A method for ossified ligamentum flavum segmentation and a method for resection path planning

Abstract

The present invention discloses a method for segmenting ossified ligamentum flavum and a method for planning resection path. Among them, the method for segmenting ossified ligamentum flavum includes: obtaining three-dimensional images of the target area of the patient, and segmenting to obtain the first bone area and the second bone area therein; setting a Hu value threshold according to the voxel characteristics of the ossified ligamentum flavum, and searching and extracting voxel points with Hu values greater than the Hu value threshold located between the first bone area and the second bone area to obtain the ossified ligamentum flavum area. The present invention can accurately identify and extract the ossified ligamentum flavum from many tissue areas of the human body, effectively ensuring the accuracy and success rate of subsequent resection of the ossified ligamentum flavum. The present invention can plan a safer and more effective resection path for the ossified ligamentum flavum, provide surgical guidance for doctors, improve the success rate of ossified ligamentum flavum resection, and thus reduce the surgical risk.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for segmenting ossified ligamentum flavum and a method for planning resection path. Background Art

[0002] The ligamentum flavum is an organizational structure in the spinal canal that maintains the stability of the spine. The ligamentum flavum is elastic and is the main ligament structure participating in the formation of the posterior wall of the spinal canal. Ossification of the ligamentum flavum refers to a disease in which the ligamentum flavum ossifies and abnormally proliferates, resulting in a reduction in the volume of the spinal canal. The hypertrophic ligamentum flavum compresses the surrounding muscles, fascia and other soft tissues, causing the patient to have low back pain. The stenosis in the spinal canal can also cause radiating pain and numbness in the lower extremities, bringing great pain to the patient.

[0003] Surgical resection of the ossified ligamentum flavum is the only effective treatment method. Currently, clinically, the patient's images are collected by CT before surgery, and the patient's body is observed by inserting an endoscope during surgery. The imaging quality of the endoscope itself is generally average, and the ligamentum flavum is usually adhered to the surrounding dura mater and is not easy to segment. Correctly identifying and removing the ossified ligamentum flavum in the endoscope depends heavily on the doctor's clinical experience, which also affects the success rate of the resection of the ossified ligamentum flavum. Therefore, how to accurately identify and extract the ossified ligamentum flavum from many tissue regions of the human body and help plan a more effective surgical path is of great significance for guiding more doctors in surgical planning and improving the success rate of the resection of the ossified ligamentum flavum in the future. Summary of the Invention

[0004] Object of the Invention: Aiming at the above deficiencies, the present invention provides a method for segmenting ossified ligamentum flavum and a method for planning resection path, which can accurately identify and extract the ossified ligamentum flavum from many tissue regions of the human body, can provide surgical guidance for doctors, and thus reduce the surgical risk.

[0005] Technical Solution:

[0006] The present invention provides a method for segmenting ossified ligamentum flavum, including:

[0007] Obtain the three-dimensional image of the target area of the patient, and segment the first bone area and the second bone area therein;

[0008] Set the Hu value threshold according to the voxel characteristics of the ossified ligamentum flavum, and search and extract the voxel points with Hu value greater than the Hu value threshold between the first bone area and the second bone area to obtain the ossified ligamentum flavum area.

[0009] Specifically, the searching and extracting the voxel points with Hu value greater than the Hu value threshold between the first bone area and the second bone area is specifically:

[0010] Traverse the voxel points in the first bone region, establish a three-dimensional search space of a set size centered on any one of the voxel points, extract the voxel points in the three-dimensional search space whose Hu value is greater than the set Hu value threshold and are not in the first bone region and the second bone region, and put them into the seed point set. After traversing all the voxel points in the first bone region, the obtained seed point set is the ossified ligamentum flavum point set.

[0011] More specifically, traverse the seed point set. If the voxel value of a certain neighborhood point of the 8-neighborhood points of any one voxel point is greater than the Hu value threshold and is not in the first bone region and the second bone region, then put this neighborhood point into the seed point set. After traversing all the voxel points in the seed point set, the obtained target point set is the ossified ligamentum flavum point set finally obtained.

[0012] More specifically, the three-dimensional search space of the set size is specifically:

[0013] In the human anatomical coordinate system, the height of the three-dimensional search space is one-third of the distance between the lowest point of the vertebral body in the second bone region and the highest point of the first bone region obtained by calculation, its width is one-fifth of the width of the first bone region, and its length is one-half of the depth of the vertebral body in the second bone region.

[0014] Specifically, the method for obtaining the ossified ligamentum flavum region is as follows:

[0015] Calculate the boundary of the minimum circumscribed rectangle of the seed point set composed of the extracted points, and determine the ossified ligamentum flavum candidate region accordingly;

[0016] Based on the Sato classification method, classify the ossified ligamentum flavum candidate region into the original classification template most similar to it, and exclude the vertebral segment region in the ossified ligamentum flavum candidate region, and finally obtain the ossified ligamentum flavum region.

[0017] More specifically, the method of classifying the ossified ligamentum flavum candidate region into the original classification template most similar to it based on the Sato classification method is as follows:

[0018] (1) For any original classification template image, perform scale and rotation transformations on it to obtain multiple candidate classification template images;

[0019] a. Perform scale transformation on each original classification template image to generate a series of scaled original classification template images;

[0020] b. Perform rotation transformation on each original classification template image to generate a series of rotated original classification template images.

[0021] (2) For each candidate classification template image obtained through scale and rotation transformation, its matching degree with the candidate region of the ossified ligamentum flavum is detected, the maximum matching coefficient is calculated, and the candidate classification template is retained, that is, the candidate region of the ossified ligamentum flavum is classified into the corresponding candidate classification template.

[0022] Specifically, the first bone region and the second bone region in the patient target region obtained by segmentation are as follows:

[0023] Define the spinous process, vertebral arch plate, and transverse process regions below the vertebral body in the vertebral segment as the first bone region, and the vertebral body region as the second bone region;

[0024] Use a pre-trained segmentation model to segment the three-dimensional image of the patient target region to obtain the first bone region and the second bone region therein;

[0025] The segmentation model is obtained by training a number of three-dimensional images with the first bone region and the second bone region pre-marked as training samples.

[0026] The present invention also provides a method for planning the resection path of the ossified ligamentum flavum, including the steps:

[0027] (1) Obtain the preoperative three-dimensional image of the patient target region, and segment the ossified ligamentum flavum region according to the aforementioned ossified ligamentum flavum segmentation method;

[0028] (2) Mark the necessary region and the constraint region of the resection path in the preoperative three-dimensional image, use this as prior knowledge, and use the ant colony algorithm to plan the resection path;

[0029] (3) During the operation, obtain the three-dimensional image of the patient target region, and register the surface point cloud of the vertebral segment therein with the surface point cloud of the vertebral segment in the preoperative three-dimensional image;

[0030] (4) Based on the registration in step (3), transform the resection path of the ossified ligamentum flavum marked in the preoperative three-dimensional image to the three-dimensional image collected during the operation.

[0031] Beneficial effects: The present invention adopts an automated image segmentation method, which has high processing efficiency and can accurately identify and extract the ossified ligamentum flavum from many tissue regions of the human body. By combining the voxel characteristics of the ossified ligamentum flavum and the shape prior features of clinical image data of different classifications, the accuracy of positioning, identifying, and segmenting the ossified ligamentum flavum is improved, which helps to formulate a more accurate surgical plan, thereby improving the success rate of the operation. The present invention can plan a safer and more effective resection path of the ossified ligamentum flavum, provide surgical guidance for doctors, improve the success rate of the ossified ligamentum flavum resection operation, and further reduce the surgical risk. Brief Description of the Drawings

[0032] To more clearly illustrate the technical solutions in the present invention, the following will briefly introduce the accompanying drawings required for description in the embodiments. Obviously, the accompanying drawings in the following description are only embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0033] Figure 1 It is a flowchart of the ossified ligamentum flavum segmentation method of the present invention;

[0034] Figure 2 It is an example diagram of the sagittal image of the CT image of the patient's affected area;

[0035] Figure 3 It is an example diagram of the segmentation of the first bone region and the second bone region of the cross-sectional image of the CT image of the patient's affected area;

[0036] Figure 4 It is an example of the segmentation effect of the ossified ligamentum flavum. Specific Embodiments

[0037] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further elaborates on this application in combination with specific embodiments and with reference to the accompanying drawings.

[0038] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present invention should have the ordinary meaning understood by those with ordinary skills in the field to which the present invention belongs.

[0039] The present invention provides an ossified ligamentum flavum segmentation method as Figure 1 shown, including:

[0040] S1. Obtain the three-dimensional image of the patient's target area, and segment the first bone region and the second bone region in the three-dimensional image;

[0041] In the present invention, the human ligamentum flavum connects from the lower edge and inner surface of the superior vertebral lamina to the upper edge and outer edge of the inferior vertebral lamina, and participates in enclosing the posterior wall and the posterolateral wall of the vertebral canal. The ligamentum flavum undergoes heterotopic ossification under specific circumstances or due to metabolic disorders. For reference, see Figure 2 , the ligamentum flavum exists in the cervical, thoracic, and lumbar vertebrae. In theory, ossification of the ligamentum flavum can occur in any part of the spine.

[0042] In the present invention, an example of the three-dimensional image of the patient's target area is the CT cross-sectional image of a certain vertebral segment of a patient with ossification of the ligamentum flavum. As Figure 3 shown, the ossified ligamentum flavum ( Figure 3(not shown in the figure) are located below the vertebral body in the CT cross-sectional image, such as the spinous process, vertebral lamina, transverse process, and articular process. Their voxel values are similar and they all have bone characteristics. Since the boundary of the ossified ligamentum flavum is not clear, it is difficult to directly segment and remove it from them. Therefore, it is necessary to first segment the spinous process, vertebral lamina, and transverse process located below the vertebral body, and then screen out the ossified ligamentum flavum area. In the present invention, it is defined that Figure 3 the area of the spinous process, vertebral lamina, and transverse process located below the vertebral body in is the first bone area, and the vertebral body area is the second bone area, as Figure 3 shown.

[0043] In the present invention, due to the different anatomical characteristics of the ossified ligamentum flavum, it does not have typical shape characteristics, and the voxel of the lesion area of the ossified ligamentum flavum accounts for a very small proportion of the background of the entire image, belonging to a small target sample area. Therefore, the correct rate of directly segmenting the ossified ligamentum flavum by training the samples of the ossified ligamentum flavum is very low. However, the first bone area and the second bone area have typical shape characteristics. Therefore, the segmentation models of these two areas are trained to segment the first bone area and the second bone area, and then the ossified ligamentum flavum can be segmented through the prior spatial information of the segmented first bone area and second bone area.

[0044] In the present invention, a pre-trained segmentation model can be used to segment the first bone area and the second bone area in the CT cross-sectional image of a patient with ossified ligamentum flavum. The segmentation model is trained with a number of 3D images pre-labeled with the aforementioned first bone area (excluding the area of the ossified ligamentum flavum in the vertebral body) and the second bone area as training samples. Specifically, the segmentation model can be a pre-trained 3d_unet deep learning network model. In the present invention, the segmentation model can also adopt a pre-trained DeepLab V2 model or a Medical SAM 2 model.

[0045] S2. Based on the first bone area and the second bone area obtained in S1, search for the ossified ligamentum flavum area located between the first bone area and the second bone area according to the voxel characteristics of the ossified ligamentum flavum;

[0046] The voxel value of the ossified ligamentum flavum is very large, generally similar to that of bone. Based on this, a Hu value threshold can be set. In this embodiment, the set Hu value threshold is between 100 and 800. Referring to Figure 1 , in the CT cross-sectional image, the area where the ossified ligamentum flavum is shown to be located near the upper part of the first bone area. Therefore, by searching and extracting the voxel points with voxel values greater than the set Hu value threshold in the set space between the first bone area and the second bone area, the ossified ligamentum flavum area can be obtained.

[0047] Specifically, search and extract the voxel points with voxel values greater than the set Hu value threshold within the set space between the first bone region and the second bone region, as follows:

[0048] Traverse the voxel points within the first bone region. For any voxel point p among them i , establish a three-dimensional search space of size w*h*l centered on this voxel point p i . Search for other voxel points in this three-dimensional search space. If a voxel point does not belong to the first bone region and the second bone region, and its voxel value is greater than the set Hu value threshold, then extract this voxel point and put it into the seed point set. After traversing all the voxel points within the first bone region, the obtained seed point set is the rough ossified ligamentum flavum point set.

[0049] In the present invention, the size of the three-dimensional search space is determined in the following manner: In the human anatomical coordinate system, its height h is one-third of the distance between the lowest point of the vertebra in the second bone region and the highest point of the first bone region calculated (i.e., the height along the anterior-posterior direction of the human body), its width w is one-fifth of the width of the first bone region (i.e., the width along the left-right direction of the human body), and its length l is one-half of the depth of the vertebra in the second bone region (i.e., the length along the head-feet direction of the human body).

[0050] In the present invention, after completing the foregoing search and extraction process and obtaining the corresponding voxel points, all the voxel points form a seed point set. Traverse this seed point set. If a certain neighborhood point of the 8-neighborhood points of any voxel point therein satisfies that the voxel value is greater than the set Hu value threshold and is not in the first bone region and the second bone region, then put this neighborhood point into the seed point set as well. After traversing all the voxel points in the seed point set, no new voxel points will be added to the seed point set. At this time, the target point set is obtained, that is, the more accurate ossified ligamentum flavum point set in the foregoing steps is obtained.

[0051] In the present invention, after obtaining the more accurate ossified ligamentum flavum point set, it is necessary to locate the candidate region of the ossified ligamentum flavum, as follows: Calculate the boundary of the minimum bounding rectangle of the foregoing obtained target point set, and its boundary is determined by obtaining the leftmost, rightmost, frontmost, rearmost, uppermost, and lowermost points of the region corresponding to the target point set. Define the coordinates of the upper left corner point of the minimum bounding rectangle in the image coordinate system as (x min , y min， z min ), and the coordinates of the lower right corner point of the minimum bounding rectangle in the image coordinate system as (x max , y max， z max ). Then the width of the minimum bounding rectangle is W = x max - x min , and the height is H = y max - y min, the length D = z max -z min , thereby determining the ossified ligamentum flavum candidate region.

[0052] The ossified ligamentum flavum candidate region determined by the above method may contain a small amount of vertebral segment regions. Therefore, it is necessary to remove the small amount of vertebral segment regions contained therein. Based on the Sato classification method, the present invention classifies the ossified ligamentum flavum candidate region into the original classification template that is most similar to it, and then the small amount of vertebral segment regions contained therein can be excluded, and then the final true ossified ligamentum flavum region can be accurately located and segmented.

[0053] According to the different morphologies of ossified ligamentum flavum, the Sato classification divides it into lateral type, extended type, enlarged type, fusion type and nodular type according to the degree of compression on the spinal cord. These classifications reflect the morphological differences of most ossified ligamentum flavum. Therefore, it is necessary to collect the clinical images of ossified ligamentum flavum corresponding to the above several classifications as the original classification templates (in the present invention, one clinical image is obtained for one classification, and a total of five clinical images are obtained for five classifications. More clinical images corresponding to different classifications can also be collected according to actual needs). The ossified ligamentum flavum candidate region of the present invention has its specific scale and rotation, and the scales and rotations of each collected original classification template are also different. It is impossible to directly classify the ossified ligamentum flavum candidate region. Therefore, it is necessary to continuously perform scale and rotation transformations on each original classification template until the matching degree with the ossified ligamentum flavum candidate region is the highest, and then the ossified ligamentum flavum candidate region can be classified into the corresponding original classification template.

[0054] The new position coordinates of the voxel points obtained after each rotation and scale transformation of the original voxel points in each original classification template image in the image coordinate system are as follows:

[0055]

[0056] In the formula: (x, y, z) are the position coordinates of the original voxel point in the image coordinate system, (x new , y new , z new ) are the position coordinates of the transformed voxel point in the image coordinate system;

[0057]

[0058] Among them:

[0059]

[0060] Among them, S is the scale transformation matrix, that is, the scale transformation step length of the original voxel point, s x is the scaling scale in the x direction, s y is the scaling scale in the y direction, s zis the scaling scale in the z direction; R x is the rotation θ around the x-axis x of the transformation matrix, R y is the rotation θ around the y-axis y of the transformation matrix, R z is the rotation θ around the z-axis z of the transformation matrix, R x 、R y 、R z together constitute the rotation transformation step size of the original voxel points; θ x ranges from 0 - 360°, θ y ranges from 0 - 360°, θ z ranges from 0 - 360°.

[0061] In the present invention, in order to ensure accuracy, a finer scale transformation step size and rotation transformation step size can also be set, that is, a scale transformation step size is set, which can be respectively applied to the x, y, and z directions, and each application is a scale transformation; similarly, a rotation transformation step size is set, which can be respectively applied to the rotations around the x, y, and z axes, and each application is a rotation transformation.

