CN1216348C - Method for utilizing 3D visual anatomy atlas in cerebral surgical operation guidance system - Google Patents

Abstract

The present invention relates to a 3D visual applying method for an anatomy spectrum in cerebral surgical operation guidance systems. The present invention is divided into two parts of prior period of work and a later period executing method, wherein in the prior period work, an advanced non-linear interpolation method is used for carrying out delicate digitizing 3-D reconstruction work for two sets of cerebral anatomy spectra (SW and TT), and a 3-D non-linear registering method is used for firstly unifying the two sets of spectra into the same coordinate system. In the executing method, formatting treatment is carried out for input images, and the cooperation of a variable proportion grid method and a sectional type local linear registering method is adopted to realize the real-time registration of the spectra and patients' images; exact registration and visualization of the anatomy spectra can be realized through mutual fine adjustment. The present invention has the advantages visualization, accuracy and easy operation and also realizes the complete unification of the TT and the SW sets of spectra after complete 3-D reconstruction; the present invention provides great convenience for preoperative diagnoses and guidance during operations for doctors and increases an important function for operation guidance systems.

Description

Three-dimensional visualization application method of anatomical atlas in brain surgery navigation system

technical field

The invention relates to a three-dimensional visualization application method, in particular to a three-dimensional visualization application method of anatomical atlas in a brain surgery navigation system, belonging to the technical field of medical image processing and application.

Background technique

Brain surgery navigation system is a complex system engineering integrating computer technology, biomedical engineering technology and modern medical technology. Among them, the three-dimensional visualization of brain anatomical atlas is a key technology, which plays a decisive role in the localization of lesions. In traditional brain surgery, doctors can only complete complex operations by reading anatomical atlases and relying on experience. This will inevitably cause the problems of low success rate and long operation time. At present, there are mainly two types of brain anatomy atlases widely used in clinic: Talairach-Tournoux (TT) brain anatomy atlas and Schaltenbrand-Wahren (SW) brain anatomy atlas. The former is a complete anatomical map covering the entire human brain, while the latter is an internal brain map of the thalamus-basal ganglia. These atlases are two-dimensional images printed on paper. Although they are essential for stereotaxic brain surgery and have been used clinically for many years, human error and inconvenience in use have greatly inhibited their effectiveness.

After literature search, it was found that Hu Zeyong and others wrote an article "Development of a Stereotactic Micro-navigation Map System in the Basal Ganglia of the Thalamus" in the "Journal of Stereotaxic and Functional Neurosurgery", 2001, 14(3): 174-174. A digital attempt has been made to the stereotaxic atlas (SW) of the thalamic basal ganglia, and preliminary results have been obtained. However, this study is still in the stage of simple correlation coordinate registration of two-dimensional images; it is limited to SW atlas; other relevant parameters such as ventricle size and head size are not considered. Since the SW map is limited to some areas of the inner brain, it is difficult to take into account other relevant parameters such as ventricle size and head size when registering it with the actual patient image, which is a difficulty in this study. In addition, the reconstruction of 2D printed atlases to 3D stereoscopic images is also a difficult point, which is expected to be resolved.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies in the prior art, to provide a three-dimensional visualization application method of the anatomical atlas in the brain surgery navigation system, to make it more intuitive, more accurate, easier to operate, and to realize the complete integration of the two sets of atlases of TT and SW. Complete and unified after 3D reconstruction, and integrate other relevant parameters such as ventricle size and skull size other than AC-PC features into the registration algorithm to provide fine-tuning and other human-computer interaction functions to achieve the optimization of its 3D visualization.

The present invention is realized through the following technical solutions. The present invention is divided into two parts: the preliminary work and the later implementation method: firstly, in the preliminary work, an advanced non-linear interpolation method is used to carry out detailed analysis of two groups of brain anatomy atlases (SW, TT). Digital 3D reconstruction work; then use the 3D nonlinear registration method to unify the two sets of atlases into the same coordinate system first, and this part of the work does not need to be repeated once completed; secondly, in the later execution method, the input image is first formatted, Improve the registration accuracy, and then use the variable-scale grid method and the segmented local linear registration algorithm to achieve real-time registration of the atlas and patient images; finally, achieve accurate registration and visualization of the anatomical atlas through interactive fine-tuning.

