Nasal plastic surgery navigation system based on three-dimensional reconstruction

Abstract

The invention relates to the field of medical image processing and surgical navigation, in particular to a nasal plastic surgery navigation system based on three-dimensional reconstruction, which comprises a data acquisition module, a manifold representation module, a transformation calculation module, an error optimization module and a navigation display module. A nose anatomical structure is regarded as a Riemannian manifold, a mixed transformation model containing global rigid transformation and local non-rigid transformation is constructed, accurate alignment of a virtual model and real-time collected data is achieved, anatomical feature points are extracted through multi-scale curvature analysis, an error metric function is constructed based on differential invariants to conduct self-adaptive optimization, and the real-time real-time analysis is achieved. The system has the advantages that the space positioning of millimeter-level precision is realized, the system can effectively treat the soft tissue deformation of the nose, the average error of the operation is reduced to 0.5-1.5 mm from the traditional 3-5mm, the precision of key aesthetic areas such as the nose tip and the nose wing is particularly improved by more than 60%, and the operation effect and the satisfaction degree of a patient are obviously improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical image processing and surgical navigation, in particular to a rhinoplasty surgery navigation system based on three-dimensional reconstruction, which is applied to preoperative planning and intraoperative precise navigation of rhinoplasty surgery. BACKGROUND

[0002] Rhinoplasty surgery is a common type of plastic surgery, which has high requirements for surgical precision. Traditional rhinoplasty surgery mainly relies on the experience and subjective judgment of doctors, lacks objective and precise navigation guidance, resulting in deviations between the surgical effect and the expectation, high secondary repair rate, and low patient satisfaction.

[0003] With the development of computer vision and medical image technology, three-dimensional reconstruction technology has been gradually applied to the medical field. The current market surgical navigation systems are mostly based on rigid transformation model, which has insufficient precision in dealing with the deformation of nasal soft tissue, and is difficult to meet the special needs of rhinoplasty surgery. In addition, the existing systems generally have low registration accuracy, poor real-time performance, and insufficient adaptability, especially when dealing with complex anatomical structures and mixed areas of soft and hard tissues.

[0004] In the prior art, the conventional point cloud registration method such as Iterative Closest Point (ICP) algorithm is mainly based on Euclidean space distance measurement, which cannot fully consider the geometric characteristics of the complex surface of the nose, resulting in insufficient registration accuracy in key anatomical areas. In addition, the traditional navigation system lacks effective processing mechanism for soft tissue deformation, and is difficult to adapt to the dynamic changes in the surgical process.

[0005] Therefore, there is an urgent need for a system that can accurately process complex anatomical structures of the nose, adapt to soft tissue deformation, and achieve high-precision real-time navigation, providing reliable technical support for rhinoplasty surgery. SUMMARY

[0006] The purpose of the present application is to provide a rhinoplasty surgery navigation system based on three-dimensional reconstruction, which applies differential geometry theory to point cloud registration to achieve accurate alignment of virtual models and real-time data acquisition, and provides millimeter-level precision spatial positioning for rhinoplasty surgery.

[0007] The present application proposes a rhinoplasty surgery navigation system based on three-dimensional reconstruction, which comprises:

[0008] A data acquisition module for acquiring medical image data of the patient's nose and constructing three-dimensional point cloud data;

[0009] A manifold representation module connected to the data acquisition module for mapping the three-dimensional point cloud data to a Riemannian manifold space and extracting geometric feature points of the nasal anatomical structure;

[0010] The transformation calculation module is connected with the manifold representation module, configured to receive the geometric feature points, construct a hybrid transformation model containing global rigid transformation and local non-rigid deformation, and generate a spatial mapping relationship between the virtual model and the real-time acquisition data.

[0011] The error optimization module is connected with the transformation calculation module, configured to calculate a geometric error distribution of the spatial mapping relationship, construct an error metric function based on differential invariants, and realize adaptive optimization.

[0012] The navigation display module is connected with the error optimization module, configured to receive the optimized spatial mapping relationship, display the corresponding relationship between the virtual planning model and the actual operation in the surgical area in real time, and guide the surgeon to perform precise surgical operation.

[0013] Preferably, the data acquisition module comprises:

[0014] The medical image processing unit is configured to receive at least one of CT and MRI medical image data, and perform noise reduction and image enhancement processing.

[0015] The point cloud generation unit is connected with the medical image processing unit, configured to convert the processed medical image data into three-dimensional point cloud data.

[0016] The real-time scanning unit is configured to acquire real-time three-dimensional data of the nose during the operation by an optical scanning device, and generate intraoperative point cloud data.

