CN108846896A - A kind of automatic molecule protein molecule body diagnostic system - Google Patents

Abstract

The invention belongs to the technical field of medical diagnosis, and discloses an automatic molecular protein molecular biology diagnosis system. The automatic molecular protein molecular biology diagnosis system includes: a protein detection module, a DNA detection module, a main control module, a DNA sequencing module, a cancer Cell detection module, expert analysis module, data storage module, display module; the detection method of protein detection module includes drawing the outer boundary contour line and internal organization of protein molecules according to the gray scale or texture characteristics of magnetic resonance or computed tomography voxel data edge line. The present invention can perform DNA sequencing quickly, accurately and at low cost through the DNA sequencing module; at the same time, the cancer cell detection module can easily detect cancer cells in peripheral blood in a short time; the expert analysis module can more professionally detect The data is analyzed to ensure the reliability of the test results.

Description

An Automatic Molecular Protein Molecular Body Diagnosis System

technical field

The invention belongs to the technical field of medical diagnosis, and in particular relates to an automatic molecular protein molecular body diagnosis system.

Background technique

At present, the existing technologies commonly used in the industry are as follows:

Molecular diagnosis refers to the technique of detecting changes in the structure or expression level of genetic material in patients by using molecular protein molecular body methods to make a diagnosis. Molecular diagnosis is the main method of predictive diagnosis, which can be used for diagnosis of individual genetic diseases, infectious diseases, etc., as well as prenatal diagnosis. Molecular diagnosis mainly refers to the genetic detection of various structural proteins, enzymes, antigen antibodies, and immune active molecules that encode various diseases. However, the existing automatic molecular protein molecular body diagnosis system has disadvantages such as slow speed of DNA sequencing, high cost, and long time-consuming detection of cancer cells at the same time.

Optical 3D imaging is an emerging optical imaging technology, which integrates the multi-angle optical signals measured on the surface of protein molecules, the anatomical structure of protein molecules and the information of tissue optical parameters, based on the precise light transmission in protein molecule tissues The model reconstructs the position and intensity distribution information of target targets in vivo in living protein molecules. Among them, the accurate description of the light transmission process in the protein molecular body organization and the accurate and fast reconstruction of the targeted target are the basis for the realization of the optical three-dimensional imaging method. Beijing University of Technology proposed a method in its patent application document "Multispectral autofluorescence tomography reconstruction method based on single view" (application number 200810116818.4, application date 2008.7.18, authorization number ZL200810116818.4, authorization date 2010.6.2) Single-view-based multispectral autofluorescence tomography reconstruction method. Based on the diffusion approximation equation, this patented technology considers the heterogeneous properties of protein molecules and the spectral characteristics of autofluorescence light sources, and uses the fluorescence data of multiple bands measured at a single angle to reconstruct the position and intensity of target targets in protein molecules distribution information. However, since the diffusion approximation equation is only suitable for describing the light transmission process in tissues with high scattering characteristics, its solution accuracy is very low for tissues with low scattering characteristics and cavities. Therefore, the patented technology has poor solution accuracy for protein molecular bodies with various scattering characteristics, and it is difficult to accurately obtain the position and intensity distribution information of the target target in the protein molecular body. Xidian University proposed a patent application document "Optical 3D Imaging Method Based on Protein Molecular Body Tissue Specificity" (application number 201110148500.6, application date 2011.6.2, authorization number ZL201110148500.6, authorization date 2013.4.3). Optical 3D imaging method based on protein molecular body tissue specificity. This patent establishes an objective function based on a mixed mathematical model of protein molecule tissue-specific optical transmission and a fully sparse regularization method, and uses a task-oriented hybrid optimization method to solve it to achieve optical three-dimensional imaging of targeted targets in vivo, which solves the existing problems Accurate and rapid optical three-dimensional imaging of complex protein molecular bodies with irregular anatomical structures and various scattering characteristics cannot be realized in the technology. However, in the optical three-dimensional imaging method based on the heterogeneous model and protein molecular body tissue specificity, accurate and effective segmentation of tissues and organs in the protein molecular body and high-quality numerical discretization of the mesh are the key to accurately construct and solve the optical imaging model. essential key steps. Organ segmentation is a complex and tedious task that requires professional software and human-computer interaction to complete. Grid discretization not only requires professional software and human-computer interaction, but also requires different quality of grid discretization for different imaging requirements. At the same time, there are uncontrollable factors in the discretization of the grid, which leads to the uncontrollable influence of the quality of the discretization of the grid on the solution and reconstruction of the model.

Computer graphics data processing involves many disciplines, mainly including: target detection and recognition, edge extraction, feature extraction and 3D reconstruction. 3D reconstruction technology is also an image-based modeling technology, which has attracted much attention since its inception. This method only needs two frames of adjacent images to more accurately restore the 3D spatial relationship between the matching feature points in the image and the camera. In this process, the number of matching feature points directly determines the quality of the point cloud acquired by 3D reconstruction, thereby determining the quality of the reconstructed model.

