We use cookies to enhance your site experience. This website stores data such as cookies to enable essential site functionality, as well as marketing, personalization and analytics. By continuing to use our website, you consent to cookies being used. Review our Privacy Policy for more information.
Xuemin Li, MD - Peking University Third Hospital, Beijing, Beijing
X. Li1, R. Cao2, M. Zhang3, R. Peng4, J. Zhou5, H. Wang6, and J. Wang4; 1Cancer Center, Peking University Third Hospital, Beijing, China, 2Shanghai United Imaging Healthcare Co., Ltd, Shanghai, China, 3Peking university, Beijing, China, 4Department of Radiation Oncology, Peking University Third Hospital, Beijing, China, 5Shanghai United Imaging Healthcare Co., Ltd., Shanghai, China, 6Beijing Key Laboratory for Interdisciplinary Research in Gastrointestinal Oncology (BLGO), Peking University Third Hospital, Beijing, China
Purpose/Objective(s):
To determine whether a tiered plan-library–first workflow reduces the fraction-level rate of triggered online ART compared with a static plan strategy under the same dosimetric acceptability criteria in a dual-target, dual-prescription postprostatectomy setting.
Materials/Methods:
A tiered plan-library–first workflow with escalation to triggered online ART was implemented for postprostatectomy radiotherapy. The retrieval algorithm first extracts the relative spatial position features of alternative plans based on the clinically concerned ROIs, then stores each feature vector of the alternative plan subsequently. These ROI structures include not only target volumes and organs at risk (OARs) but also those generated by clinically concerned dose thresholds. After acquiring the daily fan-beam computed tomography (FBCT) images, the algorithm performs feature extraction on the automatically segmented ROIs of clinical interest. Finally, similarity calculation is conducted between this feature vector and those of alternative plans, so as to recommend the optimal alternative plan. Prescription was 62.5 Gy in 25 fractions to PTV-bed (D95% = 62.5 Gy) and 45 Gy in 25 fractions to PTV-LN (D95% = 45 Gy). Fraction-level acceptability required institutional target coverage and OAR constraints, assessed with expert review and dose verification. Retrieved plans were used only if acceptable; otherwise, triggered online ART was performed and the adapted plan was added back to the library. The workflow was evaluated retrospectively in 9 postprostatectomy patients (224 fractions). Computational performance was extracted from system logs.
Results:
Triggered online ART decreased to 50/224 fractions (22.3%). Under the same constraints, the static plan would not have met fraction-level acceptability criteria in 125/224 fractions (55.8%), an absolute reduction of 33.5 percentage points and a relative reduction of 60.0% in fractions requiring online ART. Acceptability was 83.17% with top-1 retrieval and increased to 94.55% with top-2 retrieval. Based on system logs, each similarity calculation within approximately 200 ms on a platform equipped with anIntel i5-10500 CPU and NVIDIA RTX 2060 GPU, and generated plan recommendations within 6 s.
Conclusion:
A tiered plan-library–first workflow with escalation to triggered online ART reduced fractions requiring resource-intensive online adaptation while maintaining dose constrained, fraction-level dosimetric acceptability in dual-target, dual-prescription postprostatectomy radiotherapy. This approach may scale ART for complex multi-target plans while controlling clinical workload.