Below the inventive method is further described, and specific content is as follows:

1. Preliminary work:

1) Using advanced non-linear interpolation method, carry out detailed digital 3D reconstruction of two groups of brain anatomy atlases (SW, TT). This method uses convolution based nonlinear interpolation, while the influence function kernel (Kernel) uses the basic spline function (Cardinal spline):

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; sn</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; \&infin;</span>}{\overset{<span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; b</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$

Here, n=3 is selected; (b ⁿ ) ^-1 is a B-spline filter (B-Spline filter).

2) Using a three-dimensional non-linear registration method, the two sets of atlases are first unified into the same coordinate system. This not only solves the problem of insufficient consideration of global parameters such as the overall ventricle size caused by the locality of the SW atlas, but also avoids repeated operations in the real-time registration process. After the registration operation of one atlas is completed, the registration work of another atlas is also completed at the same time, realizing the real-time switching function of the two visual atlas libraries. The registration algorithm uses a point-to-point registration method between anatomic feature points. Firstly find the corresponding 200 sets of anatomical feature points in the two atlases respectively, then calculate the registration three-dimensional transformation determinant B based on their coordinate relationship, and finally transform the SW atlas into the TT coordinate system according to the determinant:

SW·B→TT.

2. Post-execution method:

1) Formatting of the input image: Because the input image cannot be guaranteed to be scanned according to the coordinate system (Ac-Pc) of the brain anatomy atlas, this method firstly uses the two key feature points set by the doctor: Ac and Pc to input the image. Correction, that is, correcting the azimuth angle of the image from the three directions of X, Y, and Z to keep it consistent with the Ac-Pc coordinate system and improve the accuracy of registration.

2) Using the PGS-Proportional Grid System, the human brain and brain anatomy atlas are physically divided into 12 pieces, which are included in the corresponding grids, laying the foundation for local registration. The establishment of the variable-scale grid structure is based on the following eight anatomical feature points:

[1] The anterior origin of the brain (AC-The Anterior Commissure)

[2] The posterior origin of the brain (PC-The Posterior Commissure)

[3] The most Left points of the temporal cortex

[4] The most Right points of the temporal cortex

[5] The most anterior point of the frontal cortex

[6] The most posterior point of the occipital cortex

[7] The highest point of the occipital cortex

[8] The lowest point of the temporal cortex

3) Using the Piecewise Linear Registration algorithm, the interactive local registration is realized on the variable-scale grid structure, which greatly improves the regional registration accuracy. Segmented local linear registration is to transform all points (P) in the original grid to the target grid according to the proportional relationship between the corresponding original (S) and target (T) two variable-scale grids. Go to grid:

T＝T ₀ +(PS ₀ )*T _ext /S _ext

Here S ₀ and T ₀ represent the origin of the original and target grids; T _ext /S _ext represents the proportional relationship between the grids.

4) Realize accurate registration and visualization of anatomical atlas through interactive fine-tuning. The adjustable design is adopted, and the fine-tuning function is designed for 36 control points (Control Point) in the variable-scale grid structure, so that doctors can interactively fine-tune the registration results completed by the algorithm and improve the registration accuracy. A change in the coordinates of any one of the 36 control points will cause a change in the proportional relationship between the grids, that is, a change in T _ext /S _ext , thereby giving the result of linear registration T and realizing the function of fine-tuning.

5) The cooperation process of the variable-scale grid method and the segmented local linear registration algorithm is as follows: the process first reads the coordinates of the AC and PC points set by the doctor, and the other 6 feature points described in 2) Coordinates (collectively referred to as Landmark), create a variable-scale grid PGS with the corresponding coordinates of the brain atlas, and then use the segmented local linear alignment detailed in 3) for each sub-region in the PGS The quasi-algorithm performs registration one by one and updates the window in real time. PGS accepts the fine-tuning operation described in 4) at the same time, and refreshes the ground coordinates and PGS according to the changed control points, thereby refreshing the registration result.

