Détermination de la limite de contrôle du taux Gamma par inférence statistique bayésienne. Application aux contrôles pré-traitement - Université d'Angers Accéder directement au contenu
Communication Dans Un Congrès Année : 2019

Détermination de la limite de contrôle du taux Gamma par inférence statistique bayésienne. Application aux contrôles pré-traitement

Résumé

Statistical Process Control (SPC) and control charts have been successfully implemented in radiotherapy in order to detect special causes of variation in processes and thus to increase their quality. Furthermore, the Gamma Index Pass Rate (GIPR), characterized by a Beta distribution, is a widely used parameter for analyzing measurements in medical physics. As a special attention must be given to the computation of the GIPR control limit (CL), we proposed two control charts in order to detect special causes. Measurements were realised on a TrueBeam v2.5 (Varian) equipped with an aS1200 EPID5. Statistical computations were performed with R (v.3.5.1) software. Pre-treatment quality controls compared the dose distribution acquired with the EPID with the calculated dose distribution using the GIPR metric generated by Portal Dosimetry v13.7. The following settings were used: (a) local 3 % / 3 mm (th. 30 %), (b) ROI: MLC + 5 mm. CLs calculations are illustrated in details. The first control chart had an adaptive CL based on Bayesian estimation. It was constructed without prior information on the distribution of the GIPR. The CLs were computed starting from the first arc and continuously adapted according to the recursive Bayesian principle. The second control chart had a fixed control limit computed as the quantile of the Beta distribution. A simple SPC methodology for monitoring GIPR was precisely described and CLs were calculated from the first arc. As a first step, Bayesian inference was used until a sufficient amount of data is collected and, in a second step, a fixed control limit was calculated. The strong points of this method lie in using the true GIPR’s distribution that minimize both: false positive and no detection of special causes. This methodology should be very useful for those who wish to set an individual control chart suited to the viability constraints of their radiotherapy department (prescription, dosimetry, equipment’s).
Fichier non déposé

Dates et versions

hal-02531018 , version 1 (03-04-2020)

Identifiants

  • HAL Id : hal-02531018 , version 1

Citer

Téodor Tiplica, Sébastien Dufreneix, Christophe Legrand. Détermination de la limite de contrôle du taux Gamma par inférence statistique bayésienne. Application aux contrôles pré-traitement. 58èmes Journées Scientifiques de la SFPM, Jun 2019, Angers, France. ⟨hal-02531018⟩
35 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More