Send to

Choose Destination
Med Phys. 2005 Apr;32(4):838-50.

Retrospective analysis of 2D patient-specific IMRT verifications.

Author information

Department of Radiation Physics, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Boulevard, Unit 94, Houston, Texas 77030, USA.


We performed 858 two-dimensional (2D) patient-specific intensity modulated radiotherapy verifications over a period of 18 months. Multifield, composite treatment plans were measured in phantom using calibrated Kodak EDR2 film and compared with the calculated dose extracted from two treatment planning systems. This research summarizes our findings using the normalized agreement test (NAT) index and the percent of pixels failing the gamma index as metrics to represent the agreement between measured and computed dose distributions. An in-house dose comparison software package was used to register and compare all verifications. We found it was important to use an automatic positioning algorithm to achieve maximum registration accuracy, and that our automatic algorithm agreed well with anticipated results from known phantom geometries. We also measured absolute dose for each case using an ion chamber. Because the computed distributions agreed with ion chamber measurements better than the EDR2 film doses, we normalized EDR2 data to the computed distributions. The distributions of both the NAT indices and the percentage of pixels failing the gamma index were found to be exponential distributions. We continue to use both the NAT index and percent of pixels failing gamma with 5%/3 mm criteria to evaluate future verifications, as these two metrics were found to be complementary. Our data showed that using 2%/2 mm or 3%/3 mm criteria produces results similar to those using 5%/3 mm criteria. Normalized comparisons that have a NAT index greater than 45 and/or more than 20% of the pixels failing gamma for 5%/3 mm criteria represent outliers from our clinical data set and require further analysis. Because our QA verification results were exponentially distributed, rather than a tight grouping of similar results, we continue to perform patient-specific QA in order to identify and correct outliers in our verifications. The data from this work could be useful as a reference for other clinics to indicate anticipated trends in 2D verifications under various conditions.

[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Wiley
Loading ...
Support Center