Department of Physics, University of Alberta, Edmonton, AB, Canada.
Phys Med Biol. 2013 Apr 7;58(7):2045-58. doi: 10.1088/0031-9155/58/7/2045. Epub 2013 Mar 4.
Although the Manchester system is still extensively used to prescribe dose in brachytherapy (BT) for locally advanced cervix cancer, many radiation oncology centers are transitioning to 3D image-guided BT, owing to the excellent anatomy definition offered by modern imaging modalities. As automatic dose optimization is highly desirable for 3D image-based BT, this study comparatively evaluates the performance of two optimization methods used in BT treatment planning--Nelder-Mead simplex (NMS) and simulated annealing (SA)--for a cervix BT computer simulation model incorporating a Manchester-style applicator. Eight model cases were constructed based on anatomical structure data (for high risk-clinical target volume (HR-CTV), bladder, rectum and sigmoid) obtained from measurements on fused MR-CT images for BT patients. D90 and V100 for HR-CTV, D2cc for organs at risk (OARs), dose to point A, conformation index and the sum of dwell times within the tandem and ovoids were calculated for optimized treatment plans designed to treat the HR-CTV in a highly conformal manner. Compared to the NMS algorithm, SA was found to be superior as it could perform optimization starting from a range of initial dwell times, while the performance of NMS was strongly dependent on their initial choice. SA-optimized plans also exhibited lower D2cc to OARs, especially the bladder and sigmoid, and reduced tandem dwell times. For cases with smaller HR-CTV having good separation from adjoining OARs, multiple SA-optimized solutions were found which differed markedly from each other and were associated with different choices for initial dwell times. Finally and importantly, the SA method yielded plans with lower dwell time variability compared with the NMS method.
虽然曼彻斯特系统仍然广泛用于局部晚期宫颈癌的近距离放射治疗（BT）剂量规定，但由于现代成像方式提供了出色的解剖定义，许多放射肿瘤学中心正在向 3D 图像引导 BT 过渡。由于 3D 图像引导 BT 高度需要自动剂量优化，因此本研究比较评估了两种用于 BT 治疗计划的优化方法--Nelder-Mead 单纯形（NMS）和模拟退火（SA）--在包含曼彻斯特式施源器的宫颈癌 BT 计算机模拟模型中的性能。基于从 BT 患者融合的 MR-CT 图像上的测量获得的解剖结构数据（用于高风险临床靶区（HR-CTV）、膀胱、直肠和乙状结肠），构建了 8 个模型病例。对于旨在以高度适形方式治疗 HR-CTV 的优化治疗计划，计算了 HR-CTV 的 D90 和 V100、OAR 的 D2cc、点 A 的剂量、适形指数和串联和卵圆内驻留时间的总和。与 NMS 算法相比，SA 被发现更优越，因为它可以从一系列初始驻留时间开始进行优化，而 NMS 的性能强烈依赖于它们的初始选择。SA 优化的计划还表现出较低的 D2cc 到 OARs，特别是膀胱和乙状结肠，并减少了串联驻留时间。对于具有与毗邻 OARs 良好分离的较小 HR-CTV 的病例，发现了多个 SA 优化解决方案，这些解决方案彼此明显不同，并且与初始驻留时间的不同选择有关。最后，重要的是，与 NMS 方法相比，SA 方法产生的计划具有较低的驻留时间变化性。