There are three types of commonly used 3D reconstruction methods: (1) Stereo vision methods. This method simulates the way the human visual system perceives objective three-dimensional objects, uses two or more cameras to image the same scene at different positions, and then converts it into a depth map according to the disparity map between the two frames of images to obtain the depth of the object information. The geometric model files generated by this method are usually relatively small and can be easily used in virtual reality. However, this method needs to overcome the problem of sparse object features. When the texture is flat, there are large blank areas in the calculated disparity map, and the density of the point cloud is very low. (2) Motion structure method. When shooting around the object, the movement of any point on the rigid object between the two frames of images is the same. By extracting several pairs of feature points between the two frames of images and matching them, the transformation of the object's motion can be calculated. Matrix, according to the transformation matrix, the positional relationship between the two cameras can be determined, and the coordinates of the feature points in the world coordinate system can be recovered through the principle of pinhole imaging. This method is relatively mature. It can calculate the movement of the camera when the internal parameters of the camera are calibrated. Processing the sparse point cloud can obtain a denser point cloud and restore a more accurate 3D model. However, it requires a larger number of matching feature points between two adjacent frames, so the number of effective points in the feature flat area is less. (3) Method based on depth image. The point cloud of the object in the current camera coordinate system can be generated through the RGB image and the depth image of each frame of the image, and the two sets of point clouds generated by the adjacent two frames of RGBD images are matched to calculate the transformation matrix of the two frames of the camera. The two sets of point clouds are fused to the world coordinate system. The point cloud calculated by this method is more accurate, and the density of the point cloud is higher. However, it requires the assistance of a depth camera and is very sensitive to the accuracy of the depth map. In a large-scale reconstruction scene, the accuracy of the depth camera is always limited, and the accuracy of the depth camera will directly correlate with the accuracy of the reconstructed point cloud.

Among the above methods, the stereo vision method requires less calculation, but there are blank areas in the disparity map obtained in the flat area of the image texture, so the calculated point cloud density is very low; the motion structure method has high universality , which includes the derivation process from sparse point cloud to dense point cloud, but the density of the obtained dense point cloud still depends on the texture complexity of the image. For images with flat texture, the obtained point cloud density is relatively low; based on depth The image reconstruction method has high precision and does not require the texture complexity of the image. However, this method is highly sensitive to the accuracy of the depth camera and is currently not suitable for 3D reconstruction of large-scale objects.

In summary, the problems in the prior art are:

The existing automatic molecular protein molecular body diagnosis system has a slow speed and high cost for DNA sequencing, and it takes a long time to detect cancer cells at the same time.

In the prior art, cumbersome organ segmentation and grid discretization are required to obtain optical three-dimensional imaging reconstruction results.

The effectiveness of the method for improving the density of 3D reconstruction point clouds in the present invention is not limited to a specific sequence of round-robin images, and does not depend too much on the adjustment of parameters, and can be improved in a relatively short period of time with a relatively low amount of calculation. The density of the effective point cloud, and the error point cloud in the original point cloud can be deleted, so that the obtained point cloud can overcome the influence of texture flatness to a certain extent.

Contents of the invention

Aiming at the problems existing in the prior art, the invention provides an automatic molecular protein molecular physical diagnosis system.

The present invention is achieved like this, a kind of automatic molecular protein molecular biology diagnosis system, comprises:

The protein detection module is connected with the main control module, and is used to detect the protein by the protein detector; the protein detector draws the outer boundary contour line and The internal tissue edge line; based on the voxel data reconstructed by magnetic resonance or computed tomography and the marked internal tissue edge line, the enrichment function of the internal boundary node is constructed; considering the structural heterogeneity and optical specificity of protein molecular organization, a method based on The adaptive light transmission mathematical model of the mixed light transport equation describes the transmission process of light particles in the protein molecular body; in view of the application advantages of the finite volume method on the hexahedral voxel grid, the adaptive light transmission mathematical model is developed by using the extended finite volume method Carry out numerical discretization and solution, establish a system equation describing the linear relationship between in vivo targets and body surface measurements; consider the sparsity of in vivo target distribution and the incompleteness of body surface measurement data, establish a strategy based on sparse regularization and fusion prior The objective function of the preliminary target localization results; use an appropriate optimization method to solve the objective function, and carry out accurate and rapid reconstruction of the target target in the protein molecular body;

The DNA detection module is connected with the main control module, and is used for detecting DNA by a DNA detector;

The main control module is connected with the protein detection module, DNA detection module, DNA sequencing module, cancer cell detection module, expert analysis module, data storage module, and display module to control the normal operation of each module;

The DNA sequencing module is connected to the main control module and is used to sequence the detected DNA;

The cancer cell detection module is connected with the main control module and is used to detect cancer cells in the peripheral blood;

The expert analysis module is connected with the main control module, and is used for online analysis of the detected data through the online expert comment network;

The data storage module is connected with the main control module and is used to store the detected data information;

The display module is connected with the main control module and is used for displaying detected data information.

Further, the detection of protein by the protein detector also includes:

(1) After reconstruction, obtain a set of round-trip image sequences through photographic equipment, extract object contours for each frame of round-trip images, set the pixel values in the contour area to 255, and set the pixel values outside the contour to 0, and obtain A frame of binary image is called the effective area map; a point cloud with low density is obtained, which is called the initial point cloud, and the rotation matrix R and translation vector t of each frame of the camera relative to the world coordinate system are also obtained, and the rotation The matrix is combined with the translation vector to form the transformation matrix M;

(2) Traverse each point in the initial point cloud, obtain the maximum and minimum values of all points in the initial point cloud on the three axes of x, y, and z, and calculate the maximum and minimum values on each axis The distance difference between them is recorded as x_dis, y_dis, and z_dis respectively, and the three distance differences are divided by 100, and the three quantities obtained are called the derived scale of the initial point cloud, and are recorded as x_scalar, y_scalar, and z_scalar;

