Measurement‐anchored Monte Carlo prediction of diagnostic x‐ray air kerma using HVL‐matched spectra and per‐electron normalization

Abstract Background Quantitative Monte Carlo (MC) prediction of diagnostic x‐ray air kerma is limited by mismatch between theoretical spectra and measured beam quality and by the lack of a reproducible normalization from source‐photon tallies to mAs‐based exposures. Purpose To test whether HVL‐matched spectra combined with measurement‐anchored per‐electron normalization can predict free‐in‐air and transmitted air kerma under selected radiographic exposure settings and to evaluate the framework in PHITS and EGS5. Methods Birch‐Marshall spectra at 50, 80, and 120 kV were adjusted by tuning the Al‐equivalent filtration parameter so that the theoretical HVL agreed with the measured HVL. The matched spectra were used as common input for PHITS and EGS5. Free‐in‐air air kerma was measured for 102 exposure conditions, and transmitted air kerma was measured after 2.0‐cm PMMA or 0.5‐cm Al for seven conditions at each tube voltage. From measured air kerma and MC air‐kerma tallies expressed per source photon, we defined the normalization constant , the per‐electron correction factor , and the tube‐voltage‐averaged coefficient . Predictions based on were compared with measurements using percentage error. Performance on the full free‐in‐air dataset was interpreted as agreement within the calibration dataset, whereas held‐out‐condition performance was assessed by leave‐one‐out cross‐validation (LOOCV) on seven selected radiographic exposure settings at each tube voltage. Combined standard uncertainty was also summarized. Results After HVL matching, the free‐in‐air MC air‐kerma tallies expressed per source photon differed between PHITS and EGS5, a finding interpreted primarily in relation to differences in source definition and source normalization. After measurement‐based normalization with , predicted air kerma was generally consistent with measurements under both free‐in‐air and transmitted conditions. The largest mean percentage error was 7.8%, slightly exceeding the predefined ± 7.5% operational benchmark. The representative combined standard uncertainty was 5.6%; when the nominal tube‐voltage setting was used directly as a simulation input, inclusion of the tube‐voltage accuracy term increased it to 7.3%. Conclusions Within the tested radiographic system and slab‐transmission geometries, HVL‐matched spectra combined with measurement‐anchored per‐electron normalization enabled prediction of air kerma under actual exposure conditions. Despite differences in the MC air‐kerma tallies expressed per source photon, the final predicted values were generally consistent with measurements in both codes under the tested conditions.