WO2016095167A2 - Traction deformation correction method based on surgical navigation system - Google Patents

Description

Stretch deformation correction method based on surgical navigation system Technical field

The invention belongs to the field of medical image processing and application, and relates to a method for correcting traction deformation based on a surgical navigation system, and particularly relates to a method for correcting deformation of brain tissue in a neurosurgical navigation system.

Background technique

Brain tissue traction deformation in clinical neurosurgery is an important factor affecting the accuracy of neurosurgical navigation systems. Intraoperative imaging and biomechanical models are often needed to correct the deformation of the brain tissue. Although intraoperative images such as intraoperative CT (iCT) and intraoperative MR (iMR) can accurately correct brain tissue deformation, the disadvantages are high cost, inconvenient operation, easy infection and prolonged infection. Surgery time, these shortcomings limit the clinical application of intraoperative image correction; biomechanical model correction method to overcome the intraoperative image by establishing a biomechanical model and then using the Finite Element Method (FEM) solution to correct brain tissue deformation The disadvantages of the method. In 2001, Miga applied the consolidation theory to establish a linear porous elastic biomechanical model, cutting the grid along the pulling path, generating twice the mesh nodes, and realizing the model update. The animal experiment showed that the model can correct an average of 66% of the brain. Tissue displacement error, reducing positioning error by 80%. In 2002, Ferrant used a linear elastic biomechanical model to model and update the model by removing all meshes on the pull path.

The finite element method (FEM) processing of brain tissue topology changes mainly includes: completely re-meshing or locally meshing the topology change regions. Regardless of the processing method used by FEM, it is necessary to change the original topology of the mesh and destroy the continuity of the original mesh. XFEM can overcome this shortcoming. It deals with the deformation of brain tissue without first considering the discontinuity caused by the topology change. Sexually, the brain tissue is directly meshed, and according to the specific shape of the crack, a new virtual degree of freedom is added to the mesh nodes passing along the crack. In this way, the crack generated by the brain tissue can accurately express the topological discontinuity generated by the crack without destroying the continuity of the original mesh. In 2009, Vigneron [3] proposed to apply XFEM to the linear elastic biomechanical model for the first time, and used the non-rigid registration technique based on iMR image to obtain the displacement of the brain tissue before and after the pulling, and used it as the boundary condition solving model. This method uses the Canny operator to perform edge detection on the images before and after registration. The verification results show that the Hausdorff distance of the image after registration is reduced. However, this method only obtained qualitative results, and did not give a quantitative result of the effectiveness of XFEM in correcting traction to cause tensile deformation of brain tissue.

The prior art related to the present invention is:

[1]M.I.Miga, D.W.Roberts, F.E.Kennedy, L.A.Platenik, A.Hartov, K.E.Lunn, and K.D.Paulsen, Modeling of retraction and resection for intraoperative updating of Images, Neurosurgery, vol. 49, (no.1), pp. 75-84; discussion 84-5, 2001-07-01 2001.

[2] M. Ferrant, A. Nabavi, B. Macq, FA Jolesz, R. Kikinis, and SKWarfield, Registration of 3-D intraoperative MR images of the brain using a finite-element biomechanical model, IEEE Trans Med Imaging, Vol.20, (no.12), pp.1384-97, 2001-12-012001.

[3]LMVigneron, MPDuflot, PARobe, SKWarfield, and JGVerly, 2D XFEM-based modeling of retraction and successive resections for preoperative image update, Comput Aided Surg, vol. 14, (no. 1-3), Pp.1-20, 2009-01-20 2009.

Summary of the invention

The object of the present invention is to provide a surgical navigation system accuracy correction method which is simple in implementation, flexible in operation, and does not need to change an existing navigation device for clinical application, and relates to a traction deformation correction method based on a surgical navigation system, and particularly relates to a method for correcting traction deformation based on a surgical navigation system, A method for correcting the deformation of brain tissue in a neurosurgical navigation system.