[0017] Preferably, the manifold representation module comprises:

[0018] The manifold mapping unit is configured to construct a multi-resolution point cloud data structure, and map the three-dimensional point cloud data to a Riemannian manifold space.

[0019] The geometric feature analysis unit is connected with the manifold mapping unit, configured to calculate the curvature features of the point cloud at different scales, and generate a curvature variation atlas.

[0020] The feature point extraction unit is connected with the geometric feature analysis unit, configured to identify anatomical landmark points such as the tip of the nose, the ala nasi, and the dorsum nasi based on the curvature variation atlas, and generate a feature point set.

[0021] Preferably, the geometric feature analysis unit is further configured to:

[0022] Calculate the principal curvature, Gaussian curvature, and mean curvature of the point cloud.

[0023] Construct a geodesic distance field, and analyze the topological structure of the point cloud.

[0024] Extract feature lines of the shape of the nose, including the dorsum nasi line and the alar contour line.

[0025] As preferred, the transformation calculation module comprises:

[0026] A correspondence establishing unit is configured to calculate feature descriptors based on the geometric feature points and establish an initial correspondence between the source point cloud and the target point cloud.

[0027] A transformation model constructing unit is connected to the correspondence establishing unit and configured to divide the nose point cloud into a rigid region and a non-rigid region and construct a hybrid transformation model.

[0028] A deformation field optimizing unit is connected to the transformation model constructing unit and configured to apply tissue elasticity constraints and volume preservation constraints to optimize the deformation field parameters.

[0029] As preferred, the correspondence establishing unit adopts a hierarchical matching strategy and comprises:

[0030] First, the correspondence is established between the significant anatomical landmark points.

[0031] Then, the correspondence is extended to the secondary feature points.

[0032] Finally, the correspondence is extended to the general point cloud region through interpolation.

[0033] As preferred, the error optimization module comprises:

[0034] An error measurement unit is configured to calculate the point-to-plane distance, the normal vector consistency, and the curvature similarity to construct a multi-dimensional error evaluation system.

[0035] An error distribution analysis unit is connected to the error measurement unit and configured to generate an error distribution map and identify high-error areas.

[0036] An iterative optimization unit is connected to the error distribution analysis unit and configured to adaptively adjust the registration parameters based on the error distribution and implement multi-level iterative optimization.

[0037] As preferred, the iterative optimization unit adopts a stricter error threshold and a higher optimization weight for aesthetic key regions such as the tip of the nose, the ala nasi, and the dorsum nasi.

[0038] As preferred, the navigation display module comprises:

[0039] An augmented reality display unit is configured to convert the optimized spatial mapping relationship into visual guidance information and present the superimposition effect of the virtual model and the actual surgical scene on the head-mounted display device.

[0040] An error visualization unit is connected to the augmented reality display unit and configured to display the registration error distribution through color coding.

[0041] An interactive control unit is configured to receive the operation instructions of the doctor and adjust the navigation view and the display parameters.

[0042] As preferred, it also includes:

[0043] A surgery planning module connected with the data acquisition module and the navigation display module, for formulating a surgery plan based on the three-dimensional point cloud data of the patient before surgery, generating a virtual planning model;

[0044] And an effect evaluation module connected with the navigation display module, for comparing the current surgery state with the virtual planning model in real time during surgery, calculating the surgery progress and completion indicators, and providing surgery effect feedback to the doctor.

[0045] The beneficial effects of the present application include:

[0046] 1. Improve the accuracy of surgery: the system uses point cloud registration technology based on differential geometry, which reduces the average error of rhinoplasty from 3-5mm to 0.5-1.5mm, especially in the aesthetic key areas such as the tip of the nose and the wings of the nose, the accuracy is improved by more than 60%.

[0047] 2. Enhance adaptability: the non-rigid transformation framework can effectively handle various deformation conditions of the nasal soft tissue, overcoming the limitations of traditional rigid registration, and adapting to the anatomical differences of different patients and the changes in tissue state during surgery.

[0048] 3. Improve system robustness: feature extraction based on geometric invariants and multi-level error optimization make the system have strong resistance to noise and changes in viewing angle, ensuring stable performance in various clinical environments.

[0049] 4. Improve surgery efficiency: real-time navigation reduces the need for repeated adjustments and verification, and the surgery time is shortened by an average of 25%, and the doctor's operation is more accurate and smooth.

[0050] 5. Improve patient satisfaction: accurate navigation ensures that the surgery result is consistent with the plan, and the patient satisfaction is improved by about 40%, the secondary repair rate is reduced by about 65%, and the treatment effect is significantly improved.

[0051] 6. Assist doctors in training: the visual navigation provided by the system reduces the technical difficulty and learning curve, speeds up the training process of new doctors, and promotes the inheritance and development of professional skills. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is the overall architecture schematic diagram of the rhinoplasty surgery navigation system based on three-dimensional reconstruction of the present application;

[0053] Figure 2 is the structure schematic diagram of the data acquisition module of the present application;

[0054] Figure 3This is a schematic diagram of the workflow of the manifold characterization module of the present invention;

[0055] Figure 4 This is a schematic diagram of the workflow of the transformation calculation module of the present invention;

[0056] Figure 5 This is a schematic diagram of the error optimization module of the present invention;

[0057] Figure 6 This is a schematic diagram of the user interface of the navigation display module of the present invention. Detailed Implementation

[0058] Please refer to the attached document. Figures 1-6 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0059] like Figure 1 As shown, the 3D reconstruction-based rhinoplasty surgical navigation system provided by this invention includes a data acquisition module 1, a manifold representation module 2, a transformation calculation module 3, an error optimization module 4, and a navigation display module 5. Data flow paths are formed between these modules, realizing a complete workflow from medical image acquisition to surgical navigation display. Furthermore, the system may also include a surgical planning module 6 and an effect evaluation module 7 to further enhance the system's functional completeness.

[0060] In one embodiment of the present invention, such as Figure 2 As shown, the data acquisition module 1 includes a medical image processing unit 11, a point cloud generation unit 12, and a real-time scanning unit 13.

[0061] The medical image processing unit 11 receives at least one type of medical image data, including CT and MRI, and performs noise reduction and image enhancement processing. Preferably, for CT data, an adaptive thresholding method is used for image segmentation, with the threshold set between -600 and 400 HU, adjusted according to the density characteristics of different tissues. For MRI data, a histogram equalization method is used for contrast enhancement to improve the display effect of soft tissue details. Furthermore, a Gaussian filter is applied for noise reduction, with the kernel size preferably set between 3×3 and 5×5, dynamically adjusted according to the image noise level.

[0062] The point cloud generation unit 12 is connected to the medical image processing unit 11 and is used to convert the processed medical image data into three-dimensional point cloud data. Preferably, the Marching Cubes algorithm is used for surface reconstruction, extracting isosurfaces and generating a triangular mesh model. Subsequently, the triangular mesh is converted into a point cloud representation by sampling the mesh vertices. The sampling density can be dynamically adjusted according to the importance of the region. The sampling density of key anatomical regions (such as the tip and alar of the nose) is set to 0.2 mm to 0.5 mm, while other regions can be appropriately reduced to 0.5 mm to 1.0 mm to achieve fine representation of key regions.

[0063] The real-time scanning unit 13 is used to acquire real-time three-dimensional data of the nose during the operation using an optical scanning device, generating intraoperative point cloud data. Preferably, a depth camera using structured light or time-of-flight (ToF) technology is used, with a scanning frequency of not less than 30Hz to ensure the real-time nature of the data. The scanning accuracy is preferably between 0.1mm and 0.3mm to meet the high precision requirements of rhinoplasty.

[0064] In another embodiment of the invention, such as Figure 3 As shown, the manifold representation module 2 includes a manifold mapping unit 21, a geometric feature analysis unit 22, and a feature point extraction unit 23.

[0065] The manifold mapping unit 21 is used to construct a multi-resolution point cloud data structure, mapping the 3D point cloud data to a Riemannian manifold space. Specifically, an octree structure is first used to spatially organize the point cloud, supporting data access at different resolutions. The maximum depth of the octree is set to 8 to 12 layers, dynamically adjusted according to the point cloud density. Then, a tangent plane coordinate system is constructed for each local region to achieve local-to-global coordinate transformation. Finally, a parameterized representation of the point cloud is established, mapping the 3D point cloud to a 2D parameter space, facilitating subsequent curvature calculation and feature extraction.

[0066] The geometric feature analysis unit 22 is connected to the manifold mapping unit 21 and is used to calculate the curvature features of the point cloud at different scales, generating a curvature variation map. In implementation, this unit calculates the principal curvature, Gaussian curvature, and mean curvature of the point cloud, constructs a curvature variation map, and identifies anatomical feature regions such as the bridge of the nose, the tip of the nose, and the alar of the nose. For each point... Its principal curvature and The calculation can be performed in the following ways:

[0067] First, estimate the point. A tangent plane within the local neighborhood of a given point. The neighborhood point set is The center point is Then the covariance matrix It can be represented as:

[0068] ,

[0069] in, The number of neighboring points is usually set to 20 to 30 nearest neighbors; For the first in the neighborhood Each point is a three-dimensional spatial coordinate vector; The center point of the neighborhood is equal to For point Position vector relative to the center point; This is the transpose of the position vector; This represents the summation operation over all neighboring points; for The covariance matrix describes the spatial distribution characteristics of a point set.

[0070] By analyzing the covariance matrix Perform eigenvalue decomposition to obtain three eigenvalues. and its corresponding eigenvectors .in Approximate as a point The normal vector at that point is a unit three-dimensional vector. It is a real number representing the variance of the point cloud along the direction of the corresponding eigenvector; They are mutually orthogonal unit three-dimensional vectors.

[0071] Based on local surface fitting, the points can be calculated. Principal curvature at and Gaussian curvature Mean curvature .in, and It is a real number representing the degree of curvature of the surface in the two principal directions at that point; Gaussian curvature, in mm -2 H represents the mean curvature, in mm. -1 ; This indicates a product operation.

[0072] In addition, this unit constructs a geodesic distance field, analyzes the topology of the point cloud, and extracts feature lines of the nose shape, including the nasal dorsum line and the nasal alar contour line. The geodesic distance is calculated using the FastMarching Method, starting from a seed point and gradually expanding to the entire point cloud region. The seed point is usually selected as a point at the root of the nose, which helps to generate a geodesic distance field centered on the root of the nose, facilitating subsequent analysis of the overall topology of the nose.

[0073] The feature point extraction unit 23 is connected to the geometric feature analysis unit 22 and is used to identify anatomical landmarks such as the nasal tip, nasal alar, and nasal dorsum based on curvature variation maps, generating a set of feature points. In specific implementations, the nasal tip is usually a local extremum point in a high curvature region, which can be located by finding the local maximum value of the average curvature. Typically, its average curvature value is around 0.15 mm. -1 up to 0.25mm -1 Within the specified range, the alar point is located at the outermost position of the alar contour line and can be determined by analyzing the curvature changes and spatial position. The nasal dorsum line is extracted by tracing the curvature ridge line between the nasal root and the nasal tip.

[0074] To improve the stability of feature points, the positions of candidate feature points at different resolutions are evaluated, and a stability score is calculated. :

[0075] ,

[0076] in, For point The stability score has a range of [0,1]. The feature point positions at the original resolution are represented by a three-dimensional spatial coordinate vector. For the first The location of a feature point at each resolution level is a three-dimensional spatial coordinate vector. This represents the feature scale at this level, indicating the neighborhood radius considered for the feature point, in mm. The number of resolution levels is typically set to 3 to 5. Represents the Euclidean distance, i.e. ; This represents the summation over all resolution levels. Stability score. The value range is [0,1], and a larger value indicates a more stable feature point. In practical applications, feature points with a stability score greater than 0.8 are usually selected for subsequent registration.

[0077] In yet another embodiment of the invention, such as Figure 4 As shown, the transformation calculation module 3 includes a correspondence establishment unit 31, a transformation model construction unit 32, and a deformation field optimization unit 33.

[0078] The correspondence establishment unit 31 is used to calculate feature descriptors based on geometric feature points and establish an initial correspondence between the source point cloud and the target point cloud. In the specific implementation, this unit adopts a hierarchical matching strategy, first establishing a correspondence between significant anatomical landmark points, then extending it to secondary feature points, and finally extending it to the ordinary point cloud region through interpolation.

[0079] Preferably, for each feature point Constructing multi-scale local descriptors It contains geometric and topological information:

[0080] ,

[0081] in, For point The feature descriptor is a multi-dimensional vector; For point The average curvature, in mm -1 ; For point Gaussian curvature, in mm -2 ; Point of A set of average curvature statistics for the neighborhood of each concentric ring, each Corresponding to the The average curvature of the concentric rings; Point of A set of Gaussian curvature statistics for the neighborhood of each concentric ring, each Corresponding to the The mean Gaussian curvature of a concentric ring; For point arrive The set of geodesic distances to key anatomical landmarks, each Indicates up to the The geodetic distance of each marker point is in mm. It is usually set to 3 to 5, indicating the number of concentric rings; It is usually set to 5 to 8, which represents the number of key markers.

[0082] The similarity between feature points is calculated using the weighted Euclidean distance of the descriptors:

[0083] ,

[0084] in, For feature points and The smaller the value, the more similar they are. and Feature points and The descriptor number Dimensional components; represents the weight of the corresponding dimension, indicating the importance of that dimension in similarity calculation; dim is the dimension of the descriptor, equal to ; This represents summing over all dimensions; This represents the square root operation. Preferably, the weight of the curvature-related dimension is set to 0.6 to 0.8, and the weight of the geodesic distance-related dimension is set to 0.2 to 0.4.

