Nowadays, most of treatments for external radiotherapy are prepared with Treatment Planning Systems (TPS) which uses a virtual patient generated by a set of transverse slices acquired with a CT scanner of the patient in treatment position 1 2 3. In the first step of virtual simulation, the TPS is used to define a ballistic allowing a good target covering and the lowest irradiation for normal tissues. This parameters optimisation of the treatment with the TPS is realised with particular graphic tools allowing to: •Contour the target, •Expand the limit of the target in order to take into account contouring uncertainties, patient set up errors, movements of the target during the treatment (internal movement of the target and external movement of the patient), and beam's penumbra, •Determine beams orientation and define dimensions and forms of the beams, •Visualize beams on the patient's skin and calculate some characteristic points which will be tattooed on the patient to assist the patient set up before treating, •Calculate for each beam a Digital Reconstructed Radiography (DRR) consisting in projecting the 3D CT virtual patient and beam limits with a cone beam geometry onto a plane. These DRR allow one for insuring the patient positioning during the treatment, essentially bone structures alignment by comparison with real radiography realized with the treatment X-ray source in the same geometric conditions (portal imaging).
Then DRR are preponderant to insure the geometric accuracy of the treatment. For this reason quality control of its computation is mandatory4 . Until now, this control is realised with real test objects including some special inclusions4 5 . This paper proposes to use some numerical test objects to control the quality DRR calculation in terms of computation time, beam angle, divergence and magnification precision, spatial and contrast resolutions. The main advantage of this proposed method is to avoid a real test object CT acquisition allowing for a drastic time reduction of the control as well as its automatic control. This method has been used to test a new method to compute DRR6 and is here presented to control a standard DRR calculation algorithm7 .