The method of the invention comprises an MRI-based three-dimensional automatic segmentation algorithm, obtains a target tissue such as a brain tissue and performs meshing, and establishes a physical model by assigning corresponding biomechanical properties to each mesh unit. The surface of the stretched tissue was indirectly tracked by the tracking algorithm, and it was used as a boundary condition and combined with the physical model to calculate the eXtended Finite Element Method (XFEM). The deformation of the entire brain tissue was obtained. Finally, the interpolation algorithm was used to update the algorithm. The front three-dimensional data field is used to guide the surgery.

Specifically, the present invention adopts a physical model based on the linear elasticity theory, and in order to conveniently and efficiently obtain boundary conditions, a three-dimensional laser imaging device LRS (Laser Range Scanner) is used to acquire a bare brain plate in real time, and is tracked by a tracking algorithm. The movement of the brain tissue to obtain the boundary conditions for driving the model; combining the boundary conditions with the linear elastic physical model and solving with XFEM can effectively correct the deformation of the brain tissue during the operation.

The method of the invention has simple implementation and flexible operation, and does not need to change the existing navigation device, and is beneficial to clinical application.

More specifically, the present invention relates to a traction deformation correction method based on a surgical navigation system, comprising the following steps:

(1) Using MRI-based 3D automatic segmentation algorithm to obtain the target organization;

(2) meshing the obtained target organization by using a multi-resolution grid algorithm;

(3) assigning corresponding biomechanical properties to each grid unit to establish a physical model of the target organization;

(4) Using a three-dimensional laser scanning device, obtaining a boundary condition by being subjected to a pulling tissue tracking algorithm;

(5) Perform XFEM calculation in combination with the boundary conditions and the physical model to obtain deformation of the target tissue at any position;

(6) The pre-operative three-dimensional data field is updated with the obtained deformation using the interpolation algorithm.

In the present invention, the target tissue is selected from brain tissue.

The first step in solving the deformation of brain tissue is to segment the target region such as brain tissue. The present invention adopts an automatic segmentation algorithm based on the principle of gray histogram combined with morphological features. In the step (1) of the method of the present invention, the three-dimensional automatic segmentation algorithm is adopted, the Gaussian curve is fitted to the histogram curve, and the threshold value is automatically determined according to the fitted Gaussian curve, and the thresholding determination formula is:

t _s =t _BF +4/5(u _GM -t _BF )

Where t _BF is the threshold between the background and the foreground, and u _GM is the gray level mean of the target tissue white matter, corresponding to the first and second peaks of the Gaussian curve, respectively;

After the thresholding, the morphological corrosion algorithm can be used to remove the existing fine connections; the brain tissue is separated from the skin and bone by setting the maximum connected domain by setting the seed point, and finally adopting the morphology for the segmented brain tissue. The expansion algorithm in the study restores the brain tissue lost in the corrosion operation.

In the present invention, the divided brain tissue is discretized by a tetrahedral unit. The linear elastic model is characterized in that each unit is a constant strain unit, and the method of the present invention adopts a mesh division algorithm with multiple variability rates for the phenomenon that the relative internal tissue changes greatly at the brain tissue boundary and the ventricle boundary. The unit density is large, and the internal unit density is small, and the surface can be accurately expressed, and the deformation at the boundary is better simulated. For the area with relatively small internal variation, the unit is divided larger, which reduces the calculation amount. In the step (2) of the method of the present invention, after the three-dimensional image space is divided into the mesh of the hexahedral element by the octree-like algorithm, each hexahedral element is divided into five tetrahedral elements, and the obtained tetrahedral mesh is obtained. Refinement at the boundary, multi-variable variability grid, using the MT-like algorithm to cut the grid, remove the background, and obtain a multi-variation grid that ultimately contains only brain tissue.

In the present invention, for the brain tissue after meshing, a physical model of the brain tissue is established by assigning corresponding biomechanical properties to each unit. Since the deformation process of the brain tissue is slow and the deformation is small, the strain and the stress are linear, and the present invention It is modeled as an elastic body based on linear elasticity theory, expressed as:

Where u is the displacement vector, F is the force vector, μ=E/2(1+v), v is the Poisson's ratio, and E is the elastic modulus.

In the present invention, after the physical model is completed, the model is driven by appropriate boundary conditions; in view of the most easily observed intracerebral pressure plate for pulling the brain tissue, since the end of the brain pressure and the pulled brain tissue are closely connected, the Obtaining the displacement of the brain tissue before and after the pulling of the brain pressure plate is used as the boundary condition of the driving model in the present invention. The method only needs to use the original probe, and the probe is used after inserting the brain pressure plate. The probe points the physical points on the brain pressure plate, and the three-dimensional imaging device LRS is introduced. The LRS is scanned after being pulled. The two devices can be removed after use, and the operation is simple, and the existing navigation device does not need to be changed and the intraoperative space is not occupied. The starting position of the brain plate tension is obtained by taking the marker points on the brain plate by the probe point. The results show that the LRS can provide rich geometric and texture information for tracking the movement of the brain plate. The initial position of the cerebral pressure plate and the LRS scan to obtain the brain pressure plate pulling point cloud can be unified into the image space by the rigid body registration algorithm;

In the present invention, the above-mentioned rigid body registration algorithm is implemented by coordinate transformation: three points for determining the initial position of the brain pressure plate in the image space by the probe point are realized by the pre-point registration (PBR) algorithm to realize the reference space to The transformation of the image space, the LRS scan to obtain the intraoperative brain pressure plate pulling point cloud first by means of the calibrator, to realize the transformation of the LRS space to the tracked space, and then transform the space of the tracker to the space locator by means of the spatial locator And the space locator space to the reference frame space transformation, and finally realize the transformation of the reference frame space to the image space by means of the pre-point registration (PBR) algorithm. After the above series of transformations, the reference frame space can be finally realized to the image space. And the transformation of the LRS space to the image space, thereby obtaining the initial position of the brain platen in the image space and the position of the brain platen in the image space after the pulling. In one embodiment of the present invention, the traction brain tissue tracking algorithm is adopted. First, after the brain plate is inserted into the brain tissue, the three physical point coordinates of the brain pressure plate are obtained by using the probe, and the brain pressure plate is obtained by three-point coordinate transformation. At the starting position of the image space, the brain tissue is then pulled by the brain pressure plate, and after the completion of the pulling, the point cloud data of the shape of the end of the brain tissue plate including the exposed portion of the brain tissue is obtained by scanning with a three-dimensional laser scanning device. Since the brain pressure plate has a certain thickness, in the image space, the end surface of the brain pressure plate is translated along the pulling direction by the distance of the thickness of the brain plate, and finally the non-rigid registration of the plane of the starting position of the brain plate and the plane after the brain plate is pulled is obtained. The displacement caused by the pulling of the cerebral platen is closely followed by the pulled brain tissue and the brain platen, so that the displacement of the pulled brain tissue is tracked;

In the present invention, the improved point-based registration method based on Coherent Point Drift (CPD: Coherent Point Drift) is used to perform non-rigid registration of the plane of the starting position of the brain plate and the plane after the brain plate is pulled, wherein the dense points are The set uses the Gaussian mixture model to represent that the sparse point set is regarded as the set of sampling points of the dense point set, thus transforming the point set registration problem into the mixed density estimation problem of estimating the probability distribution of the dense point set according to the probability distribution of the sparse point set. The objective function is:

Where the Gaussian distribution function p is:

m, n are the number of points of the sparse point set and the dense point set respectively, x _n is a point in the dense point set, y _m is a point in the sparse point set, and θ and σ ² are parameters of the Gaussian distribution, w _i And w _k user-defined parameters; bringing the Gaussian distribution function to the objective function yields:

The θ and σ ² of the minimum time of the objective function are obtained through continuous iteration, and finally θ is brought into the transformation matrix T(y _m , θ), and each point corresponding to the dense point set in the sparse point set is obtained, and the matching is obtained. Quasi-result results; then according to the registration results, the displacement difference of the end point set before and after the brain plate is pulled, that is, the pulling displacement of the pulled brain tissue, and the displacement difference is used as the boundary condition for driving the traction deformation of the brain tissue before surgery. Establish a biomechanical model.

In the present invention, according to the obtained cerebral cortex displacement, combined with the physical model, XFEM is used to calculate the displacement at any node, and then the shape function can be used to obtain the deformation of the brain tissue at any position; in the method (5) of the method, the brain is used. The cortical motion is the boundary condition. The model equation in the form of partial differential equations is solved by XFEM method. The deformation of all the unit nodes is obtained, and the deformation of the brain tissue at any position is calculated by combining the shape function.

In step (6) of the method of the present invention, the pre-operative three-dimensional data field is updated by using the interpolation algorithm: starting from the deformed grid unit, finding the display coordinate point (integer coordinate point) in the unit, and obtaining the point by using the shape function. The position before the deformation is obtained by trilinear interpolation to obtain the gray value of the point. After the above-mentioned processing is performed on the deformed unit, the original three-dimensional data field is updated by the deformed three-dimensional data field.

The invention introduces a three-dimensional laser scanning device to obtain boundary conditions, and predicts the deformation of the whole brain tissue by using a linear elastic physical model, which not only ensures the prediction accuracy of the model, but also solves the problem that the boundary condition is difficult to measure. The problem can be implemented clinically, greatly improving the accuracy of the surgical navigation system.

DRAWINGS

Figure 1 is a flow chart for solving the deformation of brain tissue.

Figure 2 shows the results of a three-dimensional automatic segmentation algorithm based on a 256 × 256 × 48 MRI data field. The three images are the cross-sectional, sagittal and coronal segmentation results from left to right.

Figure 3 is the result of multi-resolution meshing for 256 × 256 × 48 MRI data fields, in which 18,485 tetrahedral elements and 5410 nodes are divided; the left picture shows the meshing result of the fault, and the right picture shows the three-dimensional image. The result of the meshing of the data field.

Figure 4 is a rigid body registration algorithm implemented using coordinate transformation.

Figure 5 is the boundary condition, where 1 is the brain tissue to be pulled, 2 is the result of LRS scan, 3 is the point cloud after the end of the brain plate after stretching, 4 is the initial plane of the brain plate, and is subjected to non-rigid registration. A registration result of 4 is obtained, and finally processed to obtain a boundary condition of 5.

Fig. 6 is a result of three-dimensional deformation, in which the left image is the brain tissue for deformation, and the right image is the brain tissue after the traction deformation due to the tension of the brain plate.

Figure 7 shows the LRS, which contains a color camera that stacks color photos and data point clouds. The effective measurement range is 200mm to 750mm. The measurement accuracy of the instrument at 450 mm from the object to be measured is ±125 um.

Fig. 8 is a scan result of the cortex, the left view is a top view, and the right view is a right view.

detailed description

Example 1

A traction deformation correction method based on a surgical navigation system has the following steps:

1. For the 256×256×48 3D MRI data field, the 3D automatic segmentation algorithm is used to obtain the brain tissue. The threshold values correspond to the first and second peaks of the Gaussian curve respectively. The corrosion element has a radius of 5 pixels. a spherical element, the expanded element adopts a spherical element having a radius of 6 pixels;

2. Using multi-resolution gridding algorithm, the segmented brain tissue is discretized into 18485 tetrahedrons, the number of nodes is 5410, and the largest tetrahedron at the boundary is 7.5×7.5×7.5mm ³ (with four-sided external hexahedron) Size measurement), the largest internal tetrahedron is 15 × 15 × 15mm ³ ;

3. Set the biomechanical attribute parameters of brain tissue for each unit, Young's modulus = 3Kpa, Poisson's ratio = 0.45;

4. Using coordinate transformation to achieve rigid body registration, obtain the initial position of the cortex in the LRS space, The tracker is fixed on the LRS, and the spatial locator tracks the LRS through the tracker to realize the arbitrary movement of the LRS within the monitoring range of the spatial locator;

5. Perform neurosurgery, insert the brain plate into the target tissue, use the probe to take the three physical markers on the brain plate, and then pull the brain plate, keep the state with LRS for a scan, and the LRS is pulled from the brain plate. The distance above is controlled at about 250mm, the scanning time is 5-7 seconds, and the resolution is 512 (horizontal) × 512 (vertical). According to the physical point coordinates of the initial positions of the three brain pressure plates provided by the probe, the deformed LRS provides The three-dimensional surface information and the three-dimensional surface information obtained by the step 4 are combined with the non-rigid surface registration technology to indirectly track the motion of the pulled brain tissue by tracking the motion of the brain pressure plate, and finally obtain the displacement of the tensioned mesh node;

6. Use XFEM to transform the physical model into a matrix form: Au=b (A is the total rigid matrix, u is the displacement vector to be solved, b is the nodal force vector), and the solution is 5410 (number of nodes) using the conjugate gradient method. ×3 (dimension)=16230 equations of linear equations, introducing the boundary conditions obtained by step 5, eliminating the singularity of the matrix A, finding the displacement vector u, obtaining the displacement at any node inside the target tissue, and combining The shape function in XFEM can be interpolated to deform at any position inside the target tissue;

7. From the deformed data field, calculate the position before the deformation of all coordinate points (integer coordinate points) contributing to the display according to the shape function, calculate the gray value of the point by trilinear interpolation, and finally for all these integers The three-dimensional data field consisting of coordinate points is visualized using the classical Raycasting algorithm to guide the surgery.

Example 2

Another method of retracting deformation based on a surgical navigation system has the following steps:

1. For the 256×256×128 3D MRI data field, the 3D automatic segmentation algorithm is used to obtain the brain tissue. The threshold values correspond to the first and second peaks of the Gaussian curve respectively. The corrosion element has a radius of 5 pixels. a spherical element, the expanded element adopts a spherical element having a radius of 6 pixels;

2. Using multi-resolution gridding algorithm, the segmented brain tissue is discretized into 38273 tetrahedrons, the number of nodes is 11220, and the largest tetrahedron at the boundary is 8.0×8.0×8.0mm3 (using four-sided external hexahedron size) Measured), the largest internal tetrahedron is 15 × 15 × 15mm3;

3. Set the biomechanical attribute parameters of brain tissue for each unit, Young's modulus = 3Kpa, Poisson's ratio = 0.4;

The coordinate transformation is used to realize the rigid body registration, and the initial position of the cortex in the LRS space is obtained. The tracker is fixed on the LRS, and the spatial locator tracks the LRS through the tracker to realize the arbitrary movement of the LRS within the monitoring range of the spatial locator;

5. Perform neurosurgery, insert the brain plate into the target tissue, use the probe to take the three physical markers on the brain plate, and then pull the brain plate, keep the state with LRS for a scan, and the LRS is pulled from the brain plate. The distance above is controlled at about 250mm, the scanning time is 5-7 seconds, and the resolution is 512 (horizontal) × 512 (vertical). According to the physical point coordinates of the initial positions of the three brain pressure plates provided by the probe, the deformed LRS provides The three-dimensional surface information and the three-dimensional surface information obtained by the step 4 are combined with the non-rigid surface registration technology to indirectly track the motion of the pulled brain tissue by tracking the motion of the brain pressure plate, and finally obtain the displacement of the tensioned mesh node;

6. Using XFEM to transform the physical model into a matrix form: Au=b (A is the total rigid matrix, u is the displacement vector to be solved, b is the nodal force vector), and the solution is 11220 (number of nodes) using the conjugate gradient method. 3 (dimension) = 33660 equations of linear equations, introduce the boundary conditions obtained by step 5, eliminate the singularity of matrix A, find the displacement vector u, obtain the displacement at any node inside the target tissue, and then combine XFEM The shape function in the interpolation can be interpolated to deform the arbitrary position inside the target tissue;

7. From the deformed data field, calculate the position before the deformation of all coordinate points (integer coordinate points) contributing to the display according to the shape function, calculate the gray value of the point by trilinear interpolation, and finally for all these integers The three-dimensional data field consisting of coordinate points is visualized using the classical Raycasting algorithm to guide the surgery.

