Résumé
The image guided radio therapy (IGRT) Imaging System based on cone-beam computer tomography (CBCT) can reach the goal of improving the accuracy of the radiotherapy. However, because the clinical registration between CBCT images and Planning CT images is carried out manually, it inevitably reduces radiation positioning accuracy to some extent. To tackle the problem, we proposed a new feature vector selection method for the CBCT image elastic registration in the framework of hierarchical attribute matching mechanism for elastic registration (HAMMER) algorithm. We analyzed the characteristics of HAMMER algorithm and used Canny operator which has a better edge detection and positioning performance to replace the noise-sensitive gradient amplitude. Therefore, we used a new attribute vector, which consisted of the intensity, Laplacian of the Gaussian and Canny operator, to ex tract the image feature points in CBCT and planning CT images. We also presented an adaptive feature-point selection method and the choice criteria of attribute vector weights. Experimental results showed that the new feature vector effectively avoided the noise interference resulted from scattering lines in CBCT images to improve registration accuracy, and it also decreased the required feature point numbers and reduced the computation redundancy, so that it provided a new approach for the clinical elastic registration of CBCT and Planning CT rapidly and accurately.