(3) Take a point in the initial point cloud as the source point, and expand the corresponding derived scales calculated in step (2) along the positive and negative directions of the three directions of x, y, and z respectively, and obtain a point centered on the source point The cuboid, the length, width and height of the cuboid are 2*x_scalar, 2*y_scalar, 2*z_scalar respectively, the center of the source point extends to 26 directions around the cuboid, and a new point is derived in each direction, taking The normal vector of this new point is the same as the normal vector of the source point, and each derived point records its source point;

(4) all carry out step 4) described derivation operation to each point in the initial point cloud, will obtain a derived point cloud, the quantity of this point cloud midpoint is 26 times of initial point cloud quantity;

(5) For the i-th frame image in the round shot image sequence, take out the transformation matrix M _i calculated in step (1), and transform the derived point cloud obtained in step (4 ₎ into the corresponding In the camera coordinate system of , and according to the projection principle, each point in the derived point cloud is back-projected onto the effective area map of the i-th frame obtained in the accurate and fast reconstruction step of the target target in the protein molecular body;

According to the step in (5), for the points projected into the invalid area in the i-th frame effective area map, delete it from the derived point cloud, and the points projected into the valid area in the i-th frame effective area map are then reserve;

For each frame in the round shot image sequence, perform the above step (4) and delete it from the derived point cloud, and keep the points in the valid area projected to the i-th frame valid area map, by deriving The point cloud is projected around and deleted, and the 3D reconstruction obtains a derived point cloud containing internal points;

Traverse each derived point in the obtained derived point cloud, judge whether each derived point is an internal point or an external point, delete the derived point cloud that is an internal point, and keep the derived point cloud that is an external point; finally retain the point Cloud is a valid point cloud derived once;

Count the number of effective point clouds obtained. If the density meets the requirements, the effective point cloud is the final point cloud; if the density does not meet the requirements, the effective point cloud is used as the initial point cloud. Steps until the obtained effective point cloud meets the density requirement.

Further, in step (3), one of the points in the initial point cloud is used as the source point to derive new points from 26 directions of the cuboid, and the calculation formula of the new points is:

Among them, x_org, y_org, and z_org are the coordinates of a point in the initial point cloud on the x, y, and z axes, respectively, and x_scalar, y_scalar, and z_scalar are the calculated derived scales in the three directions of x, y, and z, respectively.

The 3*3*3 new point coordinates calculated by the above formula, except for the case where the source point coordinate increment is (0, 0, 0), will derive 26 new point clouds described in step (3);

In step (5), the calculation formula for transforming the derived point cloud into the camera coordinate system of the i-th frame image:

(x_cam _i ,y_cam _i ,z_cam _i )=(x_world,y_world,z_world)*R _i |t _i

Among them, (x_world, y_world, z_world) are the coordinates of the derived point cloud in the world coordinate system, R _i and t _i are the rotation matrix and translation vector of the i-th frame camera respectively, after the transformation of R _i and t _i , the world The point cloud in the coordinate system is transferred to the i-th frame camera coordinate system, that is, the coordinates of the transformed point cloud in the i-th frame camera coordinate system are (x_cam _i , y_cam _i , z_cam _i );

The point cloud in the camera coordinate system is back-projected, and each point is projected into the effective area map of the i-th frame. The calculation formula of the projection position is:

Among them, f is the focal length of the camera, C _x and _Cy are 0.5 times the resolution of the image respectively, and the calculated u and v are the projected positions of the point on the image, that is, the positions corresponding to the uth row and vth column on the image pixel location.

Further, the construction of a voxel-based physical model specifically includes:

The first step is to use the registration software in the multimodal molecular imaging system to register the 3D voxel data reconstructed by magnetic resonance or computed tomography to the existing public digital mouse atlas, so as to draw and label protein molecules External contour lines and internal organizational boundaries;

In the second step, based on the 3D voxel data and the marked internal tissue boundary lines, a boundary node enrichment function is constructed:

Among them, j is a voxel node;

ψ _j (r) is the defined inner boundary node enrichment function;

v _j (r) is the linear interpolation basis function;

is a signed distance function, defined as the distance from a node to its nearest closed boundary:

Among them, sign(r) is used to indicate the affiliation relationship between point r and boundary Γ: if the point is inside the region, the value is negative, outside the region, it is positive, and on the boundary, it is zero;

is the value of the signed distance function on the voxel node j;

The third step is to use the marked internal tissue boundary as the interface to decompose the protein molecular body into a collection of multiple organs, assign the tissue optical characteristic parameters to the corresponding organs, and construct a voxel-based optical three-dimensional imaging physical model.

Further, the construction of the adaptive optical transmission mathematical model specifically includes:

In the first step, according to the decomposed multiple organs and the corresponding tissue optical characteristic parameters, the organs are divided into three categories: high scattering, cavity and other tissues, and the classification basis is defined as:

Among them, Ω is the solution domain composed of protein molecules; Ω _hs is the high scattering tissue area; Ω _v is the cavity area; Ω _ls is the other tissue area; μ _'s is the tissue reduced scattering coefficient; ζ and χ are the classification thresholds , respectively taken as ζ=10 and χ=0.2mm ^-1 ;

In the second step, comprehensively considering the accuracy and computational complexity, the appropriate light transmission model is adaptively selected for different types of tissues to describe; among them, the diffusion approximation equation is used to describe the light transmission process in highly scattering tissues, and the free space The light transmission equation describes the transmission process of light in the cavity, and uses the third-order simplified spherical harmonic approximation equation to describe the transmission process of light in other tissues;

The third step is to construct an adaptive optical transmission mathematical model by constructing the boundary coupling conditions of physical quantities between different optical transmission models:

Among them, φ _i (r) (i=1, 2) is the node optical flux, S(r) is the energy density distribution of the protein molecule optical probe, μ _a (r) and μ _aj (r) (j=1 ,2,3) are parameters related to protein molecular body absorption, D(r) is protein molecular body diffusion coefficient, β _i (i=1,2) and α are boundary mismatch factors between SP ₃ and DA equation, G(r′ ,r) is the transfer function describing the concept of radiative transfer theory, which is used to describe the transmission process of diffuse light from the cavity tissue, B is the interface between the scattering tissue and the cavity, and σ(r) is an indicator describing the position of the solution point factor, which is:

Apply the following light transport equation coupling highly scattering and other scattering tissues:

Among them, φ ₀ (r) is the nodal optical flow for solving the diffusion approximation equation;

Apply the following light transport equation coupling the scattering tissue to the cavity:

Among them, q ₀ (r) is the Norman luminous flux formed on the interface between the cavity and the scattering tissue.

Further, the establishment of system equations by said fusion enrichment function specifically includes:

Taking the constructed voxel-based physical model as the solution domain, numerically discretize and solve the constructed adaptive light transmission mathematical model by using the finite volume method of the internal boundary node enrichment function of the fusion structure, and establish a description of the target and volume of protein molecules in vivo. A system of equations that express a linear relationship between measurements:

J = AS;

Among them, A is the system matrix, which depends on the distribution of the three types of protein molecular bodies in the protein molecule body and the corresponding optical characteristic parameters; J is the outgoing optical flow rate collected on the surface of the protein molecule body; S is the target energy density distribution.

Another object of the present invention is to provide a computer program for realizing the operation method of the automatic molecular protein molecular biology diagnosis system.

Another object of the present invention is to provide an information data processing terminal equipped with the automatic molecular protein molecular body diagnosis system.

Another object of the present invention is to provide a computer-readable storage medium, including instructions, which, when run on a computer, enable the computer to execute the method for running the automatic molecular protein molecular biology diagnosis system.

Another object of the present invention is to provide an automatic molecular protein molecular body diagnostic equipment equipped with the automatic molecular protein molecular body diagnostic system.

Advantage of the present invention and positive effect are:

The present invention can perform DNA sequencing quickly, accurately and at low cost through the DNA sequencing module; at the same time, the cancer cell detection module can easily detect cancer cells in peripheral blood in a short time; the expert analysis module can more professionally detect The data is analyzed to ensure the reliability of the test results.

Since the present invention directly performs optical three-dimensional reconstruction on the voxel data reconstructed by magnetic resonance or computed tomography imaging, it overcomes the problem in the prior art that organ segmentation and grid discreteness must be performed to complete the three-dimensional reconstruction of the targeted target, fundamentally avoiding the It eliminates the complicated organ segmentation and grid discretization, simplifies the reconstruction process of optical three-dimensional imaging, and realizes accurate, efficient and easy-to-use optical three-dimensional imaging.

Since the present invention considers the differences in the anatomical structure and tissue optical characteristic parameters of protein molecules at the same time to establish a mixed mathematical model of light transmission, it overcomes the reconstruction accuracy of optical three-dimensional imaging methods based on single approximation equations or mixed light transmission equations in the prior art. And efficiency limitations, enabling accurate and rapid imaging of targeted targets in complex protein molecular bodies organized with irregular anatomy and diverse scattering properties.

In the present invention, the detection results of magnetic resonance or computed tomography data are used as the prior preliminary target positioning results to limit the feasible range of the system equation solution, which overcomes the inaccurate problem of direct positioning and reconstruction in the prior art, and is effective Accurate positioning and quantification of the target was achieved.

The present invention is compared with the method of obtaining point cloud based on stereo vision: the method of obtaining point cloud based on stereo vision needs to provide an image sequence with complex texture, there are no points in the area without disparity map in the reconstruction process, and the reconstruction error is affected by the error of disparity map solution. However, the present invention does not have too many requirements on the texture of the image. As long as the provided initial point cloud can be relatively close to the shape of the real object, most of the information lost in the initial point cloud can be recovered to a certain extent.

Compared with the method of obtaining point cloud based on motion structure: the number of point clouds obtained by the method of obtaining point cloud based on motion structure depends on the number of effective matching feature point pairs between adjacent two frames, and the sparse point cloud to dense point cloud is adopted. The calculation of the derivation method is complicated. However, the number of derived point clouds generated in the present invention is not directly related to the texture of the image, and there are no too many requirements for the initial point cloud. As long as it is closer to the real object, the number of point clouds in places where the original point cloud distribution is sparse can be made by means of derivation. increase, to increase the number of valid point clouds.

Compared with the method based on the depth image: the method based on the depth image needs to provide the depth map of each frame image, and the algorithm is more sensitive to the accuracy of the depth map. The matching between the two point clouds uses an iterative algorithm, so that all It requires a lot of calculations, and there are many matrix operations, which need to be calculated on the GPU. However, the method proposed by the present invention has no requirements on the depth map. With the filtering of the derived point cloud, the number of points to be calculated is also reduced, and the calculation speed is increased. It can also have a faster speed without calculation on the GPU.

Description of drawings

Fig. 1 is a structural diagram of an automatic molecular protein molecular body diagnosis system provided by an embodiment of the present invention.

In the figure: 1. Protein detection module; 2. DNA detection module; 3. Main control module; 4. DNA sequencing module; 5. Cancer cell detection module; 6. Expert analysis module; 7. Data storage module; 8. Display module .

Detailed ways

In order to further understand the content, features and effects of the present invention, the following examples are given, and detailed descriptions are given below with reference to the accompanying drawings.

