Program Information
Performance Comparisons of Patient Specific IMRT QA Methodologies Using ROC Analysis
E McKenzie1*, P Balter2 , F Stingo2 , D Followill2 , J Jones3 , S Kry2 , (1) Cedars-Sinai Medical Center, Los Angeles, CA, (2) UT MD Anderson Cancer Center, Houston, TX, (3) Centura Health, Denver, CO
Presentations
TU-C-BRE-9 Tuesday 10:15AM - 12:15PM Room: Ballroom EPurpose:
To evaluate the ability of a selection of patient-specific QA methods to accurately classify IMRT plans as acceptable or unacceptable based on a multiple ion chamber (MIC) phantom.
Methods:
Twenty-four IMRT plans were selected (20 previously failed the institutional QA), and were measured on a MIC phantom to assess their dosimetric acceptability. These same plans were then measured using film (Kodak EDR2) and ion chamber (Wellhofer cc04), ArcCheck (Sun Nuclear), and MapCheck (Sun Nuclear) (delivered AP field-by-field, AP composite, and with original gantry angles). All gamma analyses were performed at 2%/2mm, 3%/3mm, and 5%/3mm. By using the MIC results as a gold standard, the sensitivity and specificity were calculated across a range of cut-off thresholds (% pixels passing for gamma analysis, and % dose difference for ion chamber), and were used to form ROC curves. Area under the curve (AUC) was used as a metric to quantify the performance of the various QA methods.
Results:
Grouping device’s AUC’s revealed two statistically significant different groups: ion chamber (AUC of 0.94), AP composite MapCheck (AUC of 0.85), ArcCheck (AUC of 0.84), and film (AUC of 0.82) were in the better performing group, while original gantry angles and AP field-by-field MapCheck (AUC of 0.65 and 0.66, respectively) matched less well with the gold standard results. Optimal cut-offs were also assessed using the ROC curves. We found that while often 90% of pixels passing is used as a criteria, the differing sensitivities of QA methods can lead to device and methodology-based optimal cutoff thresholds.
Conclusion:
While many methods exist to perform the same task of patient-specific IMRT QA, they utilize different strategies. This work has shown that there are inconsistencies in these methodologies in terms of their sensitivity and specificity to dosimetric acceptability.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by Public Health Service grants CA010953, CA081647, and CA21661 awarded by the National Cancer Institute, United States Department of Health and Human Services.
Contact Email: