Method for recognizing three-dimensional spatial posture of semicircular canal

Abstract

Provided in the present invention is a method for recognizing a three-dimensional spatial posture of a semicircular canal. The method comprises the following process: using a three-dimensional skeleton extraction algorithm for image processing to acquire disordered points of center lines of semicircular canals; on the basis of the mutual symmetry of a left semicircular canal and a right semicircular canal, distinguishing the left semicircular canal from the right semicircular canal by using the center of mass as a boundary; establishing a distance matrix between the disordered points of the center lines of the semicircular canals, and on the basis of a distance relationship, identifying intersection points and single points from the disordered points of the center lines of the semicircular canals; by means of connecting the single points, forming a center line of a rear semicircular canal, a center line of an upper semicircular canal and a center line of an outer semicircular canal; and on the basis of the center lines, fitting three-dimensional spatial models of the center lines of the semicircular canals. In the present application, a three-dimensional skeleton extraction algorithm for image processing is used to acquire disordered points of center lines of semicircular canals, adjacent disordered points are identified starting from poles of the center lines of the semicircular canals, and ordered center lines of the semicircular canals are constructed, thereby achieving a deep understanding of point-by-point spatial postures of the center lines of the semicircular canals, including spatial postures of semicircular canals in different segments.

Description

A method for recognizing the three-dimensional spatial posture of semicircular canals Technical field

[0001] The present invention relates to the technical field of medical image processing, and more specifically, it relates to a method for recognizing the three-dimensional spatial posture of semicircular canals. Background technique

[0002] Benign paroxysmal positional vertigo (BBPV) is a common peripheral vestibular disease clinically and is the most common vertigo disease originating from the inner ear. When the head moves to a certain specific position, it can induce transient vertigo, accompanied by nystagmus and autonomic nerve symptoms. This disease is self-limiting. The semicircular canals in the left and right ears are the posterior semicircular canal, the horizontal semicircular canal, and the superior semicircular canal in sequence.

[0003] Clinically, the otolith is induced to settle in a specific semicircular canal under the action of gravity by changing the head position, and the hydrodynamic force acts on the crista ampullaris to induce vertigo nystagmus to judge the position of the otolith. The Dix-Hallpike test is often used to diagnose posterior and superior semicircular canal BPPV, and the horizontal roll test is used to diagnose horizontal semicircular canal BPPV.

[0004] It can also be cured by a series of head position changes to make the displaced otolith return to the utricle again. The Epley method is commonly used to reduce posterior semicircular canal BPPV, and the roll reduction method is used to reduce horizontal semicircular canal BPPV.

[0005] Due to the large differences in the spatial directions of individual semicircular canals, it is also necessary to design individualized diagnostic and treatment methods according to the spatial directions of the patient's individual semicircular canals.

[0006] At present, Chinese Patent No. CN112288687A discloses an inner ear spatial posture analysis method and analysis system, which includes the following steps: constructing a segmentation model according to temporal bone image data. The temporal bone structures segmented by this model include: horizontal semicircular canal, posterior semicircular canal, superior semicircular canal, common canal, vestibule, cochlea and eyeball; rotating the segmented temporal bone image so that the top of the common canal and the bottom of the eyeball are located in the cross-section, and the bilateral inner ears are symmetric with respect to the sagittal plane, and accordingly constructing a spatial coordinate system; obtaining the center lines of the horizontal semicircular canal, posterior semicircular canal, and superior semicircular canal from the segmented temporal bone image and converting them into point arrays, using the least squares method to fit the planes respectively to obtain the plane equations of each semicircular canal, calculating the unit vectors of the plane normal vectors of the horizontal semicircular canal, posterior semicircular canal, and superior semicircular canal in the spatial coordinate system, constructing an individual semicircular canal spatial posture mathematical model, and formulating an individualized BPPV diagnosis and treatment plan for the patient; detecting multiple bilateral inner ear models, calculating the unit vector of the sum of the normal vectors of each semicircular canal plane in the spatial coordinate system of this model, and using this as the human semicircular canal spatial posture mathematical model to formulate a standardized BPPV diagnosis and treatment plan.

[0007] Although the above-mentioned patent method can establish the posture of the inner ear semicircular canals, it constructs a segmentation model based on temporal bone image data. The smallest segmentation structures are the horizontal semicircular canal, posterior semicircular canal, superior semicircular canal, common crus, vestibule, cochlea, and eyeball. It cannot intercept a certain segment of the semicircular canal for analysis and research as needed. Especially the body part of the posterior semicircular canal is often a position where otoliths are difficult to pass through. Understanding the spatial orientation of the body part of the posterior semicircular canal is very important for solving the problem of difficult reduction and improving the reduction effect. Therefore, it is necessary to have a deep understanding of the spatial posture of the semicircular canal and perform point-by-point analysis.

[0008] Therefore, how to deeply understand and perform point-by-point analysis on the spatial posture of the semicircular canal is exactly the problem to be solved by this application.

[0009] Summary of the Invention

[0010] Aiming at the deficiencies of the existing technology, a method for identifying the three-dimensional spatial posture of the semicircular canal is provided, which includes the following processes:

[0011] St10. Use the three-dimensional skeleton extraction algorithm of image processing to obtain the disordered points of the semicircular canal centerline.

[0012] St20. Based on the symmetry between the left and right semicircular canals, distinguish the left and right semicircular canals with the centroid as the boundary.

[0013] St30. Establish a distance matrix between the disordered points of each semicircular canal centerline, and identify the intersection points and single points of the disordered points of the semicircular canal centerline according to the distance relationship.

[0014] St40. Connect the single points to form the centerlines of the posterior semicircular canal, superior semicircular canal, and horizontal semicircular canal.

[0015] St50. Fit the three-dimensional spatial model of each semicircular canal centerline according to the centerline.

[0016] In summary, the above technical solution has the following beneficial effects: This application uses the three-dimensional skeleton extraction algorithm of image processing to obtain the disordered points of the semicircular canal centerline, and starts from the poles of each semicircular canal centerline to identify adjacent disordered points and construct an ordered semicircular canal centerline, so as to achieve a deep understanding of the point-by-point spatial posture of the semicircular canal centerline, including the spatial posture of different segments of the semicircular canal.

[0017] Since the judgment of the semicircular canal poles is very reliable, the identification of the semicircular canal in this patent is very robust and reliable, with the characteristics of small error and high reduction degree.

[0018] Most importantly, due to the in-depth understanding of the spatial attitude of each semicircular canal centerline point by point, a certain segment of the semicircular canal can be intercepted for analysis and research according to requirements. For example, studying the spatial attitude of the inferior arm of the posterior semicircular canal is beneficial to improving the diagnostic test; studying the spatial attitude of the body part of the posterior semicircular canal helps to improve the reduction technique, solve the problem of difficult reduction and enhance the reduction effect; studying the spatial attitude relationship among the utricular opening of the long arm of the lateral semicircular canal, the centerline of the common crus, and the utricle is beneficial to solving the problem of otolith entering the semicircular canal and developing a safe and effective reduction technique.

[0019] This application has the following advantages over the prior art:

[0020] 1. The existing method is to automatically identify the set of centerline points of each semicircular canal and identify different semicircular canals according to the intersection points and spatial positions. It depends on the ordered set of semicircular canal centerline points and is only applicable to special semicircular canal neutral line extraction algorithms. In this application, the three-dimensional skeleton extraction algorithm of image processing is used to obtain the disordered points of the semicircular canal centerline, and the method of extending the semicircular canal poles to identify adjacent disordered points is used to construct the ordered semicircular canal centerline. It is not only applicable to the disordered set of semicircular canal points, but also applicable to all semicircular canal centerline algorithms.

[0021] 2. The existing method uses VMTK (VMTK is an open-source C++ library for vascular structure segmentation, extraction, and analysis using ITK and VTK, and provides a plugin for the 3D Slicer software) to obtain the ordered semicircular canal centerline, and uses the vtkvmtkPolyDataCenterline function of VMTK to successively find the inscribed spheres of each semicircular canal pipeline. The centers of these inscribed spheres form the centerline of the corresponding semicircular canal. Through analysis and research, it can be known that the centerlines in the entire inner ear space include the central rings of the three semicircular canals, and there are common intersection points between the central rings. Therefore, the intersection points appear at least 3 times in the array, and the distances between the intersection points of the semicircular canal central rings have a fixed pattern. Based on this, other abnormal intersection points can be excluded. The above method depends on the VMTK algorithm and a special operating environment. The algorithm of this application does not depend on a special operating environment and has strong universality.