The present invention has substantive features and significant progress. It overcomes the low accuracy and inconvenience of the existing methods, adopts the advanced non-linear interpolation method, and carries out meticulous digital three-dimensional reconstruction work on the two groups of brain anatomical atlases. A three-dimensional non-linear registration method was developed, and the two sets of atlases were first unified into the same coordinate system, avoiding repeated registration operations for different atlases. In the process of execution, the input image is formatted and rectified first, and then the real-time registration of the atlas and the patient image is realized with the cooperation of the variable-scale grid method and the segmented local linear registration algorithm, and finally the The doctor's interventional fine-tuning function realizes the precise registration and complete visualization of the anatomical atlas and the patient's image for the first time. This provides great convenience for doctors' preoperative diagnosis and intraoperative navigation, and adds an important function to the surgical navigation system.

Description of drawings

Figure 1 is a flow chart of the present invention's late stage execution method

Figure 2. The flow chart of the cooperation between the variable-scale grid method and the segmented local linear registration algorithm

Detailed ways

As shown in Figure 1 and Figure 2, Figure 2 is the cooperation process of the variable-scale grid method and the segmented local linear registration algorithm. This figure describes the division of the input image into 12 sub-regions based on AC-PC points , after the process of creating a variable-scale grid PGS, the process of implementing registration and fine-tuning using the method of segmented local linear registration is given.

The following embodiments are provided in conjunction with the content of the execution method of the present invention:

As shown in Figure 1, when the patient's 3D image instance is read into the system, the doctor implements the method in the following six steps:

(1) AC-PC setting: When the doctor selects the "AC-PC setting" function in the system menu, the system will give the initial AC-PC point in the patient image, and display it in the window with a color mark. The physician then uses the mouse to calibrate these initial point positions.

(2) Re-formatting of the input image: Based on the calibrated AC-PC coordinate points, the system will automatically format the input patient image under the instructions of the doctor so that it can be accurately established in the AC-PC coordinate system. The system also provides the ability to undo the reformatting so that the doctor can retry the operation.

(3) Create PGS: When the doctor uses the menu provided by the system to issue the command to create a PGS, the system will automatically create the initial PGS, and it will be displayed in the window in the form of a colored grid, and the 32 control points will start to accept mouse controls. Interactive intervention operation (see Figure 2). Doctors can fine-tune the positions of all control points according to the positions of PGS feature points to meet the requirements.

(4) Loading the brain anatomy atlas: using the menu provided by the system, the doctor can choose to load the corresponding brain anatomy atlas. Each type of anatomical atlas (TT, SW) system provides two atlases of high and low resolution for different needs. This example loads a low-resolution TT atlas, which uses different colors to represent different anatomical regions. At this time, the brain atlas has not yet been registered with the patient image.

(5) Registration: When the doctor uses the menu to issue a registration command to the system, the system will automatically call the segmented local linear registration (PWL) algorithm to perform the registration operation (see Figure 2), and display the registration results in real time. anatomical atlas. 0.5 seconds after the command was executed, the TT atlas and the patient image were registered, but errors still existed in some areas.

(6) Fine-tuning calibration: After the system completes the registration, the doctor can further fine-tune the control points in the PGS, and perform local registration operations on the anatomical area of special interest (such as the target area where a tumor occurs). The system will provide real-time registration refresh to meet the special diagnosis and operation planning needs of doctors. When the doctor moved (fine-tuned) the position of the control point in the unsatisfactory local area for 0.6 seconds, the TT map was only re-registered in this area, achieving high-precision results, and the registration result satisfied the doctor.

