S2348
Physics - Quality assurance and auditing
ESTRO 2026
Conclusion: In-vivo constancy analysis using Epiqa proved to be a promising and reliable tool for patient treatment QA, avoiding additional uncertainties related to complex algorithm-based comparisons. This approach provides a valuable overview of in-vivo treatment delivery quality. Keywords: EPID, in-vivo, Epiqa
evaluated through group consensus stability, defined as Δκ after substituting each MP with ADDI. Non- inferiority was established if Δκ exceeded a − 10% margin relative to the baseline consensus. Results:
Digital Poster Highlight 1512
Multi-centre validation of an automated dose evaluation metric and analysis of treatment plan evaluation variability in breast radiotherapy Hiran CG Maladenige 1,2 , Jan Seppälä 2 , Tuomas Virén 2 , Juuso TJ Honkanen 2 , Janne Heikkilä 2 , Akseli Leino 1,2 , Karoliina Hundlinger 3 , Laura Tuomikoski 4 , Tero Vatanen 5 , Tiina Seppälä 4 , Henri Korkalainen 1,2 1 Technical Physics, University of Eastern Finland, Kuopio, Finland. 2 Cancer Center, Kuopio University Hospital, Kuopio, Finland. 3 Radiotherapy, Kymenlaakso Central Hospital, Kotka, Finland. 4 Cancer Center, Helsinki University Hospital, Helsinki, Finland. 5 Cancer Center, Joensuu Central Hospital, Joensuu, Finland Purpose/Objective: The complexity of modern radiotherapy treatment planning makes the plan evaluation process highly subjective and inconsistent [1], affecting clinical outcomes [2]. Evaluation currently relies on manual use of fragmented indices—such as homogeneity and conformity measures—and discrete DVH parameters [3]. To address these limitations, we developed the Advanced Dose Distribution Index (ADDI) [4], a holistic plan quality metric. This study aimed to quantify variability in plan evaluation across four Finnish radiotherapy centres and to assess ADDI’s performance relative to qualified medical physicists (MPs) in breast cancer radiotherapy, establishing ADDI's utility as an automated evaluation tool. Material/Methods: We performed a retrospective validation study with 21 left breast cancer patients treated with hypofractionated (15 × 2.67 Gy) tangential-VMAT [5] at the Kuopio University Hospital. For each patient, an alternative plan was created by increasing the parallel and serial cost-function isoconstraints for the ipsilateral lung and heart by 20%. Nine MPs independently evaluated both plans across two rounds, using predefined criteria (756 evaluations in total), and provided a final verdict for each patient. Inter- and intra-rater reliabilities (RR) were computed using Fleiss’ and Cohen’s kappa ( κ ), respectively. A probabilistic ground truth was derived as an intra-RR– weighted average of each MP’s decision and used to assess ADDI’s performance. ADDI’s substitutability was
The mean inter-rater Fleiss’ κ for the final verdict among MPs across both rounds was 0.61, with a mean pairwise Cohen’s κ of 0.6 (range 0.33–0.90) (Figure 1). The mean intra-rater Cohen’s κ was 0.75 (0.43–0.90). Agreement of final verdicts between each MP and the ground truth averaged 0.86 (0.65–0.94), while ADDI achieved 0.88. The consensus stability after replacing
Made with FlippingBook - Share PDF online