Method for setting initial position and posture of liver statistical shape model

Abstract

The invention discloses a method for setting the initial position and posture of a liver statistical shape model, and the method comprises the steps: carrying out the preprocessing of a to-be-segmented image, obtaining a binary image, obtaining a shape grid model of liver, selecting mark points, carrying out the manual segmentation of a plurality of groups of other liver CT images at the same time, carrying out the surface meshing, obtaining a prior shape model, and selecting mark points; taking the mark points on the prior shape model as a training set, and constructing Riemann sparse coding;further constructing a Riemann kernel function based on Stein divergence, and calculating an optimal solution coefficient of a Riemann dictionary; obtaining a deformation field of the mark points relative to the positions of the original mark points through the spatial position relation of the mark points in the updated to-be-segmented image, and mapping the deformation field to each top point inthe shape model of the to-be-segmented image, and achieving the setting of the initial position and posture. The method can achieve the quick construction of the statistical shape model of liver andthe precise determining of the initial position and posture, improves the segmentation efficiency and precision, and is suitable for the segmentation and recognition of liver parenchyma and a focus ofinfection.

Description

technical field [0001] The invention relates to a processing method of a CT image, in particular to a method for setting an initial pose of a statistical shape model of the liver. Background technique [0002] Segmentation method based on statistical shape model is a research hotspot in the field of liver segmentation. Scholars at home and abroad have designed different improvement strategies for this method, but so far, there are still problems to be solved in liver segmentation methods based on statistical shape models: 1) At present, most statistical shape models are constructed using PCA for dimensionality reduction. It is difficult to explain the principal components in the calculation results of the method, and the algorithm efficiency is low for the registration and dimensionality reduction of the large data prior shape model; 2) It is difficult to accurately set the initial pose of the statistical model in the image to be segmented. Although many researchers have pr...

AI Technical Summary

Problems solved by technology

Scholars at home and abroad have designed different improvement strategies for this method, but so far, there are still problems to be solved in liver segmentation methods based on statistical shape models: 1) At present, most statistical shape models are constructed using PCA for dimensionality reduction. It is difficult to explain the principal components in the calculation results of the method, and the algorithm efficiency is low for the registration and dimensionality reduction of the large data prior shape model; 2) It is difficult to accurately set the initial pose of the statistical model in the image to be segmented. Although many researchers have proposed different solutions, the accuracy still needs to be improved

But in reality, it is very difficult to obtain corresponding point information. Although there are some alignment algorithms that can adaptively find corresponding point information, these methods require a large number of sample points, and the computational complexity is very high. They can only be used for some specific situation or data set

Images