Improving the gamma analysis comparison using an unbinned multivariate test
Ramos Garcia LI (1), Perez Azorin JF (2), Aguilar-Redondo PB (3), Moran-Velasco V (3).
In this study we present a new procedure for the comparison of two dose matrices by means of a statistical test. A statistical distance is proposed to decide whether the difference between the two matrices is statistically significant.
This statistical test is based on the square difference between the experimental and expected gamma matrix results. The expected gamma matrix is calculated by simulating the measurement process. The criteria for significance is chosen to give the same α-type error as the usual gamma-pass rate criteria, i.e. when the proportion of detectors with gamma value less than one is greater than 90%. The performance of the statistical distance is checked against 53 VMAT.
The power of the presented test is compared using simulations with the usual gamma-pass rate criteria for two cases in which intentional errors are introduced. In both cases, the test is uniformly more powerful.
According to the test, two of the measured matrices have a significant difference with calculated matrices, although the gamma pass rate is always greater than 90%.
CITATION Phys Med Biol. 2017 Sep 1;62(18):N417-N427. doi: 10.1088/1361-6560/aa848f.