As shown in Figure 1, the automatic molecular protein molecular biology diagnosis system provided by the embodiment of the present invention includes: a protein detection module 1, a DNA detection module 2, a main control module 3, a DNA sequencing module 4, a cancer cell detection module 5, an expert An analysis module 6, a data storage module 7, and a display module 8.

The protein detection module 1 is connected with the main control module 3, and is used to detect the protein by the protein detector;

The DNA detection module 2 is connected with the main control module 3, and is used for detecting DNA by a DNA detector;

The main control module 3 is connected with the protein detection module 1, the DNA detection module 2, the DNA sequencing module 4, the cancer cell detection module 5, the expert analysis module 6, the data storage module 7, and the display module 8, and is used to control the normal operation of each module;

The DNA sequencing module 4 is connected to the main control module 3 for sequencing the detected DNA;

Cancer cell detection module 5, connected with main control module 3, for detecting cancer cells in peripheral blood;

The expert analysis module 6 is connected with the main control module 3, and is used to carry out online analysis to the detected data through the online expert comment network;

The data storage module 7 is connected with the main control module 3, and is used for storing detected data information;

The display module 8 is connected with the main control module 3 and is used for displaying the detected data information.

The present invention will be further described below in conjunction with specific analysis.

The detection method of the protein detection module includes: 1) according to the grayscale or texture characteristics of the magnetic resonance or computed tomography voxel data, drawing the outer boundary contour line and the internal tissue edge line of the protein molecule; Voxel data and marked internal tissue edge lines to construct internal boundary node enrichment functions; considering the structural heterogeneity and optical specificity of protein molecular organization, an adaptive light transmission mathematical model based on mixed light transmission equations is used to describe light particles The transmission process in the protein molecular body; in view of the application advantages of the finite volume method on the hexahedral voxel grid, the extended finite volume method is used to numerically discretize and solve the mathematical model of adaptive light transmission, and to establish a description of the target in the body and the measurement of the body surface The system equation of the linear relationship between the values; considering the sparsity of the target distribution in the body and the incompleteness of the body surface measurement data, an objective function based on the sparse regularization strategy and the fusion of the prior preliminary target localization results is established; a suitable optimization method is used to solve the problem Objective function, to realize accurate and rapid reconstruction of protein molecular targets in vivo;

2) After reconstruction, obtain a set of round-trip image sequences through photographic equipment, extract object contours for each frame of round-trip images, set the pixel values in the contour area to 255, and set the pixel values outside the contour to 0, and obtain a The frame binary image is called the effective area map; a point cloud with a very low density is obtained, which is called the initial point cloud, and the rotation matrix R and translation vector t of each frame of the camera relative to the world coordinate system, the rotation matrix Combined with the translation vector to form the transformation matrix M;

3) Traverse each point in the initial point cloud, obtain the maximum value and minimum value of all points in the initial point cloud on the three axes of x, y, and z, and calculate the difference between the maximum value and the minimum value on each axis The distance difference between them is recorded as x_dis, y_dis, and z_dis respectively, and the three distance differences are divided by 100, and the three quantities obtained are called the derived scale of the initial point cloud, which are recorded as x_scalar, y_scalar, and z_scalar;

4) Take a point in the initial point cloud as the source point, and expand the corresponding derived scales calculated in step 3) along the positive and negative directions of the three directions of x, y, and z respectively, and obtain a cuboid centered on the source point , the length, width, and height of the cuboid are 2*x_scalar, 2*y_scalar, and 2*z_scalar respectively. The center of the source point extends to 26 directions around the cuboid, and a new point is derived in each direction. Take the new The normal vector of the point is the same as the normal vector of the source point, and each derived point records its source point;

5) performing the derivation operation described in step 4) on each point in the initial point cloud, a derived point cloud will be obtained, and the number of points in the point cloud is 26 times that of the initial point cloud quantity;

6) For the i-th frame image in the round shot image sequence, take out the transformation matrix M _i calculated in step 2), and transform the derived point cloud obtained in step 5) to the corresponding camera coordinates according to the transformation matrix M _i system, and according to the projection principle, back-project each point in the derived point cloud onto the effective area map of the i-th frame obtained in step 1).

According to the steps in 6), the points projected into the invalid area in the i-th frame valid area map are deleted from the derived point cloud, and the points projected into the valid area in the i-th frame valid area map are retained ;

For each frame in the round shot image sequence, perform the above step 6) and delete it from the derived point cloud, and then retain the points in the effective area projected into the i-th frame effective area map, by deriving the points The cloud is projected around and deleted, and the 3D reconstruction obtains a derived point cloud containing internal points;

Traverse each derived point in the obtained derived point cloud, judge whether each derived point is an internal point or an external point, delete the derived point cloud that is an internal point, and keep the derived point cloud that is an external point; finally retain the point cloud is a valid point cloud derived once;

Count the number of effective point clouds obtained, if the density meets the requirements, then this effective point cloud is the final point cloud; if the density does not meet the requirements, then use the effective point cloud as the initial point cloud, repeat the above 3) to the current step, Until the obtained effective point cloud meets the density requirement;

In step 4), one of the points in the initial point cloud is used as the source point to derive new points from 26 directions of the cuboid, and the calculation formula of the new points is:

Among them, x_org, y_org, and z_org are the coordinates of a point in the initial point cloud on the x, y, and z axes, respectively, and x_scalar, y_scalar, and z_scalar are the calculated derived scales in the three directions of x, y, and z, respectively.

