Ethos智能优化引擎在宫颈癌自动计划生成中的性能评估

Performance evaluation of Ethos intelligent optimization engine in automatic plan generation

目的评估Ethos在线自适应放疗平台智能优化引擎(IOE)自动优化性能及其临床可行性。方法回顾性分析11例Halcyon加速器治疗的宫颈癌术后患者的资料,患者所用计划均为Eclipse中采用的4个全弧容积弧形调强放疗(VMAT)手工计划(Manual-4Arc),处方剂量为45 Gy分25次。将所有患者影像和结构导入Ethos模拟器,基于科室剂量限值要求添加合适的临床目标,靶区归一到95%覆盖,之后采用IOE进行自动计划生成,得到7、9、12野均分固定野调强放疗(IMRT)计划(IMRT-7F、IMRT-9F、IMRT-12F)和2、3弧的VMAT计划(VMAT-2Arc、VMAT-3Arc)。使用单因素方差分析对6种计划进行剂量学比较,并基于分析结果进行Turky事后多重比较进行优劣判断,评估IOE自动优化性能。结果高量方面,对于计划靶区(PTV),IMRT-12F计划的D 1%最低,且与Manual-4Arc计划存在显著性差异(P=0.004);靶区覆盖方面,所有方法制作出来的临床靶区均满足临床要求,Ethos在线自适应计划虽在计划制作时做了归一,但PTV靶区覆盖率略微不足。对于距离靶区较近的危及器官,比如膀胱,其V_(30 Gy)、V_(40 Gy)及D_(mean)在6种计划之间的差异有统计学意义。膀胱的受量排序基本为IMRT-12F<IMRT-9F<Manual-4Arc<IMRT-7F<VMAT-3Arc<VMAT-2Arc。直肠的V_(30 Gy)和D_(mean)在6种计划之间的差异有统计学意义,其受量排序基本与膀胱一致,除了IMRT-7F<Manual-4Arc计划。直肠V_(40 Gy)、小肠D_(max)及D_(mean)在6种计划之间的差异无统计学意义。对于离靶区较远的器官如左右股骨头、脊髓和骨髓受量排序基本为IMRT-12F<IMRT-9F<IMRT-7F<VMAT-2Arc<VMAT-3Arc<Manual-4Arc。结论Ethos IOE为宫颈癌术后放疗患者自动生成的计划可以达到与手工计划相当的质量,临床使用时优先选择自动生成的IMRT-12F和IMRT-9F计划。

Objective To evaluate the automatic optimization performance and clinical feasibility of the intelligent optimization engine(IOE)in the Ethos online adaptive radiotherapy platform.Methods Clinical data of 11 patients with postoperative cervical cancer treated with Halcyon accelerator were retrospectively analyzed.Manual planning was performed for all patients using the 4 full arc volumetric modulated arc therapy(VMAT)(Manual-4Arc)in Eclipse,with a prescription dose of 45 Gy/25F.Patient images and structures were imported into the Ethos simulator,and appropriate clinical goals were added based on clinical requirements.The target coverage was normalized to 95%.Automatic plan generation was conducted using IOE,resulting in 7,9,and 12 field intensity modulated radiotherapy(IMRT)plans(IMRT-7F、IMRT-9F、IMRT-12F),as well as 2 and 3 arc VMAT plans(VMAT-2Arc、VMAT-3Arc).Dosimetric index comparisons were made between the Manual-4Arc plans and the 5 groups of IOE-generated plans through one-way analysis of variance.Based on the analysis results,Turky post hoc multiple comparisons were performed to evaluate the automatic optimization performance of IOE.Results In terms of the high dose area,the IMRT-12F plans showed the lowest D1%for the planning target volume(PTV),and there were significant differences compared to the Manual-4Arc plans(P=0.004).Regarding target coverage,all groups produced clinical target volume(CTV)plans that met the clinical requirements.Although the Ethos online adaptive plans were normalized during planning,the PTV coverage was slightly insufficient.For organs at risk(OAR)close to the target,such as the bladder,there were significant differences in V_(30 Gy),V_(40 Gy),and D_(mean)among the 6 groups of plans.The dose ranking for the bladder was generally as follows:IMRT-12F<IMRT-9F<Manual-4Arc<IMRT-7F<VMAT-3Arc<VMAT-2Arc.There were significant statistical differences in V_(30 Gy)and D_(mean)for the rectum,and the dose ranking was generally consistent with that of the bladder,except for a switch between the IMRT-7F and Manual-4Arc plans.There were no significant differences in rectal V_(40 Gy),small intestine D_(max),and Dmean among the 6 groups of plans.For OAR distant from the target,such as the left and right femoral heads,spinal cord,and bone marrow,the dose ranking was generally as follows:IMRT-12F<IMRT-9F<IMRT-7F<VMAT-2Arc<VMAT-3Arc<Manual-4Arc.Conclusion The plans automatically generated by Ethos IOE in postoperative patients with cervical cancer can achieve similar performance to manual plans,and the automatically generated IMRT-12F and IMRT-9F plans are recommended for clinical use.