically, this algorithm includes 2 key steps: (1) an iterative optimization algorithm via simulated annealing that generates a set of potential collimator settings from 39 cases with pancreas stereotactic body radiation therapy, and (2) a multi-leaf collimator modulation scoring system that makes the final decision of the optimal collimator settings (collimator angles and jaw positions) based on organs at risk sparing criteria. For validation, the other 16 cases with pancreas stereotactic body radiation therapy were analyzed. Two plans were generated for each validation case, with one plan optimized using the proposed algorithm (Plan_opt) and the other plan with the template setting (Plan_conv). Each plan was optimized with 2 full arcs and the same set of constraints for the same case. Dosimetric results were analyzed and compared, including target dose coverage, conformity, organs at risk maximum dose, and modulation complexity score. All results were tested by Wilcoxon signed rank tests, and the statistical significance level was set to .05.

Results: Both plan groups had comparable target dose coverage and mean doses of all organs at risk. However, organs at risk (stomach, duodenum, large/small bowel) maximum dose sparing (D_{0.1 cc} and D_{0.03 cc}) was improved in Plan_opt compared to Plan_conv. Plan_opt also showed lower modulation complexity score, which suggests better capability of handling complex shape and sparing organs at risk .

Conclusions: The proposed collimator settings optimization algorithm successfully improved dosimetric performance for dual-arc pancreas volumetric modulated arc therapy plans in stereotactic body radiation therapy of pancreas. This algorithm has the capability of immediate clinical application.

Keywords: IMRT; SBRT; VMAT; collimator settings; optimization; pancreas cancer; simulated annealing; treatment planning.

Figure 1.

Flowchart of the proposed collimator angle optimization algorithm. Red blocks are inputs, and green block is output. Xd indicates the larger jaw opening value of the dual-arc pair in the X direction (multi-leaf collimator [MLC] motion direction).

Figure 2.

Dose distribution comparison of Plan_opt (left) and Plan_conv (right) from a selected case. Yellow arrows in (A) and (B) indicate the comparison of dose conformity, and blue arrows in (C) and (D) indicate the comparison of organs at risk (OAR) sparing. Primary planned target volume (PTV_{25 Gy}) is shown in cyan, and PTV_{33 Gy} is shown in red. Proximal OARs are shown in yellow (stomach), dark blue (duodenum), and dark green (large bowel). Isodose lines are shown in yellow (3300 cGy) and white (3135 cGy). Plan_opt indicates proposed algorithm; Plan_conv, plan with the template setting.

Figure 3.

Comparison of the dose statistics (normalized to prescription dose value) in Plan_conv and Plan_opt.*Statistical significance. Plan_opt indicates proposed algorithm; Plan_conv, plan with the template setting

Figure 4.

Comparisons of the modulation complexity score, weighed multi-leaf collimator (MLC) opening, and total MU between Plan_conv and Plan_opt. Both tests were of statistical significance. Plan_opt indicates proposed algorithm; Plan_conv, plan with the template setting.

Figure 5.

Illustration of the initial collimator settings in the dual collimator angles plane. Each pixel value in the figure is the occurrence count of the desirable collimator angle pairs of 39 library cases. Please note that each case could have multiple desirable collimator angle pairs with similar X_d. Red dots represent the selected initial CS_a,i.

Figure 6.

Collimator angle pair results from (A) simulated annealing optimization and (B) exhausting search (red dots) overlapped to the results from simulated annealing optimization. These 2 methods show good coincidence with each other.