Liver three-dimensional multi-modality image registration method based on discontinuous motion

Abstract

The invention discloses a liver three-dimensional multi-modality image registration method based on discontinuous motion. The liver three-dimensional multi-modality image registration method comprises the steps of: performing preprocessing and rigid registration on an acquired original liver sequence image; adopting a free-form deformation model combined with a three-order B-spline function to simulate elastic deformation of a liver image, adopting a regularization term based on a total variation and hybrid measurement for constructing a cost function, measuring a similarity degree of the two images, and optimizing and solving the cost function by adopting a limited memory quasi-Newton interpolation method. The liver three-dimensional multi-modality image registration method based on discontinuous motion disclosed by the invention avoids the result of global smooth caused by taking two norms as the regularization term in the prior art, can reserve the boundary discontinuity caused by liver organ motion, and is high in registration precision.

Description

technical field [0001] The present invention relates to the technical field of digital image processing, and more specifically, the present invention relates to a three-dimensional multi-modal liver image registration method based on discontinuous motion. Background technique [0002] In image registration, it is usually desired to obtain a global smooth displacement field, but when the human body performs breathing motion, there will be a relative sliding motion between the lungs and liver and their surrounding tissues. The movement of organs and the movement of their boundaries are broken, so a discontinuous displacement field will be generated between the lung and pleura, and between the liver and the abdominal wall. The traditional cost function constraint term that guarantees global smoothness is no longer applicable. this particular case. [0003] In the prior art, the mainstream method to deal with the registration problem of this kind of discontinuous displacement f...

AI Technical Summary

Problems solved by technology

[0002] In image registration, it is usually desired to obtain a global smooth displacement field, but when the human body performs breathing motion, there will be a relative sliding motion between the lungs and liver and their surrounding tissues. The movement of organs and the movement of their boundaries are broken, so a discontinuous displacement field will be generated between the lung and pleura, and between the liver and the abdominal wall. The traditional cost function constraint term that guarantees global smoothness is no longer applicable. this special situation

[0003] In the prior art, the mainstream method to deal with the registration problem of this kind of discontinuous displacement field is the motion masking method based on segmentation. The motion masking method can only preserve the authenticity of the organ movement, but cannot reflect the effective effect of the external tissue of the organ. The displacement field of ; the motion masking method also requires an initial segmentation operation, which is not suitable for the difficult-to-segment abdominal organs

In addition, some registration methods based on diffusion models have been proposed, such as direction-related regularization, Demons-based adaptive regularization, etc. These methods are based on non-parametric deformation models, and there is no explicit expression for optimization. solve

Images