The 3*3*3 new point coordinates calculated by the above formula, except for the case where the source point coordinate increment is (0, 0, 0), will derive 26 new point clouds described in step 4);

In step 6), the calculation formula for transforming the derived point cloud into the camera coordinate system of the i-th frame image:

(x_cam _i ,y_cam _i ,z_cam _i )=(x_world,y_world,z_world)*R _i |t _i

Among them, (x_world, y_world, z_world) are the coordinates of the derived point cloud in the world coordinate system, R _i and t _i are the rotation matrix and translation vector of the i-th frame camera respectively, after the transformation of R _i and t _i , the world The point cloud in the coordinate system is transferred to the i-th frame camera coordinate system, that is, the coordinates of the transformed point cloud in the i-th frame camera coordinate system are (x_cam _i , y_cam _i , z_cam _i );

The point cloud in the camera coordinate system is back-projected, and each point is projected into the effective area map of the i-th frame. The calculation formula of the projection position is:

Among them, f is the focal length of the camera, C _x and _Cy are 0.5 times the resolution of the image respectively, and the calculated u and v are the projected positions of the point on the image, that is, the positions corresponding to the uth row and vth column on the image pixel location.

The construction of a voxel-based physical model specifically includes:

The first step is to use the registration software in the multimodal molecular imaging system to register the 3D voxel data reconstructed by magnetic resonance or computed tomography to the existing public digital mouse atlas, so as to draw and label protein molecules External contour lines and internal organizational boundaries;

In the second step, based on the 3D voxel data and the marked internal tissue boundary lines, a boundary node enrichment function is constructed:

Among them, j is a voxel node;

ψ _j (r) is the defined inner boundary node enrichment function;

v _j (r) is the linear interpolation basis function;

is a signed distance function, defined as the distance from a node to its nearest closed boundary:

Among them, sign(r) is used to indicate the affiliation relationship between point r and boundary Γ: if the point is inside the region, the value is negative, outside the region, it is positive, and on the boundary, it is zero;

is the value of the signed distance function on the voxel node j;

The third step is to use the marked internal tissue boundary as the interface to decompose the protein molecular body into a collection of multiple organs, assign the tissue optical characteristic parameters to the corresponding organs, and construct a voxel-based optical three-dimensional imaging physical model.

The construction of the adaptive optical transmission mathematical model specifically includes:

In the first step, according to the decomposed multiple organs and the corresponding tissue optical characteristic parameters, the organs are divided into three categories: high scattering, cavity and other tissues, and the classification basis is defined as:

Among them, Ω is the solution domain composed of protein molecules; Ω _hs is the high scattering tissue area; Ω _v is the cavity area; Ω _ls is the other tissue area; μ _'s is the tissue reduced scattering coefficient; ζ and χ are the classification thresholds , respectively taken as ζ=10 and χ=0.2mm ^-1 ;

In the second step, comprehensively considering the accuracy and computational complexity, the appropriate light transmission model is adaptively selected for different types of tissues to describe; among them, the diffusion approximation equation is used to describe the light transmission process in highly scattering tissues, and the free space The light transmission equation describes the transmission process of light in the cavity, and uses the third-order simplified spherical harmonic approximation equation to describe the transmission process of light in other tissues;

The third step is to construct an adaptive optical transmission mathematical model by constructing the boundary coupling conditions of physical quantities between different optical transmission models:

Among them, φ _i (r) (i=1, 2) is the node optical flux, S (r) is the energy density distribution of protein molecule optical probe, μ _a (r) and μ _aj (r) (j=1 ,2,3) are parameters related to protein molecular body absorption, D(r) is protein molecular body diffusion coefficient, β _i (i=1,2) and α are boundary mismatch factors between SP ₃ and DA equation, G(r′ ,r) is the transfer function describing the concept of radiative transfer theory, which is used to describe the transmission process of diffuse light from the cavity tissue, B is the interface between the scattering tissue and the cavity, and σ(r) is an indicator describing the position of the solution point factor, defined as:

Apply the following light transport equation coupling highly scattering and other scattering tissues:

Among them, φ ₀ (r) is the nodal optical flow for solving the diffusion approximation equation;

Apply the following light transport equation coupling the scattering tissue to the cavity:

Among them, q ₀ (r) is the Norman luminous flux formed on the interface between the cavity and the scattering tissue.

The establishment of system equations by the fusion enrichment function specifically includes:

Taking the constructed voxel-based physical model as the solution domain, numerically discretize and solve the constructed adaptive light transmission mathematical model by using the finite volume method of the internal boundary node enrichment function of the fusion structure, and establish a description of the target and volume of protein molecules in vivo. A system of equations that express a linear relationship between measurements:

J = AS;

Among them, A is the system matrix, which depends on the distribution of the three types of protein molecular bodies in the protein molecule body and the corresponding optical characteristic parameters; J is the outgoing optical flow rate collected on the surface of the protein molecule body; S is the target energy density distribution.

In the above embodiments, all or part of them may be implemented by software, hardware, firmware or any combination thereof. When implemented wholly or partly in the form of a computer program product, said computer program product comprises one or more computer instructions. When the computer program instructions are loaded or executed on the computer, the processes or functions according to the embodiments of the present invention will be generated in whole or in part. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website, computer, server or data center Transmission to another website site, computer, server or data center by wired (eg coaxial cable, fiber optic, digital subscriber line (DSL) or wireless (eg infrared, wireless, microwave, etc.)). The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center integrated with one or more available media. The available medium may be a magnetic medium (for example, a floppy disk, a hard disk, or a magnetic tape), an optical medium (for example, DVD), or a semiconductor medium (for example, a Solid State Disk (SSD)).

