Data down sampling for a fast registration of 3D-medical images in radiotherapy.

This paper describes a data down sampling technique increasing the registration speed of two data sets either from the same or from different modalities. The proposed technique is suitable to 3D data given in the shape of 2D parallel slides or images (e.g. MRIdata, Computed TomographyData, etc.). As illustrated in this contribution, the down sampling algorithm can also be used for data (3D clouds of points) acquired with a 3D surfacing sensor. The data reduction algorithm, which selects points Laving a height local curvature, was used in a radiotherapy application. For the latter, patient´s head data obtained by a CT-scan, are registered with the patient´s face data acquired with a 3D surfacing sensor rued in the treatment room in order to determine the tumor position in the irradiation machine coordinate system. Monomodal and multimodal data were down sampled and registered in this context to assess both the impact of data reduction in the registration accuracy and the reduction of registration time.