GPP02 Presentation Time: 10:39 AM

Purpose

Fully automated adaptive treatment planning for image-guided cervical cancer brachytherapy following EMBRACE II planning aims, using ‘BiCycle’, was recently been developed. Favorable results were achieved in retrospective research studies [Oud et al., Radiother Oncol 148:143, 2020; Bijman et al., Radiother Oncol 170(S390), 2022; Rossi et al. Radiother Oncol 182(S92), 2023]. In this study, BiCycle was implemented parallel to the clinical routine to prospectively evaluate its impact on plan quality and workload. Dosimetrical parameters and planning times were prospectively collected as well as subjective plan scoring by five of the six clinicians performing cervical cancer brachytherapy.

Materials and Methods

Between June 2022 and January 2023, max. 2 patients per week were included, if compatible with clinical workload. Each patient was treated with an intracavitary applicator with interstitial needles, using a manually generated plan, without modifying clinical practice. Manual plan generation was performed by an RTT in Oncentra-Brachy (OCB) (Elekta) followed by a manual plan adjustment togehter with the treating physician (if desired), leading to a Man_Adj plan used for delivery. After treatment, an automated plan was generated using BiCycle aiming at the same CTVHR D90% as in the Man_Adj plan (Auto plan generation). The plan was then imported into OCB and evaluated, and optionally adjusted without access to the Man_Adj plan by the same treating physician who also evaluated and adjusted the manual plan, resulting in an Auto_Adj plan. Next, the treating physician compared Auto_Adj and Man_Adj plans using Visual Analogue Scales (VAS) for i) overall plan quality, ii) target structures doses only, iii) OARs doses only and iv) loading pattern only. For each VAS, the clinician first selected the favorable plan and then expressed the importance of superiority using a 0-100 scale. Both for manual planning and autoplanning, times needed to generate and adjust the plan were recorded.

Results

Data of 41 fraction plans were included, resulting in 37 evaluable plans due to 3 not evaluated in time, and 1 missing informed consent. Auto_Adj plans resulted in superior plan quality compared to Man_Adj in almost all aspects. While CTVHR D90% and GTVRES D98% improved by 0.3 and 1.7 Gy, respectively, CTVIR D98% decreased by 1.3 Gy, while always remaining within constraints (results expressed in total EBRT+BT EQD2Gy doses). Auto_Adj reduced bladder, rectum, sigmoid and bowel D2cc by on average by 3.7, 3.0, 1.0, 1.4 Gy, respectively. Treating physicians expressed an overall preference for Auto_Adj, which was preferred in 28, 17, 30 and 7 plans out of 37 for VAS overall, CTV, OARs and loading pattern respectively, while Man_Adj was preferred in 2, 1, 2 and 9 plans respectively. Planning+adjustment times significantly reduced with automated planning for all fractions (Figure 1), with a reduction in average times from 44.1 minutes (23.2 min plan generation + 22.8 min plan adjustment) to 9.4 minutes (5.3 min plan generation + 4.0 min plan adjustment).

Conclusion

The automated treatment planning workflow for cervical cancer brachytherapy, BiCycle, was validated parallel to clinical practice, meeting high dosimetrical quality and physician preference while reducing total planning time from 44.1 min to 9.4 min. The system has been released in clinical practice fulfilling EU Medical Device Regulation (MDR) requirements.