[0022] 3. The existing method only uses the least squares method to fit the plane to obtain the normal vector of the semicircular canal plane as the expression of the semicircular canal spatial attitude, lacking in-depth understanding and point-by-point analysis. Since the semicircular canal is a three-dimensional space pipeline system, it not only has a spatial structure similar to an ellipse, but also has a certain curvature. Multiple semicircular canal spatial attitude features are required, including the center position, curvature, bending degree, plane direction of a specific segment, and pipeline direction, etc. Because this application conducts in-depth understanding and point-by-point analysis of the semicircular canal spatial attitude, it can fit out the center position, curvature, bending degree, etc. of the three-dimensional space ellipse of the semicircular canal, and realize the analysis of multiple semicircular canal spatial attitude features. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a schematic flowchart of a method for identifying the three-dimensional spatial attitude of the semicircular canal;

[0024] Figure 2 is a schematic diagram of disordered points of a method for identifying the three-dimensional spatial attitude of the semicircular canal;

[0025] Figure 3 is a schematic diagram of incorporating disordered points on one side of the pole of a method for identifying the three-dimensional spatial attitude of the semicircular canal;

[0026] Figure 4 is a schematic diagram of incorporating disordered points on the other side of the pole of a method for identifying the three-dimensional spatial attitude of the semicircular canal;

[0027] Figure 5 is a schematic diagram of the center line of a method for identifying the three-dimensional spatial attitude of the semicircular canal;

[0028] Figure 6 is a schematic diagram of normal symmetry of a method for identifying the three-dimensional spatial attitude of the semicircular canal;

[0029] Figure 7 is a schematic diagram of non-normal symmetry of a method for identifying the three-dimensional spatial attitude of the semicircular canal.

[0030] Reference numerals: 10, posterior semicircular canal; 20, superior semicircular canal; 30, lateral semicircular canal; 40, intersection point; 50, single point; 60, pole; 70, useless point. Detailed implementation manners

[0031] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. The same components are denoted by the same reference numerals. It should be noted that the terms "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to the directions in the accompanying drawings, and the terms "bottom surface" and "top surface", "inner" and "outer" refer to the directions towards or away from the geometric center of a specific component respectively.

[0032] As shown in Figure 1, a method for identifying the three-dimensional spatial attitude of the semicircular canal includes the following processes:

[0033] St10: As shown in Figure 2, use the three-dimensional skeleton extraction algorithm of image processing to obtain the disordered points of the center line of the semicircular canal. The semicircular canal is an organ in the inner ear labyrinth that掌管平衡感 (responsible for balance). It consists of three mutually perpendicular annular tubes, namely the superior, posterior and lateral semicircular canals. The center line of the semicircular canal refers to the central axis of the semicircular canal, which can be used to fit the plane of the semicircular canal and calculate the normal vector of the semicircular canal. The semicircular canal includes structures such as the anterior semicircular canal, posterior semicircular canal, lateral semicircular canal, saccule and utricle, etc. The obtained disordered points are the points on the center lines of all structures.

[0034] St20: Based on the symmetry between the left and right semicircular canals, distinguish the left and right semicircular canals with the centroid as the boundary.

[0035] St30. Establish a distance matrix between the disordered points on the centerlines of each semicircular canal. According to the distance relationship, identify the intersection points 40 and single points 50 of the disordered points on the centerlines of the semicircular canals. The core code for calculating the point-to-point distance matrix is: dist_matrix = np.linalg.norm(points[:, np.newaxis, :] - points[np.newaxis, :, :], axis=-1). The single points 50 are the points that make up the centerlines of the superior semicircular canal, the posterior semicircular canal, and the horizontal semicircular canal. The intersection points 40 are the points that make up other structures such as the saccule and the utricle.

[0036] St31. Determine whether there are other disordered points within the neighborhood threshold of a certain disordered point. When it is determined that there are two disordered points within the neighborhood threshold of a certain disordered point, then determine that this disordered point is a single point 50.

[0037] St32. When it is determined that there are more than two disordered points within the neighborhood threshold of a certain disordered point, then determine that this disordered point is an intersection point 40. The neighborhood threshold refers to a spatial region within a certain range centered on a certain point, and its size can be determined according to the actual situation. Generally, there are two disordered points within the neighborhood threshold for a single point 50, and there are more than two disordered points within the neighborhood threshold for an intersection point 40. Through this feature, the single points 50 and the intersection points 40 can be identified.

[0038] St33. According to the distance relationship, it is also used to identify the useless points 70 of the disordered points on the centerlines of the semicircular canals. When it is determined that there are less than two disordered points within the neighborhood threshold of a certain disordered point, then determine that this disordered point is a useless point 70. When the semicircular canals incorporate single points 50 from the poles 60 of the centerlines of different semicircular canals to both sides, the useless points 70 are excluded. The definition of an intersection point 40 is that there are more than two disordered points within the neighborhood threshold; the definition of a single point 50 is that there are two disordered points within the neighborhood threshold; the definition of a useless point 70 is that there are less than two disordered points within the neighborhood threshold. Specifically, a terminal point is a point with one disordered point within the neighborhood threshold, and an isolated point is a point with no disordered points within the neighborhood threshold. Both the terminal point and the isolated point are useless points 70. The relevant code for identifying the intersection points 40, single points 50, and useless points 70 of the disordered points on the centerlines of the semicircular canals is: sdix = np.sum(dist_matrix < 2, axis = 1).

[0039] St40. As shown in FIGS. 3 - FIGS. 5, by connecting the single points 50, the centerlines of the posterior semicircular canal 10, the superior semicircular canal 20, and the horizontal semicircular canal 30 are formed.

[0040] St41. The first method for connecting single points 50 includes the following process: Identify the single point 50 at the rearmost of the disordered points on the semicircular canal centerline as the pole 60 of the posterior semicircular canal 10 centerline, identify the single point 50 at the uppermost of the disordered points on the semicircular canal centerline as the pole 60 of the superior semicircular canal 20 centerline, and identify the single point 50 at the outermost of the disordered points on the semicircular canal centerline as the pole 60 of the horizontal semicircular canal 30 centerline; With the pole 60 of the semicircular canal centerline as the center, incorporate the single points 50 of the disordered points on the semicircular canal centerline from the poles 60 of different semicircular canal centerlines to both sides respectively, so as to construct the centerlines of each semicircular canal until the disordered points incorporated into the semicircular canal centerline are the intersection points 40. The first method for connecting single points 50 starts from the poles 60 of each semicircular canal, and successively incorporates the disordered points within the adjacent threshold from both sides for judgment. The incorporated points are all single points 50 until the disordered points within the incorporated adjacent threshold have intersection points 40, then stop constructing the semicircular canal centerline. The prior art uses the least squares method to fit a plane. This application uses the Singular Value Decomposition (SVD) algorithm to obtain the semicircular canal plane and can fit a three-dimensional space circle or ellipse. For the convenience of identifying different semicircular canals, the centerlines of different semicircular canals can be marked with different colors respectively.

[0041] St42. The second method for connecting single points 50 includes the following process: Taking the single point 50 adjacent to the intersection point 40 as the starting point, incorporate the single points 50 of the disordered points on the semicircular canal centerline to the other side, so as to construct the centerlines of each semicircular canal until the disordered points incorporated into the semicircular canal centerline are the intersection points 40; Identify the centroid of the centerline of each semicircular canal. The centerline with the centroid at the rearmost is the centerline of the posterior semicircular canal 10, the centerline with the centroid at the uppermost is the centerline of the superior semicircular canal 20, and the centerline with the centroid at the outermost is the centerline of the horizontal semicircular canal 30. The second method for connecting single points 50 does not need to find the poles and distinguishes and identifies the posterior semicircular canal centerline, the horizontal semicircular canal centerline, and the superior semicircular canal centerline through the centroid positions of each semicircular canal.

[0042] St43. Constructing the center of each semicircular canal includes the following process: Incorporate the disordered points within the adjacent threshold and judge whether the disordered points within the adjacent threshold are intersection points 40 or single points 50. When it is judged to be a single point 50, continue to incorporate the disordered points within the adjacent threshold. When it is judged to be an intersection point 40, stop constructing the semicircular canal centerline.

[0043] St44. The near - neighbor threshold is greater than the single - distance of adjacent center - line disordered points and less than or equal to twice the distance of adjacent center - line disordered points. The near - neighbor threshold is the effective discrimination distance for the semi - circular canal center - line to incorporate disordered points. If there are no other nearby points within the near - neighbor threshold for a certain point, it can be judged as a single point and can be incorporated into the construction of the semi - circular canal center - line. If it is less than or equal to twice the minimum distance of adjacent center - line disordered points, it can ensure the identification of intersection points. Preferably, the near - neighbor threshold is equal to twice the minimum distance of adjacent center - line disordered points. In fact, the center - line disordered points are obtained according to the three - dimensional skeleton extraction algorithm in image processing, and the obtained density can be adjusted. After obtaining the center - line disordered points with a certain density, the distances between different adjacent center - line disordered points may also vary slightly. When determining that the near - neighbor threshold is greater than the single - distance of adjacent center - line disordered points, select the adjacent center - line disordered points with the maximum distance, so that there will be no situation where there are no disordered points to incorporate. When determining that the near - neighbor threshold is less than or equal to twice the distance of adjacent center - line disordered points, select the adjacent center - line disordered points with the minimum distance. In this way, the determined near - neighbor threshold can effectively incorporate single points and at the same time can determine intersection points.

[0044] St50. Fit the three - dimensional space models of each semi - circular canal according to the center - line.

[0045] St51. Import the three - dimensional space models of the semi - circular canals into a three - dimensional environment, distinguish each semi - circular canal model, and correspond the constructed center - lines of each semi - circular canal and each semi - circular canal model one by one. Use the singular - value decomposition algorithm to fit the plane of each semi - circular canal according to the center - lines of each semi - circular canal. Make cross - sections perpendicular to the semi - circular canal plane at both ends of the center - lines of each semi - circular canal, so as to obtain the three - dimensional space structure of each semi - circular canal. Analyze the complete semi - circular canal model through the constructed center - line, and split the semi - circular canal model into each component according to the center - line, obtaining the three - dimensional space models of each semi - circular canal with the center - line. Because the center - line is fitted, it is convenient to intercept any segment of the semi - circular canal and calculate the spatial directions of each part. For the convenience of observation, cross - sections can also be established in the extending direction of the center - lines of each semi - circular canal, so as to obtain the cross - section models of each semi - circular canal.

[0046] St52. As shown in Figures 6 and 7, after fitting the plane of each semi - circular canal, mark the normal vectors of each semi - circular canal plane at the center of the center - line; judge whether there is an error in the fitting result according to the symmetry of the normal vectors of the left semi - circular canal and the right semi - circular canal. The left semi - circular canal includes the upper semi - circular canal 20, the posterior semi - circular canal 10 and the lateral semi - circular canal 30 on the left side. The right semi - circular canal also includes the upper semi - circular canal 20, the posterior semi - circular canal 10 and the lateral semi - circular canal 30 on the right side. There is a natural symmetry plane between the left and right semi - circular canals, which is the sagittal plane of the human body. The normal vectors of the left semi - circular canal and the right semi - circular canal are judged whether they are symmetric with the sagittal plane of the human body as the symmetry plane.

[0047] St53. The three-dimensional space model of the semicircular canals further includes an eyeball model. A three-dimensional space coordinate system is established based on the eyeball model, and the spatial posture of the semicircular canals is calibrated. The spatial coordinate system of the semicircular canals has been described in the prior art. Specifically, through the study of a large number of samples, it is found that the plane formed by the lowest edges of the bilateral eyeballs and the top of the common canal of the bilateral semicircular canals, that is, the fundus plane of the semicircular canals, can be recognized as the horizontal plane. Therefore, the intersection point above the posterior semicircular canal and the superior semicircular canal corresponds to the top of the common canal. By rotating the segmented temporal bone image, the top of the common canal and the bottom of the eyeball are located in the cross-section, and the bilateral inner ears are symmetric with respect to the sagittal plane.

[0048] St54. Using the method of deep learning, the semicircular canals and the eyeballs are automatically segmented. The plane equation is fitted according to the centerlines of the respective semicircular canals, and the spatial posture of the semicircular canals is calibrated according to the three-dimensional space coordinate system of the semicircular canals and the eyeballs.

[0049] This application uses the three-dimensional skeleton extraction algorithm of image processing to obtain the disordered points of the semicircular canals, and uses the method of extending the poles of the semicircular canals to identify adjacent disordered points to construct an ordered centerline of the semicircular canals. The sorting of disordered points is irregular, that is, disordered. The disordered points required for fitting become the centerline and then a three-dimensional space model of the semicircular canals is constructed. Because the three-dimensional space model of the semicircular canals is constructed segment by segment from the smallest unit of disordered points, a deep understanding of the spatial posture of each point on the centerline of the semicircular canals can be achieved, including the spatial posture of different segments of the semicircular canals. Since the poles 60 of the semicircular canals are very reliable, the identification of the semicircular canals in this patent is very robust and reliable, with the characteristics of small error and high restoration degree.

[0050] Most importantly, due to the deep understanding of the spatial posture of each point on the centerline of each semicircular canal, a certain segment of the semicircular canal can be intercepted for analysis and research according to needs. For example, studying the spatial posture of the lower arm of the posterior semicircular canal 10 is beneficial to improving the diagnostic test; studying the spatial posture of the body part of the posterior semicircular canal 10 helps to improve the reduction technique, solve the problem of difficult reduction and improve the reduction effect; studying the spatial posture relationship between the oval window opening of the long arm of the lateral semicircular canal, the centerline of the common canal and the utricle is beneficial to solving the problem of otolith entering the semicircular canal and developing a safe and effective reduction technique.

[0051] This application has the following advantages over the prior art:

[0052] 1. The existing method is to automatically identify the set of centerline points of each semicircular canal, and identify different semicircular canals according to the intersection point 40 and the spatial position. It depends on the ordered set of centerline points of the semicircular canals and is only applicable to special algorithms for extracting the neutral line of the semicircular canals. However, this application uses the skeleton extraction algorithm of image processing to obtain the disordered points of the semicircular canals, so it is applicable to the disordered set of semicircular canal points and all algorithms for the centerline of the semicircular canals are applicable.

[0053] 2. The existing method uses VMTK (VMTK is an open-source C++ library that uses ITK and VTK for vascular structure segmentation, extraction, and analysis, and provides a plugin for the 3D Slicer software) to obtain the centerlines of the semicircular canals. Using the vtkvmtkPolyDataCenterline function of VMTK, the inscribed spheres of each semicircular canal are calculated in sequence, and the centers of these inscribed spheres form the centerlines of the corresponding semicircular canals. Through analysis and research, it can be seen that the centerlines in the entire inner ear space include the central rings of the three semicircular canals, and there are common intersection points 40 between the central rings. Therefore, the intersection point 40 appears at least 3 times in the array. The distances between the intersection points 40 of the central rings of the semicircular canals have a fixed pattern, based on which other abnormal intersection points 40 can be excluded. The above method depends on the VMTK algorithm and a special operating environment. However, the algorithm of this application does not depend on a special operating environment and has strong universality.

[0054] 3. The existing method only uses the least squares method to fit a plane to obtain the normal vector of the semicircular canal plane as the expression of the spatial attitude of the semicircular canal, lacking in-depth understanding and point-by-point analysis. Since the semicircular canal is a three-dimensional space pipe system, it not only has a spatial structure similar to an ellipse but also has a certain curvature, and requires multiple representations of the spatial attitude of the semicircular canal, including the center position, curvature, bending degree, plane direction of specific segments, and pipe direction, etc. Because this application conducts in-depth understanding and point-by-point analysis of the spatial attitude of the semicircular canal, it can fit out the center position, curvature, bending degree, etc. of the three-dimensional space ellipse of the semicircular canal, realizing the analysis of multiple spatial attitude characteristics of the semicircular canal.

[0055] 4. The existing technology uses VMTK to obtain the centerlines of the ordered semicircular canals, and VMTK requires a special operating environment and a specific Python version. This application uses a three-dimensional skeleton algorithm that does not require a special operating environment and supports running on smart terminals.

[0056] 5. The analysis effects of the existing technology and this application are different. The existing technology algorithm highly depends on the information of the intersection points 40 of the semicircular canal centerlines and requires the centerlines to be ordered. Its principle is to segment the semicircular canals using the intersection points 40 of the ordered semicircular canal centerlines, and distinguish the posterior, lateral, and superior semicircular canals 20 according to the lengths and positions of each segment. It is sensitive to the order and integrity of the centerlines and is easily interfered by abnormal situations such as centerline bifurcation and abnormal starting points, and does not support unordered points of unordered semicircular canals.

[0057] The algorithm of this application supports unordered semicircular canal disordered points, and uses the spatial orientation characteristics of the semicircular canals to obtain the poles 60 of each semicircular canal. The algorithm is robust and not interfered by abnormal situations. The order of point selection is crucial. First, determine the pole of the posterior semicircular canal 10, find the center line of the posterior semicircular canal 10 and extract it; then determine the pole of the superior semicircular canal 20, find the center line of the superior semicircular canal 20 and extract it; finally, determine the pole of the lateral semicircular canal 30 from the remaining other disordered points, and find the center line of the lateral semicircular canal 30; further, the common crus and other parts can be distinguished. The principle of the algorithm of this patent is that through the robust poles 60 of the semicircular canals, single points 50 can be obtained first until intersection points 40 appear, and then the center line is fitted and then the three-dimensional model is fitted. It is also possible to fit a plane first after obtaining single points 50, and gradually extend it to fit the three-dimensional space circle of the semicircular canal. Through constraint conditions including the distance between the disordered point and the plane, the distance between the disordered point and the center of the circle, the angle between the line connecting the disordered point and the center of the circle and the plane of the semicircular canal, and the spatial orientation of the line connecting the disordered point and the center of the circle, single points 50 are gradually included and the process of fitting the plane and fitting the three-dimensional space model of the semicircular canal is repeated.

[0058] 6. The analysis capabilities of the prior art and this application are different. The prior art algorithm can only obtain the center lines of each semicircular canal.

[0059] The algorithm of this application can not only obtain the center lines of each semicircular canal, but also conduct in-depth analysis on the center lines of each semicircular canal, and obtain the center lines of the semicircular canals at different positions and segments according to research needs. Through distance analysis, starting from the poles 60 of the semicircular canals, single points 50 are gradually connected, making the unordered single points 50 of the semicircular canals become ordered. Through plane fitting and curve fitting, a deep understanding of the spatial posture of the semicircular canals and point-by-point analysis are formed.

[0060] For posterior semicircular canal 10 BPPV, its diagnostic experiment is the Dix-Hallpike experiment, which induces the otolith to slide down along the lower arm of the posterior semicircular canal 10 from the ampulla of the posterior semicircular canal 10. What needs to be studied is the spatial orientation of the semicircular canal where the otolith slides. It not only requires the plane to be parallel to the gravity direction, but also requires the orientation of this segment of the semicircular canal to be consistent with the gravity direction. Similarly, during the reduction manipulation process, it is also required that the otolith sliding segments at different positions of the posterior semicircular canal 10 satisfy that the plane is parallel to the gravity direction and the orientation of this segment of the semicircular canal is consistent with the gravity direction, which can improve the reduction effect. Especially for the body part of the posterior semicircular canal 10, it is often a position where the otolith is difficult to pass through. Understanding the spatial orientation of the body part of the posterior semicircular canal 10 is very important for solving the problem of difficult reduction and improving the reduction effect. The algorithm of this patent can also conduct in-depth analysis on the spatial characteristics of the semicircular canals through the fitting of three-dimensional space circles / ellipses.

[0061] Due to the long code, only part of the code of this application is shown below. The following is the program code for the left posterior semicircular canal.

[0062] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements should also be regarded as within the protection scope of the present invention.

Claims

1. A method for identifying the three-dimensional spatial attitude of a semicircular canal, characterized in that, It includes the following processes: Adopt a three-dimensional skeleton extraction algorithm for image processing to obtain the disordered points of the semicircular canal centerline; Based on the symmetry between the left and right semicircular canals, distinguish the left and right semicircular canals with the centroid as the boundary; Establish a distance matrix between the disordered points of each semicircular canal centerline, and identify the intersection points (40) and single points (50) of the disordered points of the semicircular canal centerline according to the distance relationship; By connecting the single points (50), the centerlines of the posterior semicircular canal (10), the superior semicircular canal (20), and the lateral semicircular canal (30) are formed respectively; Fit the three-dimensional space models of each semicircular canal centerline according to the centerlines.

2. The three-dimensional spatial attitude recognition method of the semicircular canal according to claim 1, wherein Identifying the intersection points (40) and single points (50) of the disordered points of the semicircular canal centerline includes the following processes: Judge whether there are other disordered points within the neighbor threshold of a certain disordered point of the semicircular canal centerline. When it is judged that there are two disordered points within the neighbor threshold of a certain disordered point, then judge that this disordered point is a single point (50); When it is judged that there are more than two disordered points within the neighbor threshold of a certain disordered point, then judge that this disordered point is an intersection point (40).