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Any simple modifications made to the above embodiments according to the technical essence of the present invention, equivalent changes and modifications, all belong to this invention. within the scope of the technical solution of the invention.

Claims (10)

1. An automatic molecular protein molecular body diagnostic system, characterized in that, the automatic molecular protein molecular physical diagnostic system comprises: The protein detection module is connected with the main control module, and is used to detect the protein by the protein detector; the protein detector draws the outer boundary contour line and The internal tissue edge line; based on the voxel data reconstructed by magnetic resonance or computed tomography and the marked internal tissue edge line, the enrichment function of the internal boundary node is constructed; considering the structural heterogeneity and optical specificity of protein molecular organization, a method based on The adaptive light transmission mathematical model of the mixed light transport equation describes the transmission process of light particles in the protein molecular body; in view of the application advantages of the finite volume method on the hexahedral voxel grid, the adaptive light transmission mathematical model is developed by using the extended finite volume method Carry out numerical discretization and solution, establish a system equation describing the linear relationship between in vivo targets and body surface measurements; consider the sparsity of in vivo target distribution and the incompleteness of body surface measurement data, establish a strategy based on sparse regularization and fusion prior The objective function of the preliminary target localization results; use an appropriate optimization method to solve the objective function, and carry out accurate and rapid reconstruction of the target target in the protein molecular body; The main control module is connected with the protein detection module, expert analysis module, data storage module and display module, and is used to control the normal operation of each module; The expert analysis module is connected with the main control module, and is used for online analysis of the detected data through the online expert comment network; The data storage module is connected with the main control module and is used to store the detected data information; The display module is connected with the main control module and is used for displaying detected data information. 2. The automatic molecular protein molecular body diagnostic system as claimed in claim 1, is characterized in that, described automatic molecular protein molecular physical body diagnostic system further comprises: The DNA detection module is connected with the main control module, and is used for detecting DNA by a DNA detector; The DNA sequencing module is connected to the main control module and is used to sequence the detected DNA; The cancer cell detection module is connected with the main control module and is used to detect cancer cells in the peripheral blood; The detection of protein by protein detector also includes: (1) After reconstruction, obtain a set of round-trip image sequences through photographic equipment, extract object contours for each frame of round-trip images, set the pixel values in the contour area to 255, and set the pixel values outside the contour to 0, and obtain A frame of binary image is called the effective area map; a point cloud with low density is obtained, which is called the initial point cloud, and the rotation matrix R and translation vector t of each frame of the camera relative to the world coordinate system are also obtained, and the rotation The matrix is combined with the translation vector to form the transformation matrix M; (2) Traverse each point in the initial point cloud, obtain the maximum and minimum values of all points in the initial point cloud on the three axes of x, y, and z, and calculate the maximum and minimum values on each axis The distance difference between them is recorded as x_dis, y_dis, and z_dis respectively, and the three distance differences are divided by 100, and the three quantities obtained are called the derived scale of the initial point cloud, and are recorded as x_scalar, y_scalar, and z_scalar; (3) Take a point in the initial point cloud as the source point, and expand the corresponding derived scales calculated in step (2) along the positive and negative directions of the three directions of x, y, and z respectively, and obtain a point centered on the source point The cuboid, the length, width and height of the cuboid are 2*x_scalar, 2*y_scalar, 2*z_scalar respectively, the center of the source point extends to 26 directions around the cuboid, and a new point is derived in each direction, taking The normal vector of this new point is the same as the normal vector of the source point, and each derived point records its source point; (4) all carry out step 4) described derivation operation to each point in the initial point cloud, will obtain a derived point cloud, the quantity of this point cloud midpoint is 26 times of initial point cloud quantity; (5) For the i-th frame image in the round shot image sequence, take out the transformation matrix M _i calculated in step (1), and transform the derived point cloud obtained in step (4 ₎ into the corresponding In the camera coordinate system of , and according to the projection principle, each point in the derived point cloud is back-projected onto the effective area map of the i-th frame obtained in the accurate and fast reconstruction step of the target target in the protein molecular body; According to the step in (5), for the points projected into the invalid area in the i-th frame effective area map, delete it from the derived point cloud, and the points projected into the valid area in the i-th frame effective area map are then reserve; For each frame in the round shot image sequence, perform the above step (4) and delete it from the derived point cloud, and keep the points in the valid area projected to the i-th frame valid area map, by deriving The point cloud is projected around and deleted, and the 3D reconstruction obtains a derived point cloud containing internal points; Traverse each derived point in the obtained derived point cloud, judge whether each derived point is an internal point or an external point, delete the derived point cloud that is an internal point, and keep the derived point cloud that is an external point; finally retain the point cloud is a valid point cloud derived once; Count the number of effective point clouds obtained. If the density meets the requirements, the effective point cloud is the final point cloud; if the density does not meet the requirements, the effective point cloud is used as the initial point cloud. Steps until the obtained effective point cloud meets the density requirement. 