ESTRO 2025 - Abstract Book

S2664

Physics - Detectors, dose measurement and phantoms

ESTRO 2025

TPS simulation and categorized by magnitude: small (S) if 3.5%>ΔPTVD50%<5% , 7%<ΔPTVD98%<10%, or 7%<ΔPTVD2%<10%; medium (M) if 5%<ΔPTVD50%<10%, 10%<ΔPTVD98%<20%, or 10%<ΔPTVD2%<20%; and large (L) if ΔPTVD50%>10%, ΔPTVD98%>20%, or ΔPTVD2%>20%. For DE analysis, the TPS-calculated dose of error-free plans was compared with the EPID-reconstructed dose of plans delivered with errors. For IE analysis, the TPS-calculated dose of plans with errors was compared with the EPID-reconstructed dose of error-free plans. Sensitivity and specificity were evaluated through the confusion matrix for each error type and magnitude using DVH metrics. The agreement rate between the elements of the DE and IE confusion matrices was also computed. Results: Sensitivity and specificity depended on error type and magnitude. For DE, sensitivity and specificity were 0.40/0.50 for S , 0.60/0.70 for M , and 0.38/0.79 for L . For IE, these values were 0.85/0.52, 0.81/0.57, and 0.77/0.85 for small, medium, and large errors, respectively. Agreement rate between DE and IE confusion matrix for anatomical changes was 55% for S , 94% for M , and 92% for L . For setup errors, agreement was 26% for S , 54% for M , and 31% for L . For delivery errors, agreement was 14% for S , 41% for M , and 100% for L .