3. The three-dimensional spatial attitude recognition method of the semicircular canal according to claim 2, wherein According to the distance relationship, it is also used to identify the useless points (70) of the disordered points of the semicircular canal centerline. When it is judged that there are less than two disordered points within the neighbor threshold of a certain disordered point, then judge that this disordered point is a useless point (70). When the semicircular canals incorporate single points (50) from the poles (60) of different semicircular canal centerlines to both sides, the useless points (70) are excluded.

4. The three-dimensional spatial attitude recognition method for semicircular canals according to claim 1, wherein Connecting the single points (50) includes the following processes: Identify the single point (50) at the rearmost of the disordered points of the semicircular canal centerline as the pole (60) of the posterior semicircular canal (10) centerline, identify the single point (50) at the uppermost of the disordered points of the semicircular canal centerline as the pole (60) of the superior semicircular canal (20) centerline, and identify the single point (50) at the outermost of the disordered points of the semicircular canal centerline as the pole (60) of the lateral semicircular canal (30) centerline; Centering on the pole (60) of the semicircular canal centerline, incorporate the single points (50) of the disordered points of the semicircular canal centerline from the poles (60) of different semicircular canal centerlines to both sides respectively, so as to construct the centerlines of each semicircular canal until the disordered points incorporated into the semicircular canal centerline are intersection points (40).

5. The three-dimensional spatial attitude recognition method for semicircular canals according to claim 1, characterized in that Connecting the single points (50) includes the following processes: Taking the single point (50) adjacent to the intersection point (40) as the starting point, incorporate the single points (50) of the disordered points of the semicircular canal centerline to the other side, so as to construct the centerlines of each semicircular canal until the disordered points incorporated into the semicircular canal centerline are intersection points (40); Identify the centroid of the centerline of each semicircular canal. The centerline with the centroid at the rearmost is the centerline of the posterior semicircular canal (10), the centerline with the centroid at the uppermost is the centerline of the superior semicircular canal (20), and the centerline with the centroid at the outermost is the center of the lateral semicircular canal (30) centerline.

6. The three-dimensional spatial attitude recognition method of the semicircular canal according to claim 4 or 5, characterized in that Constructing the center of each semicircular canal includes the following processes: Incorporate the disordered points within the neighbor threshold, and judge whether the disordered points within the neighbor threshold are intersection points (40) or single points (50). When it is judged to be a single point (50), then continue to incorporate the disordered points within the neighbor threshold. When it is judged to be an intersection point (40), then stop constructing the semicircular canal centerline.

7. The three-dimensional spatial attitude recognition method of the semicircular canal according to claim 6, wherein The near-neighbor threshold is greater than one times the distance of the disordered points adjacent to the centerline and less than or equal to two times the distance of the disordered points adjacent to the centerline.

8. The three-dimensional spatial attitude recognition method for semicircular canals according to any one of claims 1-5, characterized in that Fitting the three-dimensional spatial models of each semicircular canal according to the centerline includes the following processes: Import the three-dimensional spatial models of the semicircular canals in a three-dimensional environment, distinguish each semicircular canal model, correspond each constructed semicircular canal centerline with each semicircular canal model one by one, fit the plane of each semicircular canal using the singular value decomposition algorithm according to the centerlines of each semicircular canal, and make cross-sections perpendicular to the semicircular canal plane at both ends of each semicircular canal centerline, so as to obtain the three-dimensional spatial structure of each semicircular canal.

9. The method for identifying the three-dimensional spatial attitude of the semicircular canal according to claim 8, wherein After fitting the centerline planes of each semicircular canal, mark the normal vector of the plane at the center of each semicircular canal centerline plane, and judge whether there is an error in the fitting result according to the symmetry of the normal vectors of the left and right semicircular canals.

10. The method for identifying the three-dimensional spatial attitude of the semicircular canal according to claim 9, wherein The three-dimensional spatial model of the semicircular canal also includes an eyeball model. Establish a three-dimensional space coordinate system according to the eyeball model and calibrate the spatial attitude of the semicircular canal.

Images