3. The automatic molecular protein molecular biology diagnosis system as claimed in claim 1, is characterized in that, in the step (3), with one point in the initial point cloud as source point derives and obtains new point to 26 directions of cuboid, The formula for calculating the new point is: Among them, x_org, y_org, and z_org are the coordinates of a point in the initial point cloud on the x, y, and z axes, respectively, and x_scalar, y_scalar, and z_scalar are the calculated derived scales in the three directions of x, y, and z, respectively. The 3*3*3 new point coordinates calculated by the above formula, except for the case where the source point coordinate increment is (0, 0, 0), will derive 26 new point clouds described in step (3); In step (5), the calculation formula for transforming the derived point cloud into the camera coordinate system of the i-th frame image: (x_cam _i ,y_cam _i ,z_cam _i )=(x_world,y_world,z_world)*R _i |t _i Among them, (x_world, y_world, z_world) are the coordinates of the derived point cloud in the world coordinate system, R _i and t _i are the rotation matrix and translation vector of the i-th frame camera respectively, after the transformation of R _i and t _i , the world The point cloud in the coordinate system is transferred to the i-th frame camera coordinate system, that is, the coordinates of the transformed point cloud in the i-th frame camera coordinate system are (x_cam _i , y_cam _i , z_cam _i ); The point cloud in the camera coordinate system is back-projected, and each point is projected into the effective area map of the i-th frame. The calculation formula of the projection position is: Among them, f is the focal length of the camera, C _x and _Cy are 0.5 times the resolution of the image respectively, and the calculated u and v are the projected positions of the point on the image, that is, the positions corresponding to the uth row and vth column on the image pixel location. 4. The automatic molecular protein molecular biology diagnostic system as claimed in claim 1, is characterized in that, described construction based on physical model of voxel specifically comprises: The first step is to use the registration software in the multimodal molecular imaging system to register the 3D voxel data reconstructed by magnetic resonance or computed tomography to the existing public digital mouse atlas, so as to draw and label protein molecules External contour lines and internal organizational boundaries; The second step is to construct a boundary node enrichment function based on the 3D voxel data and the marked internal tissue boundary lines: Among them, j is a voxel node; ψ _j (r) is the defined inner boundary node enrichment function; v _j (r) is the linear interpolation basis function; is a signed distance function, defined as the distance from a node to its nearest closed boundary: Among them, sign(r) is used to indicate the affiliation relationship between point r and boundary Γ: if the point is inside the region, the value is negative, outside the region, it is positive, and on the boundary, it is zero; is the value of the signed distance function on the voxel node j; The third step is to use the marked internal tissue boundary as the interface to decompose the protein molecular body into a collection of multiple organs, assign the tissue optical characteristic parameters to the corresponding organs, and construct a voxel-based optical three-dimensional imaging physical model. 5. The automatic molecular protein molecular physical diagnosis system as claimed in claim 1, wherein said building an adaptive light transmission mathematical model specifically comprises: In the first step, according to the decomposed multiple organs and the corresponding tissue optical characteristic parameters, the organs are divided into three categories: high scattering, cavity and other tissues, and the classification basis is defined as: Among them, Ω is the solution domain composed of protein molecules; Ω _hs is the high scattering tissue area; Ω _v is the cavity area; Ω _ls is the other tissue area; μ _'s is the tissue reduced scattering coefficient; ζ and χ are the classification thresholds , respectively taken as ζ=10 and χ=0.2mm ^-1 ; In the second step, comprehensively considering the accuracy and computational complexity, the appropriate light transmission model is adaptively selected for different types of tissues to describe; among them, the diffusion approximation equation is used to describe the light transmission process in highly scattering tissues, and the free space The light transmission equation describes the light transmission process in the cavity, and uses the third-order simplified spherical harmonic approximation equation to describe the light transmission process in other tissues; The third step is to construct an adaptive optical transmission mathematical model by constructing the boundary coupling conditions of physical quantities between different optical transmission models: Among them, φ _i (r) (i=1, 2) is the node optical flux, S(r) is the energy density distribution of the protein molecule optical probe, μ _a (r) and μ _aj (r) (j=1 ,2,3) are parameters related to protein molecular body absorption, D(r) is protein molecular body diffusion coefficient, β _i (i=1,2) and α are boundary mismatch factors between SP ₃ and DA equation, G(r′ ,r) is the transfer function describing the concept of radiative transfer theory, which is used to describe the transmission process of diffuse light from the cavity tissue, B is the interface between the scattering tissue and the cavity, and σ(r) is an indicator describing the position of the solution point factor, which is: Apply the following light transport equation coupling highly scattering and other scattering tissues: Among them, φ ₀ (r) is the nodal optical flow for solving the diffusion approximation equation; Apply the following light transport equation coupling the scattering tissue to the cavity: Among them, q ₀ (r) is the Norman luminous flux formed on the interface between the cavity and the scattering tissue. 6. The automatic molecular protein molecular physical diagnosis system as claimed in claim 1, wherein said fusion enrichment function establishes system equations specifically comprising: Taking the constructed voxel-based physical model as the solution domain, numerically discretize and solve the constructed adaptive light transmission mathematical model by using the finite volume method of the internal boundary node enrichment function of the fusion structure, and establish a description of the target and volume of protein molecules in vivo. A system of equations that express a linear relationship between measurements: J = AS; Among them, A is the system matrix, which depends on the distribution of the three types of protein molecular bodies in the protein molecule body and the corresponding optical characteristic parameters; J is the outgoing optical flow rate collected on the surface of the protein molecule body; S is the target energy density distribution. 7. A computer program for realizing the operation method of the automatic molecular protein molecular biology diagnosis system according to any one of claims 1-6. 8. An information data processing terminal equipped with the automatic molecular protein molecular biology diagnosis system according to any one of claims 1-6. 9. A computer-readable storage medium, comprising instructions, which, when run on a computer, cause the computer to execute the operating method of the automatic molecular protein molecular biology diagnosis system according to any one of claims 1-6. 10. An automatic molecular protein molecular body diagnostic equipment installed with the automatic molecular protein molecular body diagnostic system according to claim 1.