Cone-beam computed tomography image and X-ray image registration method

Abstract

The invention discloses a cone-beam computed tomography image and X-ray image registration method; the method uses a regression model based on a mixing residual error convolution nerve network to build associations between two-dimension X-ray images and three dimensional cone-beam CT image non-rigid deformation parameters, and uses a mixing residual error convolution nerve network and deformation parameter iteration optimization method to realize reliable online two-dimension three dimensional image registration; the method comprises the following steps: extracting to obtain an image channel; training the regression model based on the mixing residual error convolution nerve network; carrying out two-dimension three dimensional non-rigid registration based on regression; iterating and optimizing deformation parameters; obtaining a final body image determined by the deformation parameters, and thus realizing two-dimension three dimensional image registration based on iteration regression. The method can realize reliable online two-dimension three dimensional image registration, can be applied to oral cavity clinics, and can evaluate treatments and analyze craniofacial growth according to two-dimension three dimensional image registrations.

Description

technical field [0001] The invention relates to the technical fields of oral clinical medicine and computer vision, in particular to a method for registering cone-beam computed tomography images and X-ray images. Background technique [0002] Orthodontic treatment usually lasts for several years, and multiple images collected during treatment will be used to evaluate the difference between rigid and non-rigid shapes of maxillofacial structures, especially for minor patients. Nonrigid morphological change to structural growth. Before the widespread use of cone-beam computed tomography (cone-beam CT) images in clinical orthodontics, two-dimensional radiographs were the only medium to record changes in maxillofacial morphology. Nonrigid 2D-3D image registration is a key step in obtaining 3D images from in-treatment X-ray images. Traditional 2D and 3D image registration methods rely on iterative optimization to minimize the difference between digitally-reconstructed-radiograph...

AI Technical Summary

Problems solved by technology

The method based on convolutional neural network is used for 2D and 3D image registration, but the existing registration work based on convolutional neural network regression model only deals with a small number of parameters in rigid 2D and 3D image registration, and the single regression Also difficult to deal with the non-rigid complex structural changes involved in craniofacial 3D images

Images