Claims (8)

A traction deformation correction method based on a surgical navigation system, characterized in that it comprises the steps of: (1) Using MRI-based 3D automatic segmentation algorithm to obtain the target organization; (2) meshing the above target organization by using a multi-resolution grid algorithm; (3) For the meshed target organization, assign corresponding biomechanical properties to each grid unit, and establish a physical model of the target organization; (4) Using a three-dimensional laser scanning device to track the motion of the target tissue being pulled by a tracking algorithm to obtain boundary conditions; (5) Combine the boundary conditions with the physical model to perform XFEM calculation, obtain the deformation of all the unit node positions, and then calculate the deformation of the target tissue at any position by combining the shape function; (6) Using the interpolation algorithm to update the pre-operative three-dimensional data field with the obtained deformation, and visualizing the three-dimensional data field using the Raycasting algorithm; Where, in step (1), The three-dimensional automatic segmentation algorithm adopts a Gaussian curve to fit a histogram curve, and automatically determines a threshold value according to a fitted Gaussian curve, and the thresholding determination formula is t _s =t _BF +4/5(u _GM- t _BF ) Where t _BF is the threshold between the background and the foreground, u _GM is the mean value of the gray matter of the target tissue, and the threshold values respectively correspond to the first and second peaks of the fitted Gaussian curve; After the thresholding, the morphological corrosion algorithm is used to remove the existing fine connections; the maximum connected domain is sought by setting the seed point, and finally the morphological expansion algorithm is used for the segmented brain tissue to restore the corrosion operation. Loss of brain tissue. In the step (2), the multi-resolution grid algorithm divides the three-dimensional image space into a grid of hexahedral elements by using an octree-like algorithm, and divides each hexahedral unit into five tetrahedral units. The obtained tetrahedral mesh boundary is refined, the multi-variable variability mesh, the MT-like algorithm is used to cut the mesh, the background is removed, and the multi-variation tempo mesh finally containing only the target tissue is obtained. The etched element uses a spherical element with a radius of 5 pixels, and the swelled element uses a spherical element with a radius of 6 pixels. The method of claim 1 wherein said target tissue is selected from the group consisting of brain tissue. The method according to claim 1 or 2, wherein in the step (3), the mathematical expression of the physical model is

Where u is the displacement vector and F is the force vector. μ=E/2(1+v), v is the Poisson's ratio, and E is the elastic modulus. The method according to claim 3, wherein after the physical model is completed, the model is driven by a suitable boundary condition, and the displacement of the pulled brain tissue is calculated by obtaining the position of the brain pressure plate before and after the pulling, and the displacement is taken as Driving the boundary condition of the model, in step (4), the traction brain tissue tracking algorithm is to obtain the three physical point coordinates of the brain pressure plate by using the probe, and obtain the starting position of the brain pressure plate in the image space by three-point coordinate transformation. ; Then, the brain tissue is pulled by the brain pressure plate, and the pulling state is maintained after the pulling is completed, and the target tissue of the stretched state is scanned by the three-dimensional laser scanning device to obtain the point cloud data of the exposed area of the tissue and the end shape of the brain pressure plate, and finally utilized. Based on the point-point registration method of consistent point drift, the plane of the starting position of the brain plate and the plane after the brain plate are pulled are non-rigidly registered, and the displacement caused by the pulling of the brain plate is obtained, and the displacement of the target tissue is tracked. , Among them, the denser point set is represented by the Gaussian mixture model, and the sparse point set is used as the sampling point set of the dense point set, and the point set registration problem is transformed into the mixture of the probability distribution of the dense point set according to the probability distribution of the sparse point set. The density estimation problem, its objective function is:

Where the Gaussian distribution function p is:

m, n are the number of points of the sparse point set and the dense point set respectively, x _n is a point in the dense point set, y _m is a point in the sparse point set, and θ and σ ² are parameters of the Gaussian distribution, w _i And w _k user-defined parameters; bringing the Gaussian distribution function to the objective function yields:

By iteratively obtaining the minimum θ and σ ^{2 of} the objective function, and finally bringing θ into the transformation matrix T(y _m , θ), each point in the sparse point set corresponds to the point in the dense point set, and the registration result is obtained. . The method according to claim 4, wherein the displacement difference of the end point set before and after the tension of the brain plate is obtained according to the registration result, and the physical model is used to determine the partial differential equation by using the XFEM method as the boundary condition of the cortical layer motion. The model equation of the form obtains the deformation of the position of all the element nodes, and then combines the shape function to calculate the deformation of the target tissue at any position. The method according to claim 1, wherein in the step (6), the interpolation algorithm is: starting from the deformed grid unit, determining a display coordinate point in the unit, and obtaining the point by using a shape function. At the position before the deformation, the gray value of the point is obtained by using the trilinear interpolation, and after the above processing is performed on the deformed unit, the original three-dimensional data field is updated by the deformed three-dimensional data field. The method according to any one of claims 1 to 6, wherein the specific steps are: (1). For the 256×256×48 3D MRI data field, the 3D automatic segmentation algorithm is used to obtain the target brain tissue. The threshold values correspond to the first and second peaks of the Gaussian curve respectively. The radius of the corrosion element is 5 a spherical element of pixels, the expansion element adopting a spherical element with a radius of 6 pixels; (2). Using the multi-resolution meshing algorithm, the segmented brain tissue is discretized into 18485 tetrahedrons, the number of nodes is 5410, and the maximum tetrahedron at the boundary is 7.5×7.5×7.5mm ³ (with tetrahedron) The size of the external hexahedron is measured), and the largest internal tetrahedron is 15×15×15mm ³ ; (3). Set the biomechanical attribute parameters of brain tissue for each unit, Young's modulus = 3Kpa, Poisson's ratio = 0.45; (4). Using coordinate transformation to achieve rigid body registration, obtain the initial position of the cortex in the LRS space, the tracker is fixed on the 3D laser scanning device, and the spatial locator tracks the 3D laser scanning device through the tracker to realize the 3D laser scanning device. The space locator can move anywhere to obtain three-dimensional surface information; after inserting the brain pressure plate into the target brain tissue, use the probe to take three physical marker points on the brain pressure plate, and then pull the brain pressure plate to maintain the state by using three-dimensional laser scanning. The device performs a scan. The distance between the 3D laser scanning device and the brain plate is directly above 250mm, the scanning time is 5-7 seconds, the resolution is 512 (horizontal) × 512 (vertical), according to the three brain plate provided by the probe. The physical point coordinates of the initial position, the deformed three-dimensional surface information provided by the three-dimensional laser scanning device and the three-dimensional surface information before deformation, combined with the non-rigid surface registration technology, indirectly track the movement of the pulled brain tissue by tracking the movement of the brain pressure plate, and finally Obtaining the displacement of the pulled mesh node; (5). Use XFEM to transform the physical model into a matrix form: Au=b (A is the total rigid matrix, u is the displacement vector to be solved, b is the nodal force vector), and the solution is 5410 (the node is solved by the conjugate gradient method). Number) × 3 (dimension) = 16230 equations of linear equations, introducing the boundary conditions obtained by step 5, eliminating the singularity of the matrix A, finding the displacement vector u, and obtaining the displacement at any node inside the target tissue. Combined with the shape function in XFEM It can interpolate the deformation of any position inside the target tissue; (6). From the deformed data field, calculate the position before the deformation of all the integer coordinate points contributing to the display according to the shape function, calculate the gray value of the point by using the trilinear interpolation, and finally all the integer coordinate points. The composed 3D data field is visualized using the Raycasting algorithm. The method according to claim 7, wherein in step (2), the segmented brain tissue is discretized into 38273 tetrahedrons, the number of nodes is 11220, and the maximum tetrahedron at the boundary is 8.0×8.0×8.0. Mm3 (measured by four-sided external hexahedron size), the largest internal tetrahedron is 15 × 15 × 15mm3; In step (3), the biomechanical attribute parameters of the brain tissue are set for each unit, the Young's modulus is 3 Kpa, and the Poisson's ratio is 0.4; In step (5), the physical model is transformed into a matrix form by XFEM: Au=b (A is the total rigid matrix, u is the displacement vector to be determined, b is the nodal force vector), and the symmetry gradient method is used to solve the problem with 11220 ( Number of nodes) × 3 (dimension) = 33660 equations of linear equations, introduce the boundary conditions obtained in step 5, eliminate the singularity of matrix A, find the displacement vector u, and obtain the displacement at any node inside the target tissue. In combination with the shape function in XFEM, you can interpolate the deformation anywhere in the target tissue.

Images