[0085] A bidirectional nearest neighbor matching strategy is used to ensure the consistency of correspondence, i.e., point... The nearest neighbor is a point. At the same time The nearest neighbor is also a point In addition, geometric consistency constraints are applied to filter out incorrect matching point pairs, thereby improving the matching quality.

[0086] The transformation model construction unit 32 is connected to the correspondence establishment unit 31 to divide the nasal point cloud into rigid and non-rigid regions, constructing a hybrid transformation model. In the specific implementation, based on anatomical structural characteristics, regions such as the nasal bone are marked as rigid regions, while regions such as cartilage and soft tissue are marked as non-rigid regions. The division of rigid regions can be based on thresholds of Gaussian curvature and average curvature, typically with an absolute value of Gaussian curvature greater than 0.05 mm. -2 And the absolute value of the average curvature is less than 0.1 mm. -1 The region is considered a rigid region.

[0087] The hybrid transformation model comprises two parts: a global rigid transformation and local non-rigid deformation. The global transformation employs a similarity transformation model, including rotation, translation, and scale parameters.

[0088] ,

[0089] in, For point The position after global transformation is a three-dimensional spatial coordinate vector; A point in the source point cloud is a three-dimensional spatial coordinate vector. for A rotation matrix describes a spatial rotation transformation. for Translation vector, describing spatial translation transformation; The scale factor is a positive real number, usually close to 1, ranging from 0.95 to 1.05.

[0090] Local deformation is represented by a spatial deformation field, which is suitable for the elastic deformation characteristics of soft tissue. Preferably, a radial basis function (RBF) network is used to represent the deformation field:

[0091] ,

[0092] in, For point The deformed vector is a three-dimensional vector; For the first Each control point is a three-dimensional spatial coordinate vector; The corresponding weight vector is a three-dimensional vector. These are radial basis functions; Number of control points; Point To the control point The Euclidean distance; This represents the summation over all control points. Commonly used radial basis functions include Gaussian functions and multiple quadratic functions. In this embodiment, Gaussian radial basis functions are used:

[0093] ,

[0094] in, The output value of the radial basis function; Spatial distance, in mm; The radius of influence is measured in mm and is typically set to 5 mm to 15 mm, dynamically adjusted according to the size of the local area. The base of the natural logarithm is approximately 2.71828; This represents the negative half of the normalized squared value of the distance. Control points. Preferably placed at anatomical feature points and high curvature areas, the number of control points... The number is usually set to 30 to 50 to balance the degrees of freedom of deformation and computational complexity.

[0095] The deformation field optimization unit 33 is connected to the transformation model construction unit 32, and is used to optimize the deformation field parameters by applying tissue elasticity constraints and volume preservation constraints. In the specific implementation, a variety of constraints are introduced to make the deformation more consistent with the physical characteristics of biological tissues.

[0096] The elastic properties of tissues constrain the physical deformation behavior of cartilage and soft tissues, which can be expressed as a deformation energy minimization problem:

[0097] ,

[0098] in, This represents the elastic energy of deformation; a smaller value indicates a more natural deformation. The deformation field is a three-dimensional vector field function; The first derivative (gradient) of the deformable field is a The Jacobian matrix; The second derivative of the deformable field (the Hessian matrix) is a tensor; This is the transpose of the gradient matrix; This is the transpose of the Hessian matrix; The trace operation represents the sum of the elements on the main diagonal of a matrix. and These are weighting coefficients, controlling the elasticity and stiffness of the deformation respectively; both are positive real numbers. Set to 0.5 to 1.0. Set to 0.1 to 0.3; The spatial domain in which the point cloud resides; It represents the integral over the entire spatial domain.

[0099] Volume preservation constraints ensure that the local volume remains essentially unchanged during deformation, preventing non-physical deformation.

[0100] ,

[0101] in, To maintain volume constraints; Represents determinant operations; The gradient of the deformation field is a Jacobian matrix; for identity matrix; This represents the local volume ratio after deformation, and ideally should be close to 1. Represents the squared error of volume change; It represents the integral over the entire spatial domain.

[0102] Boundary continuity constraints ensure a natural transition of deformation at the interface between soft and hard tissues:

[0103] ,

[0104] in, For boundary continuity constraints; For the boundary point set of soft and hard tissues; For point The actual deformation vector; The ideal deformation vector of the boundary points inferred based on hard tissue transformation; This represents the squared Euclidean distance between the actual deformation and the ideal deformation; This represents summation over all boundary points.

[0105] Considering the above constraints, the objective function for optimizing the deformation field is:

[0106] ,

[0107] in, The overall optimization objective function; For each data item, measure the distance error between corresponding points; For flexible constraint terms; This is a volume preservation constraint term; For boundary continuity constraints; , and These are weighting coefficients, all positive real numbers, controlling the importance of each constraint term. In practical applications, Set to 0.5 to 1.0. Set to 0.3 to 0.7. Set it to 0.2 to 0.5.

[0108] In yet another embodiment of the invention, such as Figure 5 As shown, the error optimization module 4 includes an error measurement unit 41, an error distribution analysis unit 42, and an iterative optimization unit 43.

[0109] Error metric unit 41 is used to calculate point-to-surface distance, normal vector consistency, and curvature similarity, constructing a multi-dimensional error evaluation system. In the specific implementation, the point-to-surface distance... Calculate the spatial positional deviation between corresponding points:

[0110] ,

[0111] in, For point Time The distance between the planes, in mm; A point in the source point cloud is a three-dimensional spatial coordinate vector. The point corresponding to the target point cloud is a three-dimensional spatial coordinate vector. For point The normal vector at that point is a unit three-dimensional vector. This represents the dot product operation; This indicates taking the absolute value.

[0112] Normal vector consistency Detect the degree of surface orientation matching:

[0113] ,

[0114] in, For point and points The inconsistency of the normal vector has a range of [0,1]. and Points and points The normal vectors at each location are all unit three-dimensional vectors. Represents the dot product of two normal vectors; This represents the absolute value of the dot product. When the two normal vectors are perfectly parallel, The value is 0; when the two normal vectors are perpendicular, The value is 1.

[0115] Curvature similarity Assess the quality of local shape matching:

[0116] ,

[0117] in, For point and points Dissimilarity at curvature; and Points and points The average curvature at that point, in mm. -1 ; and Points and points Gaussian curvature at the point, in mm -2 ; and These are weighting coefficients, all of which are positive real numbers, typically... Set to 0.6 to 0.8. Set to 0.2 to 0.4; Represents the absolute value of the mean curvature difference; This represents the absolute value of the Gaussian curvature difference.

[0118] Comprehensive error The weighted sum of the above three types of errors:

[0119] ,

[0120] in, For point and points The combined error between them; Distance from a point to a surface; The normal vectors are inconsistent; For curvature dissimilarity; , and These are weighting coefficients, all positive real numbers, which can be dynamically adjusted according to the application scenario. In rhinoplasty surgery navigation, they are typically... Set to 0.5 to 0.7. Set to 0.2 to 0.3. Set it to 0.1 to 0.2.

[0121] Error distribution analysis unit 42 is connected to error measurement unit 41 and is used to generate an error distribution map and identify high-error areas. In specific implementation, a three-dimensional error distribution map is constructed to visually display the spatial distribution of registration quality. Error magnitude is represented by color coding; for example, a gradient from blue to red is used to represent error values ​​from low to high. Furthermore, focused error analysis is performed on key aesthetic areas such as the nasal tip, nasal alar, and nasal bridge, as these areas have a decisive impact on the aesthetic outcome of rhinoplasty.

[0122] By analyzing the spatial distribution and statistical characteristics of errors, regions of systematic deviation can be identified, providing direction for subsequent optimization. For example, if high errors are prevalent in the nasal alar region, it may be necessary to adjust the deformation parameters of that region or increase the density of control points.

[0123] The iterative optimization unit 43 is connected to the error distribution analysis unit 42 and is used to adaptively adjust the registration parameters based on the error distribution to implement multi-level iterative optimization. In specific implementation, an iterative optimization strategy from coarse to fine is designed to gradually improve the registration accuracy. Preferably, a regional optimization mechanism is adopted, allowing different optimization parameters to be used in different regions, especially for aesthetically critical regions such as the tip of the nose, the alar of the nose, and the bridge of the nose, where stricter error thresholds and higher optimization weights are used.

[0124] For regions with high errors, the density of local deformation control points can be increased, or the parameters of the deformation model can be adjusted. For example, in regions with large errors, the spacing between control points can be reduced to 50% to 70% of its original size, increasing the degrees of freedom of local deformation. Simultaneously, the weights of elastic and volumetric constraints can be dynamically adjusted to improve registration accuracy while ensuring the naturalness of deformation.

[0125] The convergence criteria for the optimization process employ multiple conditions, including average error, maximum error, and error rate of change. Optimization is considered convergent when the average error is less than 0.5 mm, the maximum error is less than 1.0 mm, and the average error rate of change between two consecutive iterations is less than 1%. Furthermore, to avoid overfitting due to excessive iteration, the maximum number of iterations is set to 20 to 30.

[0126] In yet another embodiment of the invention, such as Figure 6 As shown, the navigation display module 5 includes an augmented reality display unit 51, an error visualization unit 52, and an interactive control unit 53.

[0127] Augmented reality display unit 51 is used to convert the optimized spatial mapping relationship into visual guidance information, presenting the superimposed effect of the virtual model and the actual surgical scene on the head-mounted display device. In the specific implementation, the virtual model adopts a semi-transparent rendering method, and the transparency can be set from 30% to 60%, allowing doctors to see both the virtual plan and the actual tissue simultaneously. The edges of the virtual model can be emphasized using high-contrast colors (such as yellow or green) to improve visual recognition.

[0128] The error visualization unit 52 is connected to the augmented reality display unit 51 and is used to display the registration error distribution through color coding. In the specific implementation, a heatmap color scheme from blue to red is used to represent error values ​​from low to high. The error threshold can be divided into multiple levels; for example, errors less than 0.5mm are displayed in blue, 0.5mm to 1.0mm in green, 1.0mm to 2.0mm in yellow, and errors greater than 2.0mm in red. This visualization method intuitively conveys registration quality information to the doctor, helping them make reasonable judgments during the surgery.

[0129] The interactive control unit 53 receives the doctor's operating instructions and adjusts the navigation view and display parameters. In its implementation, it provides multiple interaction methods, including voice control, gesture recognition, and a touch interface. The doctor can adjust parameters such as view angle, zoom level, displayed content, and transparency using simple commands. For example, voice commands such as "zoom in," "rotate," or "switch views" can control the display effect without interrupting the surgical procedure. Furthermore, it provides surgical progress tracking, displaying the current surgical stage and completion status, assisting the doctor in controlling the overall surgical process.

[0130] In another embodiment of the present invention, the system further includes a surgical planning module 6 and an effect evaluation module 7.

[0131] The surgical planning module 6 connects with the data acquisition module 1 and the navigation display module 5 to develop a surgical plan based on the patient's 3D point cloud data before surgery, generating a virtual planning model. In practice, doctors can interactively adjust the shape of the nose, such as changing the height of the nasal bridge, adjusting the angle of the nasal tip, and modifying the width of the nasal wings. The system provides various reference templates and aesthetic indicators to assist doctors in developing personalized plans that meet the patient's facial features and aesthetic needs.

[0132] Preferably, during the surgical planning process, the target locations and morphological parameters of key anatomical landmarks can be set, such as the curvature of the nasal dorsum line, the angle of the nasolabial angle (usually set to 90° to 110°), and the nasal tip projection (usually 10mm to 14mm). Based on these parameters, the system automatically generates a virtual planning model, which can be rendered in 3D and previewed from multiple angles to help doctors evaluate the surgical outcome.

[0133] The effect evaluation module 7 is connected to the navigation display module 5 and is used to compare the current surgical status with the virtual planning model in real time during the operation, calculate the surgical progress and completion indicators, and provide feedback on the surgical effect to the doctor. In the specific implementation, a series of evaluation indicators are defined, including anatomical landmark deviation, contour matching degree, and region shape similarity.

[0134] Anatomical landmark deviation calculation: Euclidean distance between the current position and the target position.

[0135] ,

[0136] in, The deviation of the i-th marker point is expressed in mm. Let be the current position of the i-th marker, which is a three-dimensional spatial coordinate vector; Its target position is a three-dimensional spatial coordinate vector; Represents the Euclidean distance, i.e. .

[0137] Contour line matching assessment evaluates shape differences in key contours (such as the nasal dorsum line):

[0138] ,

[0139] in, The deviation in the matching degree of contour line C is expressed in mm. The current contour line is a set of points in three-dimensional space; The target outline is also a set of three-dimensional points; From point p to the target contour line The shortest distance, in mm; The number of points on the outline C; This represents the summation over all points on the contour line C; This indicates calculating the average value. Regional shape similarity is assessed by calculating differences in local curvature distributions.

[0140] ,

[0141] in, For the shape similarity deviation of region R; The region of interest is a set of points in three-dimensional space. and Let P be the current mean curvature and Gaussian curvature at point p, in mm. -1 and mm -2 ; and These are the corresponding target values, with the same units; The weighting coefficient is a positive real number, typically set to 0.2 to 0.4. Represents the absolute value of the mean curvature difference; Represents the absolute value of the Gaussian curvature difference; This represents summing over all points in region R; This indicates calculating the average value.

[0142] Based on the above indicators, the system calculates the overall completion score and displays it in real time on the navigation interface to help doctors understand the progress and effectiveness of the surgery. When all indicators reach the preset thresholds (such as deviation less than 1.0mm and similarity greater than 90%), the system will indicate that the surgical goal has been basically achieved.

[0143] The specific embodiments of the various modules and functional units of the present invention have been described in detail above. It should be noted that the present invention is not limited to the above embodiments. Those skilled in the art can make various changes and modifications to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the protection scope of the present invention.

Claims

1. A nasal surgery navigation system based on three-dimensional reconstruction, characterized in that, include: The data acquisition module is used to acquire medical imaging data of the patient's nose and construct three-dimensional point cloud data; The manifold representation module, connected to the data acquisition module, is used to map the three-dimensional point cloud data to the Riemannian manifold space and extract the geometric feature points of the nasal anatomical structure. The transformation calculation module, connected to the manifold representation module, is used to receive the geometric feature points, construct a hybrid transformation model that includes global rigid transformation and local non-rigid deformation, and generate a spatial mapping relationship between the virtual model and the real-time acquired data. An error optimization module, connected to the transformation calculation module, is used to calculate the geometric error distribution of the spatial mapping relationship, construct an error metric function based on differential invariants, and achieve adaptive optimization. The system also includes a navigation display module connected to the error optimization module. This module receives the optimized spatial mapping relationship and displays the correspondence between the virtual planning model and the actual operation in the surgical area in real time, guiding the doctor to perform precise surgical operations.

2. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 1, characterized in that, The data acquisition module includes: A medical image processing unit is used to receive at least one type of medical image data, including CT and MRI, and to perform noise reduction and image enhancement processing. A point cloud generation unit, connected to the medical image processing unit, is used to convert the processed medical image data into three-dimensional point cloud data. It also includes a real-time scanning unit, which is used to acquire real-time three-dimensional data of the nose through optical scanning equipment during surgery and generate intraoperative point cloud data.

3. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 1, characterized in that, The manifold characterization module includes: Manifold mapping unit is used to construct multi-resolution point cloud data structures and map 3D point cloud data to Riemannian manifold space; The geometric feature analysis unit, connected to the manifold mapping unit, is used to calculate the curvature features of the point cloud at different scales and generate curvature variation maps. The system also includes a feature point extraction unit, which is connected to the geometric feature analysis unit and is used to identify anatomical landmarks such as the tip of the nose, the ala of the nose, and the bridge of the nose based on the curvature change map, and to generate a set of feature points.

4. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 3, characterized in that, The geometric feature analysis unit is also used for: Calculate the principal curvature, Gaussian curvature, and mean curvature of the point cloud; Construct a geodesic distance field and analyze the topology of the point cloud; It also extracts feature lines of the nose shape, including the nasal dorsum line and the nasal alar contour line.

5. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 1, characterized in that, The transformation calculation module includes: The correspondence establishment unit is used to calculate feature descriptors based on the geometric feature points and establish an initial correspondence between the source point cloud and the target point cloud; A transformation model construction unit, connected to the corresponding relationship establishment unit, is used to divide the nose point cloud into rigid and non-rigid regions and construct a hybrid transformation model. And a deformation field optimization unit, connected to the transformation model construction unit, is used to apply tissue elasticity constraints and volume retention constraints to optimize deformation field parameters.

6. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 5, characterized in that, The correspondence establishment unit adopts a hierarchical matching strategy, including: First, establish correspondences between significant anatomical landmarks; Then extend to secondary feature points; And finally, it is extended to the ordinary point cloud region through interpolation.

7. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 1, characterized in that, The error optimization module includes: An error metric unit is used to calculate the distance from a point to a surface, the consistency of normal vectors, and the curvature similarity, thereby constructing a multi-dimensional error evaluation system. An error distribution analysis unit, connected to the error measurement unit, is used to generate an error distribution map and identify high error regions. An iterative optimization unit, connected to the error distribution analysis unit, is used to adaptively adjust the registration parameters based on the error distribution and implement multi-level iterative optimization.

8. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 7, characterized in that, The iterative optimization unit employs stricter error thresholds and higher optimization weights for key aesthetic areas such as the tip, wings, and bridge of the nose.

9. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 1, characterized in that, The navigation display module includes: Augmented reality display unit is used to convert the optimized spatial mapping relationship into visual guidance information, and to present the superimposed effect of virtual model and actual surgical scene on head-mounted display device; An error visualization unit, connected to the augmented reality display unit, is used to display the registration error distribution through color coding. And an interactive control unit, used to receive the doctor's operating instructions and adjust the navigation view and display parameters.

10. The nasal plastic surgery navigation system based on three-dimensional reconstruction according to claim 1, characterized in that, Also includes: The surgical planning module, connected to the data acquisition module and the navigation display module, is used to formulate a surgical plan based on the patient's three-dimensional point cloud data before surgery and generate a virtual planning model. It also includes an effect evaluation module, which is connected to the navigation display module and is used to compare the current surgical status with the virtual planning model in real time during the operation, calculate the surgical progress and completion indicators, and provide feedback on the surgical effect to the doctor.

Images