Département de physique, de génie physique et d'optique et Centre de recherche sur le cancer de l'Université Laval, CHU de Québec, Québec, Québec, Canada; Département de radio-oncologie et Centre de recherche du CHU de Québec, CHU de Québec - Université Laval, Québec, Québec, Canada.
Département de radio-oncologie et Centre de recherche du CHU de Québec, CHU de Québec - Université Laval, Québec, Québec, Canada.
Brachytherapy. 2022 Jul-Aug;21(4):551-560. doi: 10.1016/j.brachy.2022.04.003. Epub 2022 May 15.
Recently, a GPU-based multicriteria optimization (gMCO) algorithm was integrated in a graphical user interface (gMCO-GUI) that allowed real-time plan navigation through a set of Pareto-optimal plans for high-dose-rate (HDR) brachytherapy. This work reports on the inter-observer evaluation of the gMCO algorithm into the clinical workflow.
Twenty HDR brachytherapy prostate cancer patients were retrospectively replanned with the gMCO algorithm. The reference clinical plans were each generated by experienced physicists using inverse planning followed by graphical optimization and approved by a radiation oncologist (RO). Each case was replanned with the gMCO algorithm by generating 2000 Pareto-optimal plans with four different objective functions. Two physicists were asked to rank the objective functions according to their preferences by choosing one preferred plan for each plans pool and ranking them using gMCO-GUI. The optimized dwell positions and dwell times of the gMCO plans that were ranked first were exported to Oncentra Prostate where a blinded comparison of the gMCO plans with the clinical plans was conducted by three ROs.
The median planning time of the two physicists was 9 min. Both physicists preferred the objective function with target sub-regions to cover specific target regions. Regarding the blinded comparison, the gMCO plans were preferred 19, 17, and 12 times by the three ROs, in which eight gMCO plans were unanimously preferred compared with the clinical plans.
The plan quality and the planning time were similar between the two physicists and within what is observed in the clinic. Moreover, the gMCO plans evaluated favorably by ROs compared to the reference clinical plans.
最近，一种基于图形处理器的多准则优化（gMCO）算法被整合到一个图形用户界面（gMCO-GUI）中，该界面允许通过一组 Pareto 最优计划实时导航高剂量率（HDR）近距离放射治疗计划。本研究报告了 gMCO 算法在临床工作流程中的观察者间评估。
对 20 例 HDR 近距离治疗前列腺癌患者的病例进行回顾性再计划，参考临床计划由经验丰富的物理学家使用逆规划生成，然后进行图形优化，并由放射肿瘤学家（RO）批准。每个病例都使用 gMCO 算法生成 2000 个 Pareto 最优计划，使用四个不同的目标函数。两名物理学家被要求通过为每个计划池选择一个首选计划，并使用 gMCO-GUI 对其进行排名，根据个人偏好对目标函数进行排名。排名第一的 gMCO 计划的优化驻留位置和驻留时间被导出到 Oncentra Prostate，由三名 RO 对 gMCO 计划与临床计划进行盲法比较。
两名物理学家的中位规划时间为 9 分钟。两名物理学家都更喜欢将目标子区域覆盖特定目标区域的目标函数。关于盲法比较，三名 RO 分别 19 次、17 次和 12 次更倾向于 gMCO 计划，其中 8 次 gMCO 计划与临床计划一致受到青睐。
两名物理学家之间以及与临床实践中的计划质量和规划时间相似。此外，RO 对 gMCO 计划的评价优于参考临床计划。