[0062] Thus, after calculating the new position coordinates of the voxel points of each original segmentation template image after transformation, it may not be possible to directly obtain the voxel value. At this time, according to the voxel values of the integer position points around it, the method of linear interpolation is used to obtain its voxel value, that is, the voxel values of the new voxel points of each original segmentation template image after transformation are obtained. Furthermore, the segmentation template image data after rotation and scale transformation can be calculated and used as the candidate segmentation template image. For the above-mentioned set scale transformation step size and rotation transformation step size, different candidate segmentation template images can be obtained.

[0063] Based on this, in the present invention, the candidate regions of the ossified ligamentum flavum are classified into the corresponding original segmentation templates, specifically as follows:

[0064] (1) For any original segmentation template image, perform scale and rotation transformations on it to obtain multiple candidate segmentation template images;

[0065] a. Perform scale transformation on each original segmentation template image to generate a series of scaled original segmentation template images;

[0066] b. Perform rotation transformation on each original segmentation template image to generate a series of rotated original segmentation template images.

[0067] (2) For each candidate classification template image obtained through scale and rotation transformations, traverse the candidate regions of the ossified ligamentum flavum, perform a matching degree detection with the candidate regions of the ossified ligamentum flavum, calculate and count the maximum matching coefficient among them, and retain the candidate classification template, then the candidate regions of the ossified ligamentum flavum can be classified into the corresponding candidate classification templates.

[0068] In the present invention, the matching coefficient uses the normalized cross-correlation coefficient, which is specifically as follows:

[0069]

[0070] In the formula:

[0071]

[0072] Wherein, m is the voxel width of the candidate classification template image, n is the voxel height of the candidate classification template image, and d is the voxel depth of the candidate classification template image. m*n*d represents the window size of the candidate classification template image, f is the candidate region of the ossified ligamentum flavum, and t is the candidate classification template image. x+i, y+j, z+k are the positions in the x, y, and z directions of the coordinate system when the center point of the candidate classification template image slides on the corresponding image of the candidate region of the ossified ligamentum flavum. The value range of δ is [-1, 1]. When δ is equal to -1, it represents completely uncorrelated, and when δ is equal to 1, it represents completely correlated.

[0073] In the present invention, each original classification template image undergoes scale and rotation transformations based on the set scale transformation step size and rotation transformation step size to obtain multiple candidate classification template images. Each candidate classification template image will obtain a corresponding matching coefficient when performing a matching degree detection with the candidate regions of the ossified ligamentum flavum. Traverse all candidate classification template images to obtain the maximum matching coefficient, set the matching degree threshold. If the obtained maximum matching coefficient is greater than the matching degree threshold, then the maximum matching coefficient is valid, classify the corresponding original classification template into the corresponding candidate classification template, and based on the corresponding candidate classification template, remove a small amount of vertebral segment regions included in the candidate regions of the ossified ligamentum flavum, and then accurately locate and segment the true ossified ligamentum flavum, as Figure 4 shown.

[0074] In the present invention, by first segmenting the first bone region and the second bone region located below the vertebral body in the three-dimensional image, and then searching and extracting in the region near the first bone region according to the voxel characteristics of the ossified ligamentum flavum, an approximate ossified ligamentum flavum region can be obtained. Then, according to the shape prior features of the Sato classification of the ossified ligamentum flavum, the candidate regions of the ossified ligamentum flavum are classified into the corresponding Sato classifications, so as to remove a small amount of vertebral segments in the aforementioned approximate ossified ligamentum flavum region, thereby accurately locating, identifying, and segmenting the ossified ligamentum flavum from numerous tissue regions of the human body, and effectively ensuring the accuracy and success rate of the subsequent resection of the ossified ligamentum flavum.

[0075] The present invention also provides an ossified ligamentum flavum resection path planning method based on the aforementioned ossified ligamentum flavum segmentation method, including the steps of:

[0076] (1) Obtaining the preoperative three-dimensional image of the patient's target region, and the ossified ligamentum flavum region obtained according to the aforementioned ossified ligamentum flavum segmentation method;

[0077] (2) Marking the necessary regions and restricted regions in the resection path in the preoperative three-dimensional image, using this as prior knowledge, and using the ant colony algorithm to perform resection path planning;

[0078] Among them, the set restricted region of the resection path is the region that cannot be passed through during the resection process, such as relevant regions such as the spinal cord and nerve roots.

[0079] In this embodiment, the necessary regions of the resection path are the regions that will definitely be passed through during the resection process, and examples include the inferior articular process and lamina of the upper vertebra, the superior articular process and lamina of the lower vertebra, etc., as well as the ossified ligamentum flavum region.

[0080] (3) Obtaining the three-dimensional image of the patient's target region during the operation, and registering the vertebral segments therein with the vertebral segments in the aforementioned preoperative image;

[0081] Since the resection path of the ossified ligamentum flavum is planned in the preoperative three-dimensional image, it is necessary to transform the resection path of the ossified ligamentum flavum planned in the preoperative three-dimensional image to the three-dimensional image collected during the operation. First, it is necessary to establish the transformation relationship between the preoperative three-dimensional image and the three-dimensional image collected during the operation, and accordingly register the vertebral segments in the preoperative three-dimensional image with the vertebral segments in the intraoperative three-dimensional image, and unify the two to the same coordinate system, so as to transform the resection path of the ossified ligamentum flavum planned in the preoperative three-dimensional image to the three-dimensional image collected during the operation for navigation.

[0082] In the present invention, the three-dimensional image of the patient's target region collected during the operation is a CBCT image.

[0083] Specifically, the registration of the vertebral segment surface point cloud in the three-dimensional image collected during the operation with the vertebral segment surface point cloud in the aforementioned preoperative image is as follows:

[0084] (31) Adopt the above-mentioned segmentation method for the first bone region and the second bone region to separately segment the surface point clouds of the vertebral segment in the preoperative image and the three-dimensional image collected during the operation.

[0085] In the present invention, after obtaining the vertebral segments in the preoperative image and the three-dimensional image collected during the operation by adopting the above-mentioned segmentation method for the first bone region and the second bone region, the coordinates at the junction of the foreground and the background can be directly calculated as the position of the surface point cloud of the vertebral segment.

[0086] (32) Use the ICP algorithm to register the two sets of surface point clouds of the vertebral segments obtained in step (31) to obtain the initial registration result T0.

[0087] (33) Use the bounding box generation algorithm to generate the bounding box of the vertebral segment in the preoperative image respectively, and calculate the bounding box of the vertebral segment in the three-dimensional image collected during the operation according to the initial registration result T0 obtained in step (32).

[0088] (34) Input the bounding boxes of the vertebral segments in the preoperative image and the three-dimensional image collected during the operation obtained in step (33) into the voxel-based gradient descent optimization algorithm for optimization to obtain the final registration result T1.

[0089] In the present invention, since the rough registration in step (32) has been performed, the preoperative image and the three-dimensional image collected during the operation are already roughly aligned. Set the initial value of the optimization variable as T0, and this optimization variable T0 is already near the global optimal solution, which ensures that no local optimal solution will be obtained in the subsequent optimization process. In addition, in the present invention, through the initial registration result T0 obtained by rough registration, this initial registration result T0 can reduce the number of optimization iterations, thereby reducing the time consumption of the optimization algorithm.

[0090] In the present invention, in the voxel-based gradient descent optimization algorithm, the cross-correlation coefficient between the two images can be used to determine whether the two images are registered in place.

[0091] (4) Transform the resection path of the ossified ligamentum flavum planned in the preoperative three-dimensional image to the three-dimensional image collected during the operation according to the registration in step (3).

[0092] In the present invention, in order to further facilitate the actual application operation, after transforming the resection path of the ossified ligamentum flavum planned in the preoperative three-dimensional image to the three-dimensional image collected during the operation, the resection path of the ossified ligamentum flavum planned in the preoperative three-dimensional image can also be transformed to the two-dimensional image displayed by the endoscope during the operation. This process can obtain the transformation relationship between the three-dimensional image collected during the operation and the two-dimensional image of the endoscope through camera calibration, and accordingly transform the resection path of the ossified ligamentum flavum planned in the preoperative three-dimensional image to the two-dimensional image of the endoscope.

[0093] In the present invention, camera calibration can be achieved by any means of the prior art, which will not be elaborated herein.

