Statistical control of the spectral quality index in electron beams.

BACKGROUND AND PURPOSE In radiotherapy with electron linear accelerators, it is mandatory to guarantee the stability of beam output spectra. In this work we study the quality control of the magnitude that provides the beam spectral quality, by following statistical techniques usual in industry process. METHODS We establish the differences between the measurement of the reference quality index, R(50,ion), which involves usually a costly procedure, and that of the index used for the statistical control, which can be done in an easier way and with a higher frequency. The methodology developed is applied to a Siemens Mevatron KDS electron linear accelerator, specifically for the 6, 12 and 18 MeV modes. The uncertainties of both procedures are evaluated and are correlated. A retrospective analysis of the data registered during more than 13 years is carried out using CUSUM algorithmic and moving range charts. RESULTS The larger uncertainties in R(50,ion) come from the measurement device and the observer. The uncertainty in the measurement of the statistical control variable, which is mainly due to the geometry setup, is negligible against its intrinsic variability. The application of CUSUM charts pointed out different out of control situations linked in much cases to assignable causes. In addition, moving range charts showed up situations of anomalous functioning of the LINAC. CONCLUSIONS Algorithmic CUSUM charts permit a nice monitoring of R(50,ion). If the variability is not controlled, by using charts such as the moving range ones, the number of false alarms could increase without a change in the mean value of R(50,ion).