Texture analysis on fluence maps was performed to evaluate the degree of modulation for volumetric modulated arc therapy (VMAT) plans.
A total of six textural features including angular second moment, inverse difference moment, contrast, variance, correlation, and entropy were calculated for fluence maps generated from 20 prostate and 20 head and neck VMAT plans. For each of the textural features, particular displacement distances (d) of 1, 5, and 10 were adopted. To investigate the deliverability of each VMAT plan, gamma passing rates of pretreatment quality assurance, and differences in modulating parameters such as multileaf collimator (MLC) positions, gantry angles, and monitor units at each control point between VMAT plans and dynamic log files registered by the Linac control system during delivery were acquired. Furthermore, differences between the original VMAT plan and the plan reconstructed from the dynamic log files were also investigated. To test the performance of the textural features as indicators for the modulation degree of VMAT plans, Spearman’s rank correlation coefficients (rs ) with the plan deliverability were calculated. For comparison purposes, conventional modulation indices for VMAT including the modulation complexity score for VMAT, leaf travel modulation complexity score, and modulation index supporting station parameter optimized radiation therapy (MISPORT) were calculated, and their correlations were analyzed in the same way.
There was no particular textural feature which always showed superior correlations with every type of plan deliverability. Considering the results comprehensively, contrast (d = 1) and variance (d = 1) generally showed considerable correlations with every type of plan deliverability. These textural features always showed higher correlations to the plan deliverability than did the conventional modulation indices, except in the case of modulating parameter differences. The rs values of contrast to the global gamma passing rates with criteria of 2%/2 mm, 2%/1 mm, and 1%/2 mm were 0.536, 0.473, and 0.718, respectively. The respective values for variance were 0.551, 0.481, and 0.688. In the case of local gamma passing rates, the rs values of contrast were 0.547, 0.578, and 0.620, respectively, and those of variance were 0.519, 0.527, and 0.569. All of the rs values in those cases were statistically significant (p < 0.003). In the cases of global and local gamma passing rates, MISPORT showed the highest correlations among the conventional modulation indices. For global passing rates, rs values of MISPORT were −0.420, −0.330, and −0.632, respectively, and those for local passing rates were −0.455, −0.490 and −0.502. The values of rs of contrast, variance, and MISPORT with the MLC errors were −0.863, −0.828, and 0.795, respectively, all with statistical significances (p < 0.001). The correlations with statistical significances between variance and dose-volumetric differences were observed more frequently than the others.
The contrast (d = 1) and variance (d = 1) calculated from fluence maps of VMAT plans showed considerable correlations with the plan deliverability, indicating their potential use as indicators for assessing the degree of modulation of VMAT plans. Both contrast and variance consistently showed better performance than the conventional modulation indices for VMAT.
This research was supported by a grant from the National Research Foundation of Korea (NRF), which is funded by the Korea government (MEST, Grant No. 2014M2A2A7055063).
1. INTRODUCTION 2. MATERIALS AND METHODS 2.A. VMAT plans for prostate and H&N cancer 2.B. Textural features from fluence maps 2.B.1. Generation of fluence maps and GLCM 2.B.2. Textural features 2.C. Plan deliverability of VMAT 2.C.1. 2D pretreatment QA 2.C.2. Differences in modulating parameters between VMAT plans and dynamic log files 2.C.3. Differences in dose-volumetric parameters between VMAT plans and reconstructed VMAT plans with DICOM-RT format log files 2.D. Data analysis 3. RESULTS 3.A. Textural features and conventional modulation indices 3.B. Plan deliverability of VMAT 3.C. Correlations 3.C.1. Textural features vs global gamma passing rates of pretreatment QA 3.C.2. Textural features vs local gamma passing rates of pretreatment QA 3.C.3. Textural features vs differences in modulating parameters between VMAT plans and dynamic log files 3.C.4. Textural features vs differences in dose-volumetric parameters between prostate VMAT plans and reconstructed DICOM-RT format plans using dynamic log files 3.C.5. Textural features vs differences in dose-volumetric parameters between H&N VMAT plans and reconstructed DICOM-RT format plans using dynamic log files 3.D. Sensitivity and specificity 4. DISCUSSION 5. CONCLUSIONS
Data & Media loading...
Article metrics loading...
Full text loading...