ESTRO 2025 - Abstract Book

S3586

Physics - Quality assurance and auditing

ESTRO 2025

696

Digital Poster Does the tolerance level for beam output checks of the Elekta Unity with its on-board megavoltage imager hold in long-term clinical practice? Guido C. Hilgers, Ilona Potters, Marijke Ikink, Danny Schuring, André W. Minken Medical Physics, Radiotherapiegroep, Arnhem/Deventer, Netherlands Purpose/Objective: It has been shown that the on-board megavoltage imager (MVI) of the Unity MRI-linac (Elekta, Stockholm, Sweden) is suitable for beam output checks with a 2% tolerance level if a monthly cross calibration is performed against an output measurement in water [1, 2]. Our objective was to evaluate this tolerance level in long-term clinical practice. Material/Methods: During a period of ~20 months (late September 2022 - early June 2024), 412 daily beam output checks were performed with the MVI (ΔO MVI ), and 29 in water (ΔO w ), of which 23 were for cross calibrations. Before applying a new cross calibration, we calculated the output deviation difference (ΔO MVI - ΔO w ). A Shapiro-Wilk test was conducted to determine whether these differences were normally distributed. A Gaussian was fitted and compared with the Gaussian (μ=0.1, σ=0.3 pp) on which the 2% tolerance level was based [1]. Results: Figure 1 shows that the spread in ΔO MVI is larger than in ΔO w . So, a larger tolerance on ΔO MVI than on ΔO w holds in clinical practice. All positive ΔO MVI outliers were traced back to the cross calibration on 07/05/2024 of which the associated ΔO w was an outlier as well (+0.8%). The other ΔO w outlier (+1.5%) was a stand-alone check in water for an output adjustment after a magnetron replacement.

Fig. 1. Distribution of the output deviations.

Figure 2 shows the distribution of the output deviation differences. The Shapiro-Wilk test nearly showed evidence of non-normality (p=0.06). A Gaussian (μ=0.0, σ=0.4 pp) was fitted to the data. Interestingly, the minimum (-1.0%) and maximum (+1.2%) differences also originated from the cross calibration on 07/05/2024. This suggests that output deviation differences can be used to validate cross calibrations. If both extremes are omitted, evidence for normality improved (p=0.15), and the σ of the Gaussian slightly decreased to 0.3 pp. Thus, the original Gaussian was reproduced. A practical criterion for approving cross calibrations then is that its accompanying output deviation difference is within 2σ (0.6 pp).