Automated Position Control in Conformal Radiotherapy

This work is part of a collaboration with the cancer institute L eon B erard (Lyon, France) and concerns conformal radiotherapy. The current medical process begins by acquiring a reference computed tomog-raphy volume of a patient, and by computing an irradiation procedure. During several weeks, the patient comes every day for an irradiation session. In order to have a safe and eecient irradiation, the patient must be perfectly positionned. This is currently achieved by manual positioning. We propose in this work an automated system to characterize and to estimate the displacement of the patient as compared to the reference position. The presented method involves three parts: a calibration procedure, a volume rendering stage to produce Digitally Reconstructed Radiographs and an intensity-based registration step. We experiment in this last step several similarity measures in order to make the registration robust and accurate. First experiments show that our automated method leads to a far better accuracy than the manual one, and requires low computation times compatible with a clinical use of the system.