ESTRO 2025 - Abstract Book

S3698

Physics - Quality assurance and auditing

ESTRO 2025

methods used will aid in reassessing QA protocols and procedures, both for ongoing daily practice and if new technologies are implemented clinically.

Keywords: Machine QA, PSQA, efficiency

References: 1] AAPM TG 100 (2016) Medical Physics 43: 4209-4262: 2] NCS Report 25: https://radiationdosimetry.org/ncs/documents/ncs-25-process-management-and-quality assurance-for-intracranial-stereotactic-treatment 3] Proposition paper; 2024; Internal PowerPoint; available on request by s.heukelom@amsterdamumc.nl 4] NCS Report 33: https://radiationdosimetry.org/ncs/quality-control-for-linear-accelerators

4313

Proffered Paper Data-driven patient treatment optimization using patient specific quality assurance Esther Decabooter, Michel C Öllers, Bas MJJG Nijsten Department of Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, Netherlands Purpose/Objective: Currently, patient specific quality assurance (PSQA) is primarily used to detect patient specific treatment errors. This work shows how retrospective analysis of PSQA results can be used to improve patient workflows by identifying failure modes. Material/Methods: Retrospective analysis is performed on PSQA results collected from February 2024 until November 2024. Four different PSQA methods are analyzed: secondary dose calculations (SDC) with a Monte Carlo dose engine XVMC (1) (N=3130), pretreatment QA performed with an electronic portal imaging device (EPID) (N=873), QA based on machine log-files (N=3569) and SDC on CBCTs (N=13704). All methods, except the EPID QA, use a model to predict 2D dose images (DIs) using either the treatment plan or the treatment log files. (2) Subsequently a 3D dose reconstruction method is used to reconstruct the dose in a planning CT or a CBCT based on 2D predicted or measured DIs. (3) These are then compared either to the TPS dose or the SDC dose (for the CBCT recalculation) using a (3%,3mm) gamma analysis. Results are flagged as “fail” if the agreement score is lower than 95 % in the high dose region. These PSQA methods are implemented in an automated PSQA framework and results are saved along with treatment plan specific parameters in a database. A large-scale database analysis using Microsoft Power BI is used to group bulk QA results in order to identify common failure modes. Results: Figure 1 illustrates the different failure modes per PSQA method. 19 % of the failed reports are due to technical limitations of the physics models or applied workflow (FM1-7). Failure mode 8 (2% of failed reports) consists of patients with an inaccurate delineation of immobilization devices in our TPS. EPID and log file based QA identified a failure mode (1%) related to deliverability of treatment plans. Most failure modes (77 %) are defined for patient related dose delivery errors. Dose differences due to differences in 3D dose algorithms are eliminated by using the SDC as a reference dose making it possible to filter out issues solely related to patient anatomy for this PSQA method. The failure modes can then subsequently be used to improve patient treatment workflows by investigating failure modes per treatment site (Figure 2).