New frameless patient positioning algorithm in conformal radiotherapy using 3D-data registration

This paper describes a patient positioning method in the field of conformal radiotherapy. The proposed algorithm is dedicated to the treatment of cancers located in the head. The use of stereotactic frames can be avoided by acquiring 3D-data related to the patient's face in the treatment room. The tumor localization in the coordinate system of the radiotherapy machine is computed by registering this 3D-face data with the 3D-head data obtained during a CT-scan performed usually before the treatment. The directed Hausdorff distance has been chosen as the measure of similarity between the two surfaces by considering the nature of the data (two sets of 3D-points of different resolutions and representing surfaces of different sizes). First experimental results have shown that, even if the proposed algorithm gives promising results, the computation time related to the proposed method is not suited to clinical treatment since about fourteen hours are required for the data registration. For a drastic reduction of the computation time, the points giving the most important surface curvature information are selected from the 3D-face data set acquired in the treatment room. This data reduction, obtained with the Fan-algorithm, leads to a typical registration time of fourteen minutes.