CN-118570164-B - Method, system and terminal for generating organization mechanical characteristics

CN118570164BCN 118570164 BCN118570164 BCN 118570164BCN-118570164-B

Abstract

The invention discloses a method, a system and a terminal for generating tissue mechanical characteristics, wherein the method comprises the steps of acquiring a tissue information image, preprocessing the tissue information image to obtain a preprocessed information image, acquiring mechanical characteristic information of the preprocessed information image, constructing a tissue mechanical characteristic generation model, and training the tissue mechanical characteristic generation model according to the preprocessed information image and the mechanical characteristic information to obtain a target network model; obtaining a target tissue information image of the tissue to be processed, inputting the target tissue information image into a target network model for calculation to obtain the mechanical characteristics of the target tissue of the tissue to be processed, and obtaining the analysis result of the tissue to be processed according to the mechanical characteristics of the target tissue. The invention combines the physical principle and the deep learning to perform the reasoning calculation of the tissue mechanics characteristic information, can show higher robustness and accuracy under the condition of poor image quality and more spots, and improves the accuracy and reliability of the tissue mechanics characteristic reasoning.

Inventors

ZHOU YONGJIN
QIU ZHUOHUA
CHEN HAOXIN
WU JUNYANG
BAI ZIQI
LI JIALING
WANG ZILI

Assignees

深圳大学

Dates

Publication Date: 20260512
Application Date: 20240530

Claims (6)

1. The tissue mechanics characteristic generating method is characterized by comprising the following steps: Obtaining a tissue information image obtained by sampling a tissue, preprocessing the tissue information image to obtain a preprocessed information image, and obtaining mechanical characteristic information of the preprocessed information image; constructing a tissue mechanical feature generation model, and training the tissue mechanical feature generation model according to the preprocessing information image and the mechanical feature information to obtain a target network model; acquiring a target tissue information image of a tissue to be processed, and inputting the target tissue information image into the target network model for calculation to obtain the mechanical characteristics of the target tissue of the tissue to be processed; Analyzing and evaluating the tissue to be treated according to the mechanical characteristics of the target tissue to obtain an analysis result of the tissue to be treated; training the tissue mechanical feature generation model according to the preprocessing information image and the mechanical feature information to obtain a target network model, wherein the training comprises the following steps of: Taking the preprocessing information image as sample data, taking the mechanical characteristic information as a tag, and establishing a data set according to the sample data and the tag; Dividing the data set into a training set and a testing set according to a preset proportion, inputting the preprocessing information image into the tissue mechanical characteristic generation model, and outputting dimensionless stress and dimensionless form parameters to reduce errors, wherein the dimensionless stress comprises transverse normal stress Longitudinal normal stress Shear stress The parameters of the dimensionless form comprise a parameter of a pull Mei Di of the dimensionless form And two parameters of pull Mei Di in dimensionless form ; According to the transverse normal stress Said longitudinal normal stress Said shear stress A parameter of the non-dimensional form of the pull Mei Di Two parameters of the pull Mei Di in the dimensionless form And the mechanical characteristic information to calculate a loss function : ; Wherein, the The number of coordinate points representing the participation of the primary region in the network training, The number of coordinate points of the upper and lower boundary areas participating in the network training is represented, The number of coordinate points of the left and right boundary areas participating in the network training is represented, Representation of Or alternatively , Representation of Or alternatively , 、 Respectively the abscissa and the ordinate are indicated, Representing the symbol kronecker, Represents the stress of the dimensionless type, A trace representing all of the strain tensors, Representing stress tensors Relative to The partial derivative of the coordinates is used, Representing stress tensors Relative to The partial derivative of the coordinates is used, Representing known stress components on the left and right sides, Representing the known stress distribution at the top and bottom, Representing the mechanical characteristic information corresponding to the dimensionless stress; Adjusting the tissue mechanical feature generation model according to the loss function until the loss function meets a preset requirement to obtain the trained tissue mechanical feature generation model, and performing performance evaluation on the trained tissue mechanical feature generation model by using the test set to obtain the target network model meeting the requirement; The mechanical characteristics of the target tissue comprise elastic modulus, poisson ratio, transverse normal stress, longitudinal normal stress and shear stress; Inputting the target tissue information image into the target network model for calculation to obtain the target tissue mechanical characteristics of the tissue to be processed, wherein the method specifically comprises the following steps: Inputting the target organization information image into the target network model to obtain target dimensionless stress and parameters of a target dimensionless form; Obtaining maximum normal stress of a top boundary according to average anatomical data, calculating according to the maximum normal stress and the parameters of the target dimensionless form to obtain a first parameter of pull Mei Di and a second parameter of pull Mei Di, and calculating according to the first parameter of pull Mei Di and the second parameter of pull Mei Di to obtain the elastic modulus and the poisson ratio; And calculating the transverse normal stress, the longitudinal normal stress and the shear stress according to the maximum normal stress and the target dimensionless stress.
2. The method for generating mechanical characteristics of tissues according to claim 1, wherein the steps of obtaining a tissue information image obtained by sampling tissue, and preprocessing the tissue information image to obtain a preprocessed information image comprise: Acquiring a plurality of pieces of tissue information obtained by scanning a plurality of tissue specific positions through imaging equipment; envelope taking, logarithmic stretching and region selecting processing are carried out on the plurality of tissue information to obtain a plurality of tissue information images; sampling and resampling a plurality of the tissue information images to obtain a plurality of adjusted tissue information images; and performing iterative back projection and denoising processing on the plurality of adjusted tissue information images to obtain a plurality of preprocessed information images.
3. The method for generating mechanical characteristics of tissue according to claim 1, wherein the acquiring mechanical characteristic information of the preprocessed information image specifically includes: Processing the preprocessed information image by using an automatic image processing technology and a high-resolution ultrasonic imaging technology to obtain intermediate characteristic information; And verifying the intermediate characteristic information by using a time sequence analysis and frequency domain analysis method, and taking the intermediate characteristic information after verification as the mechanical characteristic information of the preprocessing information image.
4. A tissue mechanics feature generation system for implementing the tissue mechanics feature generation method of any one of claims 1-3, the tissue mechanics feature generation system comprising: The data acquisition module is used for acquiring a tissue information image obtained by sampling a tissue, preprocessing the tissue information image to obtain a preprocessed information image, and acquiring mechanical characteristic information of the preprocessed information image; The model training module is used for constructing a tissue mechanical feature generation model, and training the tissue mechanical feature generation model according to the preprocessing information image and the mechanical feature information to obtain a target network model; The model prediction module is used for acquiring a target tissue information image of the tissue to be processed, inputting the target tissue information image into the target network model for calculation, and obtaining the mechanical characteristics of the target tissue of the tissue to be processed; and the analysis and evaluation module is used for analyzing and evaluating the tissue to be processed according to the mechanical characteristics of the target tissue to obtain an analysis result of the tissue to be processed.
5. A terminal comprising a memory, a processor and an tissue mechanics feature generation program stored on the memory and operable on the processor, which when executed by the processor, implements the steps of the tissue mechanics feature generation method of any one of claims 1-3.
6. A computer-readable storage medium, characterized in that the computer-readable storage medium stores an tissue mechanics feature generation program which, when executed by a processor, implements the steps of the tissue mechanics feature generation method of any one of claims 1 to 3.

Description

Method, system and terminal for generating organization mechanical characteristics Technical Field The present invention relates to the field of image processing technologies, and in particular, to a method, a system, a terminal, and a computer readable storage medium for generating mechanical characteristics of an organization. Background Advanced computing techniques, such as machine learning and deep learning, are changing the field of muscle function assessment. These techniques accurately measure mechanical characteristic information of tissues by automatically analyzing medical images, which is critical for understanding the condition and function of the human body. Taking the measurement of the elastic modulus of tissue as an example, the application of PINNs (physical-informedNeural Network, PINNs) and ultrasonic imaging techniques in the evaluation of the elastic modulus of tissue is becoming the leading edge of research. But now traditional methods of quantifying tissue elastic modulus, such as shear wave elastography and gray scale ultrasound based methods, while technically advancing, face high sensitivity to image quality and accuracy and robustness challenges in processing images with speckle noise. Furthermore, these techniques tend to employ simplified geometric assumptions, such as treating the tissue as an isotropic material or following a simple elastic modulus strain distribution relationship function, which may not be sufficient to describe the complex deformation of the tissue at maximum contraction or complex motion. While in deep learning based medical imaging methods, despite advances in some aspects, accuracy still needs to be improved at different stages of the disease. These methods rely on extensive data training and complex model structures, making them difficult to meet the needs of real-time processing. Furthermore, these techniques are highly sensitive to image quality and may perform poorly on low quality or speckle-rich images. Accordingly, the prior art is still in need of improvement and development. Disclosure of Invention The invention mainly aims to provide a tissue mechanics characteristic generation method, a system, a terminal and a computer readable storage medium, which aim to solve the problems that the robustness is not high, the tissue mechanics characteristic cannot be comprehensively obtained when the traditional deep learning method is used for processing graphs with different quality or graphs with more noise points in the prior art, the accuracy of the obtained mechanics characteristic is lower, and the calculation efficiency is not high. In order to achieve the above object, the present invention provides a method for generating a mechanical feature of a tissue, the method comprising the steps of: Obtaining a tissue information image obtained by sampling a tissue, preprocessing the tissue information image to obtain a preprocessed information image, and obtaining mechanical characteristic information of the preprocessed information image; constructing a tissue mechanical feature generation model, and training the tissue mechanical feature generation model according to the preprocessing information image and the mechanical feature information to obtain a target network model; acquiring a target tissue information image of a tissue to be processed, and inputting the target tissue information image into the target network model for calculation to obtain the mechanical characteristics of the target tissue of the tissue to be processed; and analyzing and evaluating the tissue to be treated according to the mechanical characteristics of the target tissue to obtain an analysis result of the tissue to be treated. Optionally, in the method for generating a mechanical feature of a tissue, the acquiring a tissue information image obtained by sampling a tissue, and preprocessing the tissue information image to obtain a preprocessed information image specifically includes: Acquiring a plurality of pieces of tissue information obtained by scanning a plurality of tissue specific positions through imaging equipment; envelope taking, logarithmic stretching and region selecting processing are carried out on the plurality of tissue information to obtain a plurality of tissue information images; sampling and resampling a plurality of the tissue information images to obtain a plurality of adjusted tissue information images; and performing iterative back projection and denoising processing on the plurality of adjusted tissue information images to obtain a plurality of preprocessed information images. Optionally, the method for generating mechanical characteristics of tissue, wherein the acquiring mechanical characteristic information of the preprocessed information image specifically includes: Processing the preprocessed information image by using an automatic image processing technology and a high-resolution ultrasonic imaging technology to obtain intermediate characteristic