Machine learning approaches for improving dosimeter reading accuracy conform with new operational quantities...

This research investigated machine learning (ML) for improving dosimeter reading accuracy under new operational quantities from ICRU 95. Qulxel, InLight, and OSLN dosimeters were irradiated with various photon energies, angles, and accumulated doses. Two ML models were developed: ML1 bases on K-means clustering and linear regression, and ML2 with random forest regression. Both models improved dosimeter reading accuracy compared to the existing algorithm, especially for low-energy photons. ML2, using key features for predictive dose reading, outperformed ML1. The study suggests that machine learning can be a valuable tool for enhancing dosimeter reading and connecting existing quantities to new quantities smoothly.