PV QA 3 - Poster Viewing Q&A 3
Purpose/Objective(s):To develop a plan optimization algorithm tailored to a new rotatable, intensity modulated intracavitary brachytherapy (IMBT) for cervical cancer. It mainly aims to minimize the number of source dwelling positions and angles without compromising the plan quality.
Materials/Methods:The novel tandem applicator has a 7 cm long cylindrical tungsten shield with a groove, rotatable tube driven by a servo motor, whose end is connected to a hollow flexible shaft. It enables dose delivery in discrete fashion by rotating Ir-92 source placed on the designated positions and modulating its dwelling time. Namely, the plan is optimized to find the optimal dwelling time on two degrees of freedom: 1) source position, and 2) rotation angles. From clinical perspective for efficiency and safety, the number of dwelling positions and angles needs to be minimized, which mathematically means to have the minimum number of non-zero elements in solution. This notion is identical to the definition of L0-norm minimization that is an ideal solver for the sparse signal recovery. Despite its non-convex nature, this work optimized the L0-norm minimizing problem by the alternating direction algorithm and delicate hyper-parameter tuning. We validated the L0-norm based optimization against the algorithms with L1-norm constraint and without any constraint in terms of the number of dwelling positions, and plan quality for a MR retrospective cervical cancer patient dataset. The source positions were distant by 2.5mm and 20o interval, where the dose distribution at a certain position was estimated from in-house Monte-Carlo simulation.
Results:The other two optimizing algorithms without any constraint and with L1-norm constraint produced the resulting plans with 40 and 23 dwelling positions, respectively. On the contrary, the L0-norm constrained algorithm significantly reduced the source dwelling positions to 12 [Table 1]. With the reduced number of source positions, the plan quality was entirely preserved with no damage relative to the other two algorithms in dose volume histograms (DVHs) of the high-risk CTV, bladder, rectum, sigmoid and bowel. Of note, the optimal plan from the proposed algorithm has no additional expense regarding the sum of dwelling time.
Conclusion:Our proposed planning algorithm of rotational IMBT for cervical cancer successfully reduced the number of source dwelling positions and angles, while preserving the plan quality. Table 1. Comparison of different planning optimization algorithms for rotational IMBT in the optimal number of source positions and angles and entire dwelling time
|Number of non-zero elements (source positions/angles)||40||23||12|
|Sum of solutions (entire source dwelling time)||625||623||623|
The asset you are trying to access is locked. Please enter your access key to unlock.