Claims (6)

1. A three-dimensional visualization application method of anatomical atlas in a brain surgery navigation system, which is characterized in that it is divided into two parts: preliminary work and post-execution method: firstly, in the preliminary work, a non-linear interpolation method is used to analyze two groups of brain anatomical atlases. SW and TT carry out digital 3D reconstruction work; then use 3D nonlinear registration method to unify the two sets of atlases into the same coordinate system first; secondly, in the execution method, first format the input image; then use variable The combination of the proportional grid method and the segmented local linear registration algorithm realizes the real-time registration of the atlas and patient images; finally, the precise registration and visualization of the anatomical atlas is realized through interactive fine-tuning, Using the non-linear interpolation method, the digital 3D reconstruction of the two groups of brain anatomical atlases SW and TT is performed as follows: Convolution-based nonlinear interpolation is used, while the influence function kernel uses basic splines: ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; sn</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; \&infin;</span>}{\overset{<span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; b</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$ Select n=3 here; (b ⁿ ) ^-1 is a B-type spline function filter, Accurate registration and visualization of anatomical maps through interactive fine-tuning are as follows: Adjustable design is adopted to fine-tune the 36 control points in the variable-scale grid structure, so that the doctor can fine-tune the registration results completed by the algorithm in an interactive way, and the coordinates of any one of the 36 control points will change. Trigger the proportional relationship between the grids, that is, the T _ext /S _ext changes, thereby changing the result of the linear registration T and realizing fine-tuning. 2. The three-dimensional visualization application method of the anatomical atlas in the brain surgery navigation system according to claim 1, wherein the three-dimensional non-linear registration method is used to first unify the two groups of atlases into the same coordinate system, specifically as follows: The registration method adopts the point-to-point registration method between anatomical feature points. Firstly, find out the corresponding 200 sets of anatomical feature points in the two atlases respectively, and then calculate the registration three-dimensional transformation determinant B based on their coordinate relationship. , and finally transform the SW spectrum to the TT coordinate system according to the determinant: SW·B→TT. 3. The three-dimensional visualization application method of the anatomical atlas in the brain surgery navigation system according to claim 1, wherein the format of the input image is specifically as follows: First, correct the input image according to the two key feature points set by the doctor: Ac and Pc, that is, correct the azimuth angle of the image from the three directions of X, Y, and Z to make it consistent with the Ac-Pc coordinate system , to improve the accuracy of registration. 4. The three-dimensional visualization application method of the anatomical atlas in the brain surgery navigation system according to claim 1, characterized in that the human brain and the brain anatomical atlas are physically divided into 12 pieces by adopting a variable-scale grid method, which are listed in The corresponding grid lays the foundation for local registration; the variable-scale grid structure is established based on the following eight anatomical feature points: [1] The front origin of the brain; [2] The rear origin of the brain; [3] The leftmost point of the temporal cortex; [4] The rightmost point of the temporal cortex; [5] The frontmost point of the forebrain cortex; [6] ] the last point of the occipital cortex; [7] the highest point of the occipital cortex; [8] the lowest point of the temporal cortex. 5. The three-dimensional visualization application method of anatomical atlas in the brain surgery navigation system according to claim 1, characterized in that, the segmented local linear registration algorithm is used to realize interactive Local registration, segmented local linear registration, is to transform all points P in the original grid to the target grid according to the proportional relationship between the corresponding original S and target T two variable-scale grids Go to: T＝T ₀ +(PS ₀ )*T _ext /S _ext Here S ₀ and T ₀ represent the origin of the original and target grids; T _ext /S _ext represents the proportional relationship between the grids. 6. The three-dimensional visualization application method of anatomical atlas in the brain surgery navigation system according to claim 1, characterized in that the cooperation process of the variable-scale grid method and the segmented local linear registration algorithm is as follows: First read the set coordinates of the front origin of the brain and the rear origin of the brain, as well as the leftmost point of the temporal cortex, the rightmost point of the temporal cortex, the frontmost point of the forebrain cortex, the last point of the occipital cortex, and the occipital cortex. The highest point of the cerebral cortex and the lowest point of the temporal cortex are 8 feature point coordinates. These 8 feature point coordinates are collectively referred to as ground coordinates. Cooperate with the corresponding coordinates of the brain atlas to create a variable-scale grid PGS. Each sub-area is registered one by one using the segmented local linear registration algorithm, and the window is updated in real time; the PGS accepts the fine-tuning operation in the interactive fine-tuning at the same time, and refreshes the ground coordinates and PGS according to the changed control points, thereby refreshing the registration. quasi-result.

Images