ESTRO 2020

Session Item

Physics track: Dose measurement and dose calculation
00:00 - 00:00
Is it possible to reduce the number of DQAs with Tomotherapy HD using Statistical Process Control?
Sandra Losa, France


Is it possible to reduce the number of DQAs with Tomotherapy HD using Statistical Process Control?
Authors: Pascal Francois.(CHU Poitiers, Radiation Therapy, Poitiers, France), Sandra Losa.(Institut Curie Ensemble Hospitalier, Radiothérapie, Paris cedex 05, France), Noelle Pierrat.(Institut Curie Ensemble Hospitalier, Radiothérapie, Paris cedex 05, France), Philip Poortmans.(Institut Curie Ensemble Hospitalier, Radiothérapie, Paris cedex 05, France)
Show Affiliations
Purpose or Objective

In our institution, patient pre-treatment control (DQA) is currently performed for every patient plan prior to treatment delivery with helical tomotherapy. Keeping machine QA within tolerance levels in terms of output, energy, field width and MLC is mandatory to ensure beam matching to the TPS. The purpose of this work is to (1) evaluate the performance of our DQA process over time based on SPC (2) review our DQA pass/fail criteria tolerance levels and action limits (TG 218) and (3) determine a baseline to eventually reduce the number of DQAs

Material and Methods

Since 2014, major hardware (Dose Servo System or DCS and dynamic jaws) and software upgrades were performed on our two Tomotherapy machines. SPC has been implemented as part of our QA program. DQA measurements were performed with a 2D array of ion chambers (IBA Matrixx) in most cases. For small PTVs , Std. Imaging A1SL ion chamber combined with Gafchromic films was preferred because of its higher spatial resolution. A large series of data (n=968 and n=1100 patients in each machine) were followed up using process capability indicators calculated from absolute dose differences (measured/calculated ratios) in the high dose- low dose gradient regions (3% local dose difference) and LOG (Gamma Index failing rate) with 3%/3mm global analysis criteria. Data was plotted as an histogram and Chi-squared and Kolmogorof-Smirnof statistical test was calculated to check that these parameters followed a normal distribution. EWMA control charts were constructed and correlated to major machine service (MLC/linac replacement, jaw actuator, output, energy). A weekly end-to-end test was performed for constancy check and also after a TPS upgrade.


Measured to calculated dose ratios follow a gaussian distribution with estimated 1.013 mean value and 0.007 std. deviation. Lower control limits (LCL) and upper control limits (UCL) were determined (1.006 and 1.020 respectively). LOG of GI failing rate was fitted to a normal distribution. The resulting 95% confidence limit was 2% failing rate using 3%/3mm global gamma pass/fail criteria


Analysis of these data demonstrated that our process is currently stable and under control. However, we were not able to reduce the number of DQAs as a consequence of  frequent machine service that include major components as linac or MLC replacements. Trending analysis through EWMA control charts is a powerful tool for risk assessment. It is a robust method to anticipate drifts and detect out-of-control deviations with clinical impact