Automatic detection of device setup errors in radiotherapy by matching three-dimensional CT data and portal images

In order to achive best results in radiotherapeutic treatment of tumor tissue, one needs to precisely reconstruct the setup defined in therapy planning before the treatment can start. Inaccuracies in either defining the treatment parameters or transferring positioning markers to the patient’s body, errors in treatment device setup or patient motion during the treatment result in a displacement of the irradiated volume. If the dose applied to the clinical target volume (CTV) is reduced, the tumor may not be controlled as predicted and the risk of recidivists rises. In view of the different possible inaccuracies, a planning target volume (PTV) is defined which is larger than CTV and incorporates it. The major goal of the described project is to minimize the difference of PTV and CTV. In conventional radiotherapy, the patient’s setup is performed by using isocenter laser pointers and light projections of the treatment beam which have to match markers on the patient’s skin. Then, two or more portal images (PI) are generated to control the patient’s positon and to correct setup errors. To detect errors at least semi-automatically, artificial markers must be fastened to the patient or anatomical landmarks have to be selected. This paper discusses an approach where a computer detects and corrects errors at high precision without the need of manual interaction or use of artificial landmarks.