ESTRO 38 Abstract book

S963 ESTRO 38

order to understand whether is necessary to perform the verifications for each VMAT plans or whether these can be replaced by a brief calculation of one or more parameters. Material and Methods 168 VMAT clinical plans that include 48 prostate tumor bed, 40 prostate pelvis, 40 rectums and 40 head and neck have been planned with Philips Pinnacle 3 v9.0 for Varian Clinac iX with 120 Millenium MLC. For all plans, pre- treatment verification have been performed with ArcCHECK (Sun Nuclear Corp.) with 2D local gamma analysis (3%-3mm) with a threshold of 10%. RTPlan files were analyzed with a homemade MATLAB (MathWorks®) program to obtain plan complexity indices: MCS and Leaf Travel LTi defined by Masi et al. (2013), and Monitor Units (MU). The correlations between gamma passing rate and MCS, LTi and MU have been investigated by Pearson coefficient. The correlations were considered weak for | r |<0.4, moderate for 0.4≤| r |≤0.7 and strong for | r |>0.7. Three ranges of values of the passing rate have been established:  · γ<88% rejection area;  · 88%≤γ≤92% verification area;  · γ>92% acceptability area; A machine-learning algorithm known as classification trees was used to identify threshold values for the calculated parameters that could predict at which range the study plane belongs. The discriminative performances were tested with the Leave One Out cross validation procedure. The accuracy of the results was expressed in terms of sensitivity and specificity, choosing to impose the values of 95% and 90% respectively. Results The table shows the Pearson coefficients obtained. For the prostate tumor bed the correlations found with MCS and LTi were moderate and strong with MU. For the other districts, r is around 0.4 for MCS while MU appears to be moderately correlated with the exception of the rectum plans. The correlation with LTi was found to be weak or absent. Considering all plans the correlation with MCS is moderate while it appears weak for the other two parameters. The results show that the correlation found between complexity parameters and ϒ passing rate is strongly dependent on the anatomical district analyzed. For each parameter, it was possible to identify a threshold value. The best results, analyzing the data by categories, were obtained with MCS and MU in the pelvis district in identifying the rejected category: for MCS a threshold of 0.20 identifies the rejected plans with specificity 97% and sensitivity 46%; for MU, the threshold 752 achieves a specificity of 83% and a sensitivity of 46%. Analyzing all the data, the best result is achieved by LTi, whose threshold identifies the plans to be accepted with specificity 97% and sensitivity 50%. Conclusion It was not possible to predict the results of the pre- treatment QA through MCS, LTi and MU with statistical reliability, not allowing us to avoid making the verification for each VMAT plans. EP-1782 Comparison of two commercial detectors and the influence of grid spacing calculations in SBRT R. Gómez Pardos 1 , E. Ambroa Rey 1 , D. Navarro Giménez 1 , A. Ramírez Muñoz 1 , J. García-Miguel Quiroga 1 , M. Colomer Truyols 1

were assessed in pairs by employing a Student’s T-test (p<0.05). Results As the complexity of the plan grows the differences between the results obtained by the three instruments increase considerably (Fig. 1).

For the prostate tumor bed the correlation found with MCS and MU for the three devices were moderate and strong. With LTi the correlation was weak except for D 4 . For pelvis a moderate correlation was found with MCS for A and O while for the MU the correlation was moderate for A and strong for O. In the rectums a moderate correlation was found with MCS for A, the other correlations were weak or absent. For head and neck plans, the only device who found a moderate correlation with MCS and LTi and a strong one with MU was Delta 4 . Considering all 50 plans for A the correlation with MCS, MU and LTi was weak, a strong correlation was found for O with MCS and MU and a moderate one with LTi and finally for D 4 the correlation was moderate with MCS and weak for the other parameters. A T-test was computed for the pairs: A-O, A-D 4 and O-D 4 . Considering all the 50 plans, we found statistically significance for all the pairs suggesting that the averages of the two distributions compared are different. Performing the test for individual anatomical districts, the results obtained show that for the prostate tumor bed all three devices seem to provide the same information (p>0.05). The same result was achieved in the pelvis for the couple A-O and for the head-neck for couples O-D. Conclusion A strong dependence of the result of the gamma analysis from the instrument used was found. In fact the correlation was found only in few cases bringing out the limitations of the gamma metric, including the volumetric reconstruction algorithms and the differences in the measured dose. EP-1781 Ability of Modulation Complexity Score to predict the result of pre-treatment QA for VMAT plans I. Vacchieri 1 , M. Liotta 2 , A. Malovini 3 , P. Tabarelli De Fatis 2 1 University of Milan, post graduate Medical Physics School, Milan, Italy ; 2 IRCCS Istituti Clinici Scientifici Maugeri, Medical Physics Unit, Pavia, Italy ; 3 IRCCS Istituti Clinici Scientifici Maugeri, Computer Science and Systems Laboratory for Clinical Research, Pavia, Italy Purpose or Objective to evaluate the ability of the Modulation Complexity Score (MCS) to predict the result of the pre-treatment QA in