Positioning a patient accurately in treatment devices is crucial forradiological treatment, especially if accuracy vantages of particlebeam treatment are exploited. To avoid sub-millimeter misalignments,X-ray images acquired from within the device are compared to a CTto compute respective alignment corrections. Unfortunately, deviationsof the underlying geometry model for the imaging system degrade theachievable accuracy. We propose an automatic calibration routine,which bases on the geometry of a phantom and its automatic detectionin digital radiographs acquired for various geometric device settingsduring the calibration. The results from the registration of thephantom's X-ray projections its known geometry are used to updatethe model of the respective beamlines, which is used to compute thepatient alignment correction. The geometric calibration of a beamlinetakes all nine relevant degrees of freedom into account, includingdetector translations in three directions, detector tilt by threeaxes and three possible translations for the X-ray tube. Introducinga stochastic model for the calibration we are able to predict thepatient alignment deviations resulting from inaccuracies inherentto the phantom design and the calibration. Comparisons of the alignmentresults for a treatment device without calibrated imaging systemsand a calibrated device show, that an accurate calibration can enhancealignment accuracy.     «

Positioning a patient accurately in treatment devices is crucial forradiological treatment, especially if accuracy vantages of particlebeam treatment are exploited. To avoid sub-millimeter misalignments,X-ray images acquired from within the device are compared to a CTto compute respective alignment corrections. Unfortunately, deviationsof the underlying geometry model for the imaging system degrade theachievable accuracy. We propose an automatic calibration routine,which bases on the geometry of...     »