Department of Information Technology and Electrical Engineering, Swiss Federal Institute of Technology (ETH), Zurich, Switzerland.
Foundation for Research on Information Technologies in Society (IT'IS), Zurich, Switzerland.
Int J Hyperthermia. 2022;39(1):758-771. doi: 10.1080/02656736.2022.2080284.
Healthy tissue hotspots are a main limiting factor in administering deep hyperthermia cancer therapy. We propose an optimization scheme that uses time-multiplexed steering (TMPS) among minimally correlated (nearly) Pareto-optimal solutions to suppress hotspots without reducing tumor heating. Furthermore, tumor heating homogeneity is maximized, thus reducing toxicity and avoiding underexposed tumor regions, which in turn may reduce recurrence.
The novel optimization scheme combines random generation of steering parameters with local optimization to efficiently identify the set of (Pareto-) optimal solutions of conflicting optimization goals. To achieve simultaneous suppression of hotspots, multiple steering parameter configurations with minimally correlated hotspots are selected near the Pareto front and combined in TMPS. The performance of the novel scheme was compared with that of a multi-goal Genetic Algorithm for a range of simulated treatment configurations involving a modular applicator heating a generic tumor situated in the bladder, cervix, or pelvic bone. SAR cumulative histograms in tumor and healthy tissue, as well as hotspot volumes are used as metrics.
Compared to the non-TMPS optimization, the proposed scheme was able to reduce the peak temperature in healthy tissue by 0.2 °C-1.0 °C (a thermal dose reduction by at least 26%) and, importantly, the hotspot volume above 42 °C in healthy tissue by 41%-86%. At the same time, tumor heating homogeneity was maintained or improved.
The extremely rapid optimization (5 s for TMPS part, on a standard PC) permits closed-loop treatment reoptimization during treatment administration, and empowers physicians with a selection of optimal treatment scenarios reflecting different weighting of conflicting treatment goals.
健康组织热点是深部热疗癌症治疗的主要限制因素。我们提出了一种优化方案，该方案使用时间复用转向（TMPS）在最小相关（几乎）帕累托最优解之间进行优化，以在不降低肿瘤加热的情况下抑制热点。此外，最大限度地提高了肿瘤加热的均匀性，从而降低了毒性并避免了未暴露的肿瘤区域，这反过来又可能降低了复发的风险。
新的优化方案将转向参数的随机生成与局部优化相结合，以有效地识别具有冲突优化目标的（帕累托）最优解集。为了同时抑制热点，在帕累托前沿附近选择具有最小相关热点的多个转向参数配置，并在 TMPS 中进行组合。将新方案的性能与多目标遗传算法进行了比较，模拟了各种治疗配置，涉及模块化加热器加热位于膀胱、宫颈或骨盆中的通用肿瘤。使用肿瘤和健康组织中的 SAR 累积直方图以及热点体积作为指标。
与非 TMPS 优化相比，所提出的方案能够将健康组织中的峰值温度降低 0.2°C-1.0°C（热剂量降低至少 26%），并且重要的是，健康组织中高于 42°C 的热点体积降低了 41%-86%。同时，保持或改善了肿瘤加热的均匀性。
极快的优化（TMPS 部分 5 秒，在标准 PC 上）允许在治疗过程中进行闭环治疗重新优化，并为医生提供了一系列最佳治疗方案，反映了不同的权重冲突的治疗目标。