[0094] Based on the aforementioned ossified ligamentum flavum segmentation method, the present invention extracts the ossified ligamentum flavum region. On this basis, by marking the necessary regions and constraint regions of the resection path as prior knowledge, the ant colony algorithm is used for resection path planning. Then, through registration, the ossified ligamentum flavum resection path planned in the preoperative three-dimensional image is transformed to the three-dimensional image acquired during the operation, so as to plan a safer and more effective ossified ligamentum flavum resection path, provide surgical guidance for doctors, improve the success rate of ossified ligamentum flavum resection, and thus reduce the surgical risk.

[0095] Those of ordinary skill in the art should understand that the discussion of any above embodiment is only exemplary and is not intended to imply that the scope of the present invention (including the claims) is limited to these examples; under the idea of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the embodiments of the present invention as described above, which are not provided in detail for the sake of brevity.

[0096] The embodiments of the present invention are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omission, modification, equivalent substitution, improvement, etc. made within the spirit and principle of the embodiments of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for ossified ligamentum flavum segmentation, characterized in that, Including: Obtain the three-dimensional image of the patient's target area. Define the spinous process, lamina, and transverse process area below the vertebral body in the vertebral segment as the first bone area, and the vertebral body area as the second bone area. Use a pre-trained segmentation model to segment the three-dimensional image of the patient's target area to obtain the first bone area and the second bone area therein. Set the Hu value threshold according to the voxel characteristics of the ossified ligamentum flavum. Traverse the voxel points in the first bone area, establish a three-dimensional search space of a set size centered on any one of the voxel points, extract the voxel points whose Hu values in the three-dimensional search space are greater than the set Hu value threshold and are not in the first bone area and the second bone area, and put them into the seed point set. Traverse the seed point set. If the voxel value of a certain neighborhood point of the 8-neighborhood points of any one of the voxel points is greater than the Hu value threshold and is not in the first bone area and the second bone area, then put this neighborhood point into the seed point set. After traversing all the voxel points in the seed point set, obtain the seed point set. Calculate the boundary of the minimum circumscribed rectangle of the seed point set composed of the extracted points, and determine the candidate area of the ossified ligamentum flavum accordingly; classify the candidate area of the ossified ligamentum flavum into the original classification template most similar to it based on the Sato classification method, and exclude the vertebral segment area in the candidate area of the ossified ligamentum flavum accordingly to obtain the area of the ossified ligamentum flavum.

2. The ossified ligamentum flavum segmentation method according to claim 1, wherein The specific three-dimensional search space of the set size is: In the human anatomical coordinate system, the height of the three-dimensional search space is one-third of the distance between the lowest point of the vertebral body in the calculated second bone area and the highest point of the first bone area, its width is one-fifth of the width of the first bone area, and its length is one-half of the depth of the vertebral body in the second bone area.

3. The ossified ligamentum flavum segmentation method according to claim 1, wherein The classification of the candidate area of the ossified ligamentum flavum into the original classification template most similar to it based on the Sato classification method is as follows: (1) For any original classification template image, perform scale and rotation transformations on it to obtain multiple candidate classification template images; a. Perform scale transformation on each original classification template image to generate a series of scaled original classification template images; b. Perform rotation transformation on each original classification template image to generate a series of rotated original classification template images; (2) For each candidate classification template image obtained by scale transformation and rotation transformation each time, perform a matching degree detection with the candidate area of the ossified ligamentum flavum, calculate the maximum matching coefficient among them, and retain this candidate classification template, that is, classify the candidate area of the ossified ligamentum flavum into the corresponding candidate classification template.

4. According to the ossified ligamentum flavum segmentation method described in claim 1, characterized in that The segmentation model is trained by using a number of three-dimensional images with the first bone area and the second bone area marked in advance as training samples.

5. A method for planning the resection path of ossified ligamentum flavum, characterized in that, Including the steps: (1) Obtain the preoperative three-dimensional image of the patient's target area, and segment the area of the ossified ligamentum flavum according to the ossified ligamentum flavum segmentation method described in any one of claims 1-4; (2) Mark the necessary area and the constraint area of the resection path in the preoperative three-dimensional image, use this as prior knowledge, and use the ant colony algorithm to plan the resection path. (3)Obtain the three-dimensional image of the patient's target area during the operation, and register the surface point cloud of the vertebral segment therein with the surface point cloud of the vertebral segment in the preoperative three-dimensional image; (4)Based on the registration in step (3), transform the resection path of the ossified ligamentum flavum marked in the preoperative three-dimensional image to the three-dimensional image acquired during the operation.

Images