ESTRO 2024 - Abstract Book

S4939

Physics - Quality assurance and auditing

ESTRO 2024

The average pass rate was 86.6% and yearly average pass rate has not statistically changed (R2 = 0.31; p = 0.12) over the data cohort’s timeframe. Statistically significant differences in the means were seen between pass-fail for film pass rate and target coverage (p < 0.001). This is notable because both TLD and all complexity metrics individually struggle to differentiate acceptable and unacceptable dose delivery plans. 3D conformal radiation therapy (3D-CRT) had statistically significant higher pass rates compared to sliding window and segmental irradiation techniques (~8 10% higher). Random forest evaluation of the results, and the interplay between different parameters, revealed that the mean MLC gap (average rank = 2.6) and target coverage (1.0) ranked as the two most important variables for predicting phantom pass result. These same variables were also best at predicting film pass rate (mean MLC gap average rank = 1.1; tumor coverage average rank = 2.0), but TLD agreement was driven almost exclusively by complexity metrics (particularly MU (1.0) and modulation index (2.0)). Random forest modeling by separate treatment parameters showed the top ranked variables of importance, depending on the output parameter of phantom performance, were observed to consistently populate the higher ranks.

Conclusion: