A Cesarean Section Surgery Training Method and System Based on Mixed Reality

Abstract

The present invention discloses a cesarean section surgery training method and system based on mixed reality. The method includes: acquiring medical image data of a pregnant woman's pelvis and abdomen, identifying and marking the boundaries and regions of key organizational structures, constructing a three-dimensional model of the key organizational structures, generating key indicators for cesarean section surgery training, importing the constructed three-dimensional model into the MR system, collecting motion data, generating real-time feedback prompts, receiving a training end signal, generating an evaluation result, recording operations that do not meet the evaluation criteria, and providing targeted improvement suggestions for the trainees. The present invention can effectively improve the quality and efficiency of surgery training. Through the precise processing of medical images, the efficient reconstruction of three-dimensional models, real-time feedback, and personalized operation evaluation, it not only provides a safe, efficient, and immersive training platform for trainees, but also can provide personalized guidance according to their performance, further improving the training effect of surgical skills.

Description

Technical Field

[0001] The present invention relates to the field of mixed reality technology, and in particular, to a cesarean section surgery training method and system based on mixed reality.

Background Art

[0002] With the continuous development of medical technology, cesarean section surgery, as a common and important obstetrics and gynecology surgery, is widely used to solve the problems of the lives of pregnant women and fetuses. However, the complexity of cesarean section surgery requires doctors to have a high level of professional skills and rich experience. Traditional surgical training methods mainly rely on actual operations at the surgical site. Limited by factors such as training time, individual differences of patients, and operation risks, they cannot provide sufficient practice opportunities.

[0003] In recent years, mixed reality (MR) technology has achieved remarkable applications in medical education. Mixed reality technology can integrate the virtual environment with the real world, providing trainees with a highly immersive and interactive training experience, which is particularly suitable for the simulation training of complex surgeries. Nevertheless, existing mixed reality surgical training methods still face many challenges, such as how to accurately extract the key organizational structures in the image data, how to ensure the authenticity and safety of the training process, and how to provide real-time feedback and evaluation on the operations of the trainees.

[0004] Therefore, there is an urgent need for a cesarean section surgery training method and system based on mixed reality, which can help trainees quickly master surgical skills and improve the training effect through accurate image data analysis, real-time operation feedback, and evaluation.

Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides a cesarean section surgery training method and system based on mixed reality.

[0006] In a first aspect, an embodiment of the present invention provides a cesarean section surgery training method based on mixed reality, the method comprising:

[0007] S1. Obtain the medical image data of the pregnant woman's pelvis and abdomen, extract the characteristic values of the key organizational structures after preprocessing, and identify and mark the boundaries and regions of the key organizational structures;

[0008] S2. Initially segment the key organizational structures from the image data, then perform fine segmentation, and use a reconstruction algorithm to construct a three-dimensional model of the key organizational structures;

[0009] S3. Calculate the reference evaluation value for cesarean section surgery training based on the characteristic values of the key organizational structures, and generate the key indicators for cesarean section surgery training;

[0010] S4. Convert the constructed 3D model into a format recognizable by the mixed reality system for importing into the MR system, and perform scene setting and spatial registration;

[0011] S5. Receive a training start signal, detect the surgical operations of the trainee to collect action data, compare the operation data of the trainee with key indicators, generate real-time feedback prompts, receive a training end signal, and generate an evaluation result;

[0012] S6. Record the operations that do not meet the evaluation criteria, provide targeted improvement suggestions for the trainee, and update and generate the score curve of the trainee.

[0013] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. Specifically, S1 includes:

[0014] Obtain the 3D volume data D(x, y, z) of the pregnant woman's pelvis and abdomen, where x, y, and z represent spatial coordinates, and D represents the intensity value of each voxel;

[0015] Apply the anisotropic diffusion filter F AD Reduce noise to obtain D f (x, y, z), perform intensity normalization on D f (x, y, z) to obtain D n (x, y, z), and use histogram equalization to enhance the contrast to obtain D e (x, y, z);

[0016] Extract features from the preprocessed image data D e (x, y, z). Specifically: calculate the gray histogram, mean, and variance statistical features of the image, use the Sobel operator to calculate the gradient map of the image to obtain edge information, and use the gray-level co-occurrence matrix to extract texture features;

[0017] Determine the regions where the key organizational structures are located according to the extracted features. Among them, the list of key organizational structures: uterus (U): including the uterine wall and uterine cavity, placenta (P): location and scope, fetal head (F): position and orientation, bladder (B): avoid injury during surgery, abdominal wall layer (A): skin, subcutaneous tissue, fascia, and muscle;

[0018] Perform non-linear enhancement on the gradient map to enhance the edge information, and combine the enhanced edge information and the regions of the key organizational structures to determine the region of interest.

[0019] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. Specifically, S2 includes:

[0020] Use the determined region of interest as the initial condition for initial segmentation to generate a preliminary segmentation mask;

[0021] Select the segmentation algorithm again according to the organizational structure characteristics, iteratively solve to generate an accurate segmentation mask, and post-process the segmentation results;

[0022] Use the Marching Cubes algorithm to convert the segmentation mask into a three-dimensional model of the key organizational structure and optimize the model.

[0023] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. Specifically, S3 includes:

[0024] Calculate the volume, surface area, and centroid position of the key organizational structure, extract the evaluation of tissue compactness, and calculate the distance and relative position between tissues;

[0025] Calculate the incision position and the reference value of the operating force, and set the reference value of the operation time;

[0026] Generate the key indicators for surgical training: the allowable value of the incision position error, the range of the operating force, and the limit of the operation time.

[0027] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. Specifically, S4 includes:

[0028] Convert the constructed three-dimensional model from the original format to the target format according to the requirements of the mixed reality system MR, and convert the coordinate system of the three-dimensional model into the coordinate system used by the MR system;

[0029] Build a virtual operating room scene in the MR system, including the following elements: operating table: set the size and position of the operating table; medical devices: surgical instrument models; lighting settings: configure ambient light and spotlight sources; simulate the lighting conditions of the operating room; other scene elements: walls and floors;

[0030] Import the converted three-dimensional model into the virtual scene of the MR system, and adjust the position, rotation, and scaling parameters of the model to match the virtual operating room environment;

[0031] Design a user interface in the MR system, including operation guides, information displays, and feedback prompts. The operation guides are used to display surgical steps and precaution information. The information displays are used to display the operation data and evaluation results of the trainees in real time. The feedback prompts are used to provide visual, auditory, or tactile feedback to indicate operation errors or successes;

[0032] Set interaction methods in the MR system, including gesture recognition, voice control, and handle operations. The gesture recognition is used to support gesture control of the movement and rotation of the model. The voice control is used to interact with the system through voice commands. The handle operations are used to interact using a handle or a controller;

[0033] Set rendering effects in the MR system, including material properties, lighting effects, shadows, reflections, and texture mapping. The material properties are used to set the specular and transparency material parameters of the model. The lighting effects are used to adjust the intensity, color, and position of the light source. The shadows and reflections are used to enable shadow and reflection effects. The texture mapping is used to apply high-resolution texture maps.

[0034] Define the real-space coordinate system: a coordinate system based on the real world, in which all positions and postures in the real environment are represented. Define the virtual-space coordinate system: the coordinate system of the virtual model in the MR system.

[0035] Select a corresponding set of landmark points in the real space and the virtual space respectively, and establish a set of real-space landmark points Establish a set of virtual-space landmark points Use the least squares method to calculate the rigid body transformation matrix T = {R, Τ} such that Calculate the centroids of the virtual and real landmark points Decentralize the coordinates Calculate the covariance matrix Perform singular value decomposition on the covariance matrix H = UΣV Τ , calculate the rotation matrix R = VU Τ , calculate the translation vector V a = R·V v + Τ, calculate the root mean square error where, P i r represents the coordinate of the i-th landmark point in the real space, P i v represents the coordinate of the corresponding i-th landmark point in the virtual space, N represents the number of landmark points, U, Σ, and V represent the SVD decomposition results of the covariance matrix, R represents the rotation matrix, Τ represents the translation vector, V v represents the vertex coordinate matrix of the virtual model, V a represents the vertex coordinate matrix after registration, ||·|| represents the Euclidean distance,; set the registration error threshold. If the root mean square error is less than the registration error threshold, it is considered that the registration is successful; otherwise, re-registration is required.

[0036] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. The S5 specifically includes:

[0037] Configure a sensor device to collect operation data such as the position and force of the trainee during surgical training in real time. The sensor device includes a position sensor and a force sensor. The position sensor is used to capture the position and posture of the trainee's hand and surgical instrument in real time, and the force sensor is used to measure the force exerted by the trainee during the operation;

[0038] Receive a training start signal, preprocess the collected operation data, and extract key features;

[0039] Compare the real-time operation data of the trainee with the reference evaluation value to generate real-time feedback prompts. Set the position deviation warning threshold and set it to 80% of the maximum allowable value. When it is determined that the position deviation exceeds the position deviation warning threshold, trigger a warning, display a warning message on the interface and play a prompt sound, and record the warning event; Set the force deviation warning threshold and set it to 80% of the maximum allowable value. When it is determined that the force deviation exceeds the force deviation warning threshold, trigger a warning, display a warning message on the interface and play a prompt sound, and record the warning event;

[0040] Receive a training end signal, calculate and display the total surgical training score S of the trainee on the interface 总 , the total surgical training score S 总 The calculation formula is:

[0041] S 总 =w 1 S p +w 2 S f +w 3 S t , w 1 +w 2 +w 3 =1,

[0042]

[0043]

[0044] where, w 1 , w 2 and w 3 represent weight coefficients, S p represents the incision position score, S f represents the operation force score, S t represents the surgical time score, ΔP represents the Euclidean distance between the actual incision position of the trainee and the reference incision position, δF represents the allowable force deviation, F a represents the average force exerted by the trainee during the operation, F r represents the reference force, T u represents the actual time taken by the trainee to complete the surgery, δT represents the allowable time deviation, Tr Indicates the reference time;

[0045] Set the passing score threshold S 1 and set the excellent score threshold S 2 If S 总 > S 2 then it is evaluated as "excellent". If S 1 ≤ S 总 < S 2 then it is evaluated as "qualified". If S 总 < S 1 then it is evaluated as "unqualified";

[0046] Generate and display the evaluation result.

[0047] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. The S6 specifically includes:

[0048] Record the hand or instrument position trajectory of the trainee during the entire training process, record the change of the operation force applied by the trainee over time, record the time of each data point, and save the data in a structured format;

[0049] Receive the training end signal, compare the operation data of the trainee with the key indicators to analyze the reasons for errors;

[0050] Classify the types of operation errors of the trainee during the training process, including positioning deviation errors where the position deviation exceeds the position deviation warning threshold, force control errors where the force deviation exceeds the force deviation warning threshold, operation process errors of surgical steps errors, and time errors exceeding the time threshold, and calculate the number of times of different operation error types of the trainee during the training process;

[0051] Provide a positioning practice module for positioning deviation errors, including positioning tasks of different difficulties;

[0052] Provide a force practice module for force control errors, including simulation tissue practice tasks of different hardnesses;

[0053] Provide a video module for operation process errors and time errors, including learning tasks of the standard cesarean section surgical process;

[0054] Based on the trainee's established ID file, take the total score of each surgical training of the corresponding trainee as a data point, and update and generate the score curve of the trainee.

[0055] For the aspects and any possible implementation manners as described above, a further implementation manner is provided. The method further includes:

[0056] Set an initial warning threshold D for position deviation at the initial stage of training p0The initial warning threshold D for the sum operation intensity f0 ;

[0057] According to the set total training time length T t The training process is divided into multiple time periods, and the length of each time period is T s , at the end of each time period, judge the operation compliance rate of the trainee during this time period. If the operation compliance rate reaches the preset standard, update the warning threshold and enter the next time period;

[0058] Update the warning threshold using the dynamic warning threshold update formula;

[0059] During each time period, the operation deviation of the trainee is detected in real time. If the position intensity deviation exceeds the position deviation warning threshold D p (t) during this time period, or the intensity deviation exceeds the intensity deviation warning threshold D f (t) during this time period, immediately issue a feedback prompt sound, highlight the specific value of the deviation on the user interface, and at the same time suggest that the trainee adjust the operation. If the number of times the deviation exceeds the threshold in a certain time period exceeds 3 times, the system will record this time period as unqualified;

[0060] If the operation compliance rate of consecutive N 1 time periods is higher than the first preset compliance rate Q 1 , then adjust the attenuation rate constant of the threshold in the dynamic warning threshold update formula, that is, increase λ and γ by 20%. If the operation compliance rate of consecutive N 2 time periods is lower than Q 2 , then relax the adjustment rate of the threshold, that is, decrease λ and γ by 10% respectively, where N 1 > N 2 and Q 1 > Q 2 ;

[0061] At 80% of the total training time T t , reduce λ and γ to 50% of their initial values respectively;

[0062] At the end of the training, the system generates an operation stability report based on the deviation records throughout the training, and statistically calculates the average value, standard deviation, and compliance rate of the position and intensity deviations.

[0063] For the aspects and any possible implementation manners described above, a further implementation manner is provided. The dynamic warning threshold update formula includes a dynamic update formula for the position deviation threshold and a dynamic update formula for the intensity deviation threshold, where,

[0064] The dynamic update formula for the position deviation threshold is

[0065] The dynamic update formula of the force deviation threshold is as follows

[0066] where D p (t) represents the position deviation warning threshold at time t, and D f (t) represents the force deviation warning threshold at time t, D p0 represents the initial position deviation warning threshold, D f0 represents the force deviation warning threshold, k and m represent the dynamic adjustment coefficients of position and force feedback, T t represents the total training time length, T s represents the duration of a single time period, and λ and γ represent the decay rate constants of position and force.

[0067] Second, the embodiment of the present invention provides a cesarean section surgery training system based on mixed reality. The system includes:

[0068] An image data acquisition module for acquiring medical image data of a pregnant woman's pelvis and abdomen;

[0069] A preprocessing module for preprocessing the acquired medical image data;

[0070] A key organizational structure extraction module for extracting the eigenvalue of the key organizational structure, identifying and marking the boundary and area of the key organizational structure;

[0071] A three-dimensional model construction module for initially segmenting the key organizational structure from the image data, then performing fine segmentation, and constructing a three-dimensional model of the key organizational structure using a reconstruction algorithm;

[0072] An index generation module for calculating the reference evaluation value of cesarean section surgery training based on the eigenvalue of the key organizational structure and generating the key index of cesarean section surgery training;

[0073] A mixed reality conversion module for converting the constructed three-dimensional model into a format recognizable by the mixed reality system to be imported into the MR system, and performing scene setting and spatial registration;

[0074] A training analysis module for receiving a training start signal, detecting the surgical operations of the trainee to collect action data, comparing the operation data of the trainee with the key index, generating real-time feedback prompts, receiving a training end signal, and generating an evaluation result;

[0075] An evaluation feedback module for recording the operations that do not meet the evaluation criteria, providing targeted improvement suggestions for the trainee, and updating and generating the score curve of the trainee.

[0076] One of the above technical solutions has the following beneficial effects:

[0077] The present invention provides a cesarean section surgery training method and system based on mixed reality. The present invention can effectively improve the quality and efficiency of surgical training. Through precise processing of medical images, efficient reconstruction of three-dimensional models, real-time feedback, and personalized operation evaluation, it not only provides a safe, efficient, and immersive training platform for trainees, but also can provide personalized guidance according to their performance, further improving the training effect of surgical skills.

BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0079] Figure 1 It is a schematic flowchart of the cesarean section surgery training method S1-S6 based on mixed reality provided by the embodiments of the present invention;

[0080] Figure 2 It is a functional block diagram of the cesarean section surgery training system based on mixed reality provided by the embodiments of the present invention.

DETAILED DESCRIPTION

[0081] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments and the corresponding drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0082] Please refer to Figure 1 , which is a schematic flowchart of a cesarean section surgery training method S1-S7 based on mixed reality provided by the embodiments of the present invention. As Figure 1 shown, the method includes the following steps:

[0083] S1. Obtain the medical image data of the pregnant woman's pelvis and abdomen, extract the characteristic values of the key tissue structures after preprocessing, and identify and mark the boundaries and regions of the key tissue structures;

[0084] S2. Initially segment the key tissue structures from the image data, then perform fine segmentation, and use a reconstruction algorithm to construct a three-dimensional model of the key tissue structures;

[0085] S3. Calculate the reference evaluation value for cesarean section surgery training based on the characteristic values of the key tissue structures, and generate the key indicators for cesarean section surgery training;

[0086] S4. Convert the constructed 3D model into a format recognizable by the mixed reality system for importing into the MR system, and perform scene setting and spatial registration;

[0087] S5. Receive the training start signal, detect the surgical operations of the trainee to collect action data, compare the operation data of the trainee with the key indicators, generate real-time feedback prompts, receive the training end signal, and generate an evaluation result;

[0088] S6. Record the operations that do not meet the evaluation criteria, extract the action feature values to analyze the reasons for errors, provide targeted improvement suggestions for the trainee, and update and generate the score curve of the trainee.

[0089] In the embodiments of the present invention, by acquiring medical image data of a pregnant woman's pelvis and abdomen, and performing effective preprocessing and eigenvalue extraction, key organizational structures can be accurately identified and marked; by adopting specific image processing techniques and steps, effective image features can be extracted from noise, ensuring that the boundaries and regions of key organizational structures can be clearly and accurately demarcated, avoiding errors caused by noise or poor image quality; through initial segmentation and fine segmentation techniques, combined with a reconstruction algorithm, a three-dimensional model of key organizational structures is accurately constructed, which not only shows the geometric shapes of each anatomical structure, but also provides the spatial relationships of the organizational structures, greatly enhancing the realism and operation accuracy of the virtual scene during training. Through fine segmentation and modeling techniques, it is ensured that the geometric shapes and positions of each anatomical structure highly conform to the actual situation; through a surgical training system based on mixed reality, trainees can perform surgical operations in a highly simulated virtual environment, thus effectively avoiding potential safety hazards in traditional surgical training. The high immersion and real-time feedback mechanism provided by the MR system enable trainees to practice complex surgical operations in a risk-free environment, greatly reducing the risks caused by insufficient experience or unskilled skills during actual operations; by comparing the surgical operation data of trainees with key indicators, operation feedback can be generated in real time to help trainees promptly understand the correctness and operation level of their surgical operations. This real-time assessment can provide personalized feedback prompts for the specific operation behaviors of each trainee, thereby improving learning efficiency and avoiding the accumulation of operation errors. It can not only enhance the confidence of trainees, but also strengthen their understanding and mastery of the surgical process; by recording operations that do not meet the assessment criteria and analyzing the reasons for operation failures, targeted improvement suggestions are provided for each trainee. These suggestions can not only help trainees identify deficiencies in operations, but also provide corresponding improvement strategies according to the types of their operation errors. By generating the score curve of trainees, the learning progress of trainees can be clearly reflected. The present invention can help doctors simulate various surgical situations in a non-real surgical scenario by providing a highly simulated and intelligent surgical training platform, improving their actual surgical operation ability. Especially for complex or high-risk surgical operations, doctors can practice repeatedly, accumulate experience, and improve their clinical surgical emergency response ability.

[0090] Among them, acquiring medical image data of a pregnant woman's pelvis and abdomen specifically includes:

[0091] Pregnant women are selected and informed consent is obtained. Their medical imaging data is used for educational and training purposes. Volume data of the pelvis and abdomen is acquired using magnetic resonance imaging (MRI). Imaging parameters are set to ensure that the resolution and contrast are sufficient to distinguish key tissue structures. For example, MRI parameters use T1-weighted and T2-weighted sequences with a slice thickness of less than 1 mm. Three-dimensional volume data D(x, y, z) is obtained, where x, y, and z represent spatial coordinates, and D represents the intensity value of each voxel.

[0092] In a preferred embodiment of the present invention, S1 specifically includes:

[0093] Obtain three-dimensional volume data D(x, y, z) of the pregnant woman's pelvis and abdomen, where x, y, and z represent spatial coordinates, representing the position of the voxel, and D represents the intensity value of each voxel;

[0094] Image data preprocessing:

[0095] Apply the anisotropic diffusion filter F AD Reduce noise to obtain D f (x, y, z), defined as: D f (x, y, z) = F AD [D(x, y, z)], where D f (x, y, z) represents the denoised image data, and F AD represents the anisotropic diffusion filter operator; based on the Perona-Malik model, the anisotropic diffusion equation: where, represents the rate of change of the image intensity with respect to time, represents the gradient magnitude of the image, and k represents the parameter controlling the diffusion rate; set the noise threshold to judge the noise level. If the image noise level exceeds the noise threshold, increase the number of filtering iterations, otherwise use the default parameters;

[0096] Perform intensity normalization on D f (x, y, z) to obtain D n (x, y, z), D n (x, y, z) represents the normalized image data, and the intensity value ranges from 0 to 1. D min and D max represent the minimum and maximum intensity values of the denoised image data; if D min = D max , then the image intensity distribution is abnormal, and the denoising parameters are adjusted;

[0097] Use histogram equalization to enhance the contrast to obtain D e (x, y, z), D e (x, y, z) = HistEqual[Dn (x, y, z)], D e (x, y, z) represents the image data after contrast enhancement, and HistEqual represents the histogram equalization operator;

[0098] For the preprocessed image data D e (x, y, z) for feature extraction, specifically:

[0099] Calculate the gray-level histogram H(g), mean μ, and variance σ of the image 2 Statistical features,

[0100] where g represents the gray level, δ represents the Kronecker delta function, μ represents the mean, N represents the total number of voxels, and σ 2 represents the variance;

[0101] Use the Sobel operator to calculate the gradient map G(x, y, z) of the image to obtain edge information. The gradient calculation is as follows:

[0102]

[0103] G x = D e (x + 1, y, z) - D e (x - 1, y, z),

[0104] G y = D e (x, y + 1, z) - D e (x, y - 1, z),

[0105] G z = D e (x, y, z + 1) - D e (x, y, z - 1),

[0106] Set the gradient threshold. If G(x, y, z) is greater than or equal to the gradient threshold, then this point is considered an edge point; otherwise, it is a non-edge point.

[0107] Use the gray-level co-occurrence matrix (GLCM) P(i, j, d, θ) to extract texture features, including energy, contrast, entropy, and inverse difference. Use the texture features to distinguish different tissue structures. Among them, energy Contrast Entropy Inverse difference

[0108] According to the statistical features, texture features, and edge information of the image, set the intensity thresholds T imin and Timax and the texture feature threshold T ti , then determine the region R where the key organizational structure is located i , R i = {(x, y, z) D e (x, y, z) ∈ [T imin , T imax , T(x, y, z) ≥ T ti}, where T(x, y, z) represents the texture feature value calculated at the voxel (x, y, z), and T ti represents the texture feature threshold for the organizational structure i; among them, the list of key organizational structures: Uterus (U): including the uterine wall and uterine cavity, Placenta (P): location and scope, Fetal head (F): position and orientation, Bladder (B): avoid injury during surgery, Abdominal wall layer (A): skin, subcutaneous tissue, fascia and muscle; Define spatial position constraints: The uterus (U) should be located within the pelvic cavity and contain the placenta (P) and fetal head (F), and the bladder (B) should be located in the front lower part of the uterus (U);

[0109] Perform non - linear enhancement on the gradient map G(x, y, z), and the edge enhancement function G e (x, y, z) = G(x, y, z) γ , where γ represents the enhancement coefficient, and combine the determined key organizational structure region R i and the enhanced edge information G e (x, y, z) to determine the region of interest ROI i , ROI i = {(x, y, z)(x, y, z) ∈ R i and G e (x, y, z) ≥ T e}, where T e represents the edge intensity threshold.

[0110] In the embodiments of the present invention, noise removal is performed using an anisotropic diffusion filter, significantly improving the quality of the image, reducing noise interference in the image, and facilitating subsequent feature extraction and segmentation; intensity normalization and histogram equalization further enhance the contrast of the image, making key tissue structures clearer, and ensuring accurate identification and marking of key parts such as the placenta and uterus in complex images; the gray histogram and statistical feature extraction can provide a quantitative description of the overall brightness and texture, helping the subsequent segmentation algorithm to perform preliminary classification when identifying different tissue structures; the gradient map and texture features play an important role in edge detection, making the boundaries of key tissue structures more prominent and providing accurate edge information for segmentation; by combining gray and texture features with the organizational structure position constraint conditions, the specific positions and boundaries of each key tissue can be more accurately located, ensuring that the segmentation result can clearly reflect the spatial relationship of different tissue structures and improving the accuracy and rationality of subsequent 3D modeling; by processing the gradient map with a non-linear edge enhancement function, the boundary information of key tissue structures is further enhanced, making the segmentation algorithm more accurate in boundary determination.

[0111] In the preferred embodiment of the present invention, step S2 specifically includes:

[0112] Performing initial segmentation with the determined region of interest as the initial condition to generate a preliminary segmentation mask Preliminary segmentation mask The generation formula is

[0113] Performing fine segmentation using the level set method according to the characteristics of the organizational structure, defining the level set function, and initializing it as: ε represents a positive number, and the energy function is defined Ω represents the image space, μ represents the weight of the smoothing term, λ i represents the weight of the region term, c i represents the average gray value of the target region, represents the Heaviside function, and the level set evolution equation represents the Dirac function;

[0114] Iteratively solving until convergence to obtain an accurate segmentation result of the organizational structure;

[0115] Performing post-processing on the segmentation result, performing opening and closing operations, removing noise, smoothing the edges, retaining the largest connected region, and removing isolated small regions;

[0116] Using the Marching Cubes algorithm to convert the segmentation mask into a 3D model of the key organizational structure, and performing mesh optimization of the model, using Laplacian smoothing V i represents the vertex coordinates, Ni represents the number of neighboring vertices, and λ represents the smoothing coefficient, which reduces the number of polygons and improves the rendering efficiency.

[0117] In the embodiments of the present invention, by combining initial segmentation and fine segmentation, the Marching Cubes algorithm is used to generate a high-precision three-dimensional model, which can truly restore the three-dimensional structure of the key tissues of the pregnant woman's pelvis and abdomen. The accuracy and integrity of the model improve the authenticity of the virtual scene during the training process, enabling the trainee to intuitively and accurately observe the spatial relationship of each anatomical structure in the simulated surgery; in addition, the model optimization further improves the rendering efficiency, ensuring smooth visual effects and interactive experiences in the mixed reality environment.

[0118] In the preferred embodiment of the present invention, S3: Calculate the reference evaluation value for cesarean section surgery training based on the eigenvalue of the key organizational structure, and generate the key indicators for cesarean section surgery training, specifically including:

[0119] Extract anatomical feature parameters, specifically including: calculating the volume of the key organizational structure ΔV = Δx·Δy·Δz, where Vi represents the volume of the key organizational structure i, and M i (x, y, z) represents the segmentation mask, ΔV represents the volume of a single voxel, and Δx, Δy, and Δz represent the voxel spacing of the image in the x, y, and z directions; calculate the surface area of the key organizational structure S i represents the surface area of the key organizational structure i, and A k represents the area of the kth triangular patch, and v 1 , v 2 and v 3 represent the three vertex coordinates of the triangular patch, × represents the vector cross product, and ||·|| represents the vector norm; calculate the centroid position of the key organizational structure (x i , y i , z i ) represents the centroid coordinates of the organizational structure i; extract the evaluation of the compactness of the key organizational structure C i represents the compactness of the organizational structure i, V i represents the volume, and S i represents the surface area; and calculate the distance and relative position between tissues;

[0120] Calculate the incision position P i = P u + R θ ·d, P i represents the proposed incision position, P u represents the centroid position of the uterus, and R θThe rotation matrix constructed with the fetal head orientation angle θ is denoted as, and d represents the preset incision offset vector, d y represents the offset distance of the incision relative to the centroid of the uterus in the y direction;

[0121] Calculate the reference value F of the operating force r = k t ·T l , F r represents the reference operating force, k t represents the tissue hardness coefficient, T l represents the total thickness of the abdominal wall layer;

[0122] Set the reference value of the operation time according to the standard surgical procedure;

[0123] Set the key indicators of surgical training according to clinical experience and teaching requirements: the allowable value of the incision position error ΔP max , that is, the maximum allowable value of the incision position error; the operating force range F min = F r -δF, F max = F r +δF, δF represents the allowable deviation of the force; the operation time limit T max = T r +δT, δT represents the allowable deviation of the time.

[0124] In the embodiment of the present invention, by extracting anatomical parameters such as volume, surface area, centroid position, and compactness, the spatial characteristics and morphological features of key tissues can be accurately reflected, providing a reliable basis for tissue recognition and positioning in training. These characteristic parameters help to improve the simulation of the training model, enabling the trainee to operate in an anatomical structure close to the real one; calculating the incision position based on the fetal head orientation and uterine position, and setting the operating force in combination with tissue hardness and abdominal wall thickness can effectively help the trainee master the appropriate incision position and force control during training. The setting of reasonable reference values and ranges reduces the risk of incorrect operation; setting reasonable reference values and limits for the operation time not only standardizes the operation process of the trainee, but also helps them gradually improve the operation efficiency and the fluency of the surgical operation.

[0125] In the preferred embodiment of the present invention, the S4 specifically includes:

[0126] Convert the constructed three-dimensional model from the original format to the target format according to the requirements of the mixed reality system MR, and convert the coordinate system of the three-dimensional model to the coordinate system used by the MR system; it should be noted that if the model coordinate system is consistent with the MR system, no conversion is required. If not, coordinate system conversion is required, and the coordinate system conversion formula is V c = R c ·V o + Τo , V o represents the vertex coordinate vector of the original model, V c represents the vertex coordinate vector of the transformed model, R c represents the coordinate system rotation matrix, Τ o represents the coordinate system translation vector;

[0127] Build a virtual operating room scene in the MR system, including the following elements: operating table: set the size and position of the operating table; medical devices: surgical instrument models; lighting settings: configure ambient light and spotlight light sources; simulate the lighting conditions of the operating room; other scene elements: walls and floors;

[0128] Import the transformed 3D model into the virtual scene of the MR system, and adjust the position, rotation, and scaling parameters of the model to match the virtual operating room environment;

[0129] Design a user interface in the MR system, including operation guides, information displays, and feedback prompts. The operation guides are used to display surgical steps and precautionary information. The information displays are used to display the operation data and evaluation results of the trainee in real time. The feedback prompts are used to provide visual, auditory, or tactile feedback to indicate operation errors or successes;

[0130] Set interaction methods in the MR system, including gesture recognition, voice control, and handle operations. The gesture recognition is used to support gesture control of the movement and rotation of the model. The voice control is used to interact with the system through voice commands. The handle operations are used to interact using a handle or controller;

[0131] Set rendering effects in the MR system, including material properties, lighting effects, shadows, reflections, and texture mapping. The material properties are used to set the specular and transparency material parameters of the model. The lighting effects are used to adjust the intensity, color, and position of the light source. The shadows and reflections are used to enable shadow and reflection effects. The texture mapping is used to apply high-resolution texture maps;

[0132] Define a real-world space coordinate system: a coordinate system based on the real world, and all positions and postures in the real environment are represented in this coordinate system. Define a virtual space coordinate system: the coordinate system of the virtual model in the MR system;

[0133] Select a corresponding set of landmark points in the real space and the virtual space respectively, and establish a set of real-space landmark points {P i r} , i = 1, 2, ..., N, establish a set of virtual-space landmark points {P i v} , i = 1, 2, ..., N, and use the least squares method to calculate the rigid body transformation matrix T = {R, Τ} such that Calculate the centroids of virtual and real fiducial points Decentralized coordinates Calculate the covariance matrix Perform singular value decomposition on the covariance matrix H = UΣV Τ , calculate the rotation matrix R = VU Τ , calculate the translation vector V a = R·V v + Τ, calculate the root mean square error where, P i r represents the coordinates of the i-th fiducial point in the real space, P i v represents the corresponding i-th fiducial point in the virtual space, N represents the number of fiducial points, U, Σ, and V represent the SVD decomposition results of the covariance matrix, R represents the rotation matrix, Τ represents the translation vector, V v represents the vertex coordinate matrix of the virtual model, V a represents the vertex coordinate matrix after registration, ||·|| represents the Euclidean distance,; set the registration error threshold, if the root mean square error is less than the registration error threshold, it is considered that the registration is successful; otherwise, registration needs to be performed again.

[0134] In the embodiment of the present invention, a highly simulated virtual operating room scene is created in the MR system, enabling trainees to train in an environment close to reality, increasing the immersion and engagement of training. The settings of elements such as the operating table, medical devices, and lighting ensure the integrity and authenticity of the scene, helping trainees adapt to the actual operating environment; through the design of gesture recognition, voice control, and handle operation, trainees can interact with the virtual environment naturally and obtain real-time operation feedback; adding rendering effects such as materials, lighting, and shadows in the virtual surgery scene makes the model and environment more realistic, and the fine rendering effect improves the visual expressiveness of the training system and increases the immersion experience; spatial registration is performed by the least squares method to achieve the position matching between the virtual operating room and the real operating room; spatial registration ensures the coordinate synchronization between the virtual and real spaces, so that when trainees operate in the real space, the movement and position changes of virtual objects can be accurately reflected in the MR system.

[0135] In the preferred embodiment of the present invention, S5 specifically includes:

[0136] Configure sensor devices to collect operation data such as the position and force of trainees during surgical training in real time. The sensor devices include position sensors and force sensors. The position sensors are used to capture the position and posture of trainees' hands and surgical instruments in real time, and the force sensors are used to measure the force exerted by trainees during operation;

[0137] Receive the training start signal, preprocess the collected operation data, and extract key features;

[0138] Compare the real-time operation data of the trainee with the reference evaluation value to generate real-time feedback prompts. Set the position deviation warning threshold and set it to 80% of the maximum allowable value. When it is determined that the position deviation exceeds the position deviation warning threshold, trigger a warning, display a warning message on the interface and play a prompt sound, and record the warning event; set the force deviation warning threshold and set it to 80% of the maximum allowable value. When it is determined that the force deviation exceeds the force deviation warning threshold, trigger a warning, display a warning message on the interface and play a prompt sound, and record the warning event;

[0139] Receive the training end signal, calculate and display the total surgical training score S of the trainee on the interface 总 , the total surgical training score S 总 The calculation formula of is:

[0140] S 总 = w 1 S p + w 2 S f + w 3 S t , w 1 + w 2 + w 3 = 1,

[0141]

[0142] where, w 1 , w 2 and w 3 represent weight coefficients, S p represents the incision position score, S f represents the operation force score, S t represents the surgical time score, ΔP represents the Euclidean distance between the actual incision position of the trainee and the reference incision position, δF represents the allowable force deviation, F a represents the average force applied by the trainee during the operation, F r represents the reference force, T u represents the actual time taken by the trainee to complete the surgery, δT represents the allowable time deviation, T r represents the reference time;

[0143] Set the passing score threshold S 1 , and set the excellent score threshold S 2 , if S 总 > S 2 then the evaluation is "excellent", if S 1 ≤ S 总 < S 2It is evaluated as "qualified" if S 总 <S 1 It is evaluated as "unqualified";

[0144] Generate and display the evaluation result.

[0145] In the embodiment of the present invention, the operation behavior of the trainee is monitored in real time through the position sensor and the force sensor, ensuring the accurate acquisition of the data of the hand, the position of the surgical instrument and the operation force. The preprocessing improves the accuracy of the real-time feedback; by comparing the operation data with the reference value, the system can immediately provide the feedback of the position and force deviation, which helps the trainee to adjust the operation in time during the training process, avoid the accumulation of errors, and the real-time feedback greatly improves the effectiveness of the training, enabling the trainee to master the standard operation method faster; the scoring system for the incision position, operation force and operation time can comprehensively evaluate the quality of the surgical operation, help the trainee to understand the advantages and disadvantages of the operation, and the evaluation mechanism based on the scoring formula provides an objective operation evaluation; the recording of warning events and the interface prompt enable the trainee to review the specific operation deviation after the training, further improving the training effect.

[0146] In the preferred embodiment of the present invention, the S6 specifically includes:

[0147] Record the position trajectory of the hand or instrument of the trainee during the entire training process, record the change of the operation force applied by the trainee over time, record the time of each data point, and save the data in a structured format;

[0148] Receive the training end signal, compare the operation data of the trainee with the key indicators to analyze the reasons for the errors;

[0149] Classify the types of operation errors of the trainee during the training process, including positioning deviation errors where the position deviation exceeds the position deviation warning threshold, force control errors where the force deviation exceeds the force deviation warning threshold, operation process errors of surgical steps, and time errors exceeding the time threshold, and calculate the number of times of different operation error types of the trainee during the training process;

[0150] Provide a positioning practice module for positioning deviation errors, including positioning tasks of different difficulties;

[0151] Provide a force practice module for force control errors, including simulation tissue practice tasks of different hardnesses;

[0152] Provide a video module for operation process errors and time errors, including learning tasks of the standard cesarean section operation process;

[0153] According to the ID file established for the trainee, take the total score of each surgical training of the corresponding trainee as a data point, and update and generate the score curve of the trainee.

[0154] Embodiments of the present invention record the position, force change and timestamp of the hand or instrument in real time, providing detailed process data for the operation behavior of the trainee, classifying different types of operation deviations, enabling the system to accurately identify the deficiencies of the trainee, facilitating the provision of personalized improvement suggestions, providing corresponding training modules and learning videos according to different types of errors, enabling the trainee to improve operation skills targeted, generating the score curve of the trainee, intuitively showing the improvement of the operation level, and enabling the trainee and the instructor to better understand the operation progress trend.

[0155] In a preferred embodiment of the present invention, the method further includes:

[0156] Setting an initial warning threshold D for position deviation at the initial stage of training p0 and an initial warning threshold D for operation force f0 ;

[0157] Dividing the training process into multiple time periods according to the set total training time length T t Each time period has a length of T s , at the end of each time period, judging the operation compliance rate of the trainee during this time period. If the operation compliance rate reaches the preset standard, update the warning threshold and enter the next time period;

[0158] Updating the warning threshold using the dynamic warning threshold update formula;

[0159] During each time period, real-time detect the operation deviation of the trainee. If the position-force deviation exceeds the position deviation warning threshold D p (t) during the current time period, or the force deviation exceeds the force deviation warning threshold D f (t) during the current time period, immediately issue a feedback prompt sound, highlight the specific value of the deviation on the user interface, and at the same time suggest that the trainee adjust the operation. If the number of times the deviation exceeds the threshold in a certain time period exceeds 3 times, the system will record this time period as non-compliant;

[0160] If the operation compliance rate of consecutive N 1 time periods is higher than the first preset compliance rate Q 1 , then adjust the decay rate constant of the threshold in the dynamic warning threshold update formula, that is, increase λ and γ by 20%. If the operation compliance rate of consecutive N 2 time periods is lower than Q 2 , then relax the adjustment rate of the threshold, that is, decrease λ and γ by 10% respectively, where N 1 > N 2 and Q 1 > Q 2 ;

[0161] At 80% of the total training time T t reduce λ and γ to 50% of their initial values respectively;

[0162] At the end of the training, the system generates an operation stability report based on the deviation records throughout the training, and statistically calculates the average value, standard deviation, and compliance rate of the position and force deviations.

[0163] Among them, the dynamic warning threshold update formula includes the dynamic update formula of the position deviation threshold and the dynamic update formula of the force deviation threshold, where

[0164] The dynamic update formula of the position deviation threshold is

[0165] The dynamic update formula of the force deviation threshold is

[0166] Among them, D p (t) represents the position deviation warning threshold at time t, D f (t) represents the force deviation warning threshold at time t, D p0 represents the initial position deviation warning threshold, D f0 represents the force deviation warning threshold, k and m represent the dynamic adjustment coefficients of the position and force feedback, T t represents the total length of the training time, T s represents the duration of a single time period, and λ and γ represent the decay rate constants of the position and force.

[0167] In the embodiments of the present invention, the phased adjustment enables the trainee to gradually adapt to the training standards of each stage, improves the step-by-step nature of skill improvement, enhances the scientificity and sustainability of the training process. The dynamic update formula using sine and cosine functions and time decay makes the feedback threshold gradually decrease during the training process, making the operation standard more stringent over time. The reasonable design of the change form of the formula avoids the situation of threshold mutation, making the training process smoother and more natural, ensuring that the trainee can gradually improve the operation accuracy and force control ability in a stable environment; adaptively adjusting the tightening speed of the threshold according to the compliance of the trainee in multiple time periods makes the training process more flexible and personalized. If the trainee performs well, the tightening speed is accelerated, otherwise it is slowed down, avoiding the training being too aggressive or too lenient. The adaptive adjustment improves the adaptability of the training feedback, enabling the system to flexibly respond to the operation level of the trainee; slowing down the decay rate of the threshold at the end of the training makes the feedback threshold tend to be stable, ensuring that the trainee can perform stable operations under the conditions of high-precision requirements. This design helps the trainee to reach the final surgical accuracy standard in a high-requirement operation environment.

[0168] The embodiments of the present invention further provide device embodiments for implementing the steps and methods in the above method embodiments.

[0169] Please refer to Figure 2 , which is a functional block diagram of a cesarean section surgery training system based on mixed reality provided by the embodiments of the present invention. As Figure 2 shown, the system includes:

[0170] An image data acquisition module, configured to acquire medical image data of a pregnant woman's pelvis and abdomen;

[0171] A preprocessing module, configured to preprocess the acquired medical image data;

[0172] A key organizational structure extraction module, configured to extract eigenvalue of key organizational structures, identify and mark the boundaries and regions of key organizational structures;

[0173] A three-dimensional model construction module, configured to initially segment key organizational structures from the image data, then perform fine segmentation, and use a reconstruction algorithm to construct a three-dimensional model of key organizational structures;

[0174] An index generation module, configured to calculate reference evaluation values for cesarean section surgery training based on the eigenvalue of key organizational structures, and generate key indicators for cesarean section surgery training;

[0175] A mixed reality conversion module, configured to convert the constructed three-dimensional model into a format recognizable by a mixed reality system for importing into the MR system, and perform scene setting and spatial registration;

[0176] A training analysis module, configured to receive a training start signal, detect the surgical operations of the trainee to collect motion data, compare the operation data of the trainee with the key indicators, generate real-time feedback prompts, receive a training end signal, and generate an evaluation result;

[0177] An evaluation feedback module, configured to record operations that do not meet the evaluation criteria, provide targeted improvement suggestions for the trainee, and update and generate a score curve for the trainee.

[0178] The mixed reality-based cesarean section surgery training system of the present invention provides an efficient, accurate and immersive surgery training platform through technical means such as precise image data processing, three-dimensional modeling, real-time feedback and personalized evaluation. This system can not only significantly improve the safety and effectiveness of surgery training, but also achieve personalized guidance during the training process, helping trainees quickly master surgical skills and improve their actual surgical operation ability.

[0179] Although the present invention has been described based on a limited number of embodiments, those skilled in the art in this technical field will understand that other embodiments can be envisioned within the scope of the present invention thus described. For the scope of the present invention, the disclosure of the present invention is illustrative rather than restrictive, and the scope of the present invention is defined by the appended claims.

Claims

1. A cesarean section training method based on mixed reality, characterized in that: The method comprises: S1. Obtain medical imaging data of the pelvis and abdomen of pregnant women, extract characteristic values ​​of key tissue structures after preprocessing, and identify and mark the boundaries and regions of key tissue structures; S2, initially segmenting the key tissue structure from the image data, then performing fine segmentation, and using a reconstruction algorithm to construct a three-dimensional model of the key tissue structure; S3. Calculate the reference evaluation value of cesarean section training based on the characteristic values ​​of the key tissue structures, and generate key indicators of cesarean section training; S4, converting the constructed 3D model into a format recognizable by the mixed reality system to import into the MR system, and performing scene setting and spatial registration; S5, receiving a training start signal, detecting the trainee's surgical operation to collect motion data, comparing the trainee's operation data with key indicators, generating real-time feedback prompts, receiving a training end signal, and generating evaluation results; S6. Record the operations that do not meet the evaluation standards, provide targeted improvement suggestions for trainees, and update the trainees' score curve; The S1 specifically includes: Obtain three-dimensional volume data D(x, y, z) of the pregnant woman's pelvis and abdomen, where x, y, and z represent spatial coordinates, and D represents the intensity value of each voxel; Apply an anisotropic diffusion filter F AD Reduce the noise to get D f (x, y, z), for D f (x, y, z) is normalized to obtain D n (x, y, z), use histogram equalization to enhance the contrast and get D e (x, y, z); The preprocessed image data D e (x, y, z) for feature extraction, specifically: calculating the grayscale histogram, mean and variance statistical features of the image, using the Sobel operator to calculate the gradient map of the image to obtain edge information, and using the grayscale co-occurrence matrix to extract texture features; Determine the area where the key tissue structure is located according to the extracted features, wherein the key tissue structure list includes: uterus (U): including uterine wall and uterine cavity, placenta (P): positioning and range, fetal head (F): position and orientation, bladder (B): to avoid damage during surgery, abdominal wall layer (A): skin, subcutaneous tissue, fascia and muscle; The gradient image is nonlinearly enhanced to enhance the edge information, and the region of interest is determined by combining the enhanced edge information and the key tissue structure region.

2. The method according to claim 1, characterized in that The S2 specifically includes: Perform initial segmentation using the determined region of interest as an initial condition to generate a preliminary segmentation mask; According to the characteristics of the tissue structure, the segmentation algorithm is selected again for segmentation, and the accurate segmentation mask is generated by iterative solution, and the segmentation result is post-processed; The Marching Cubes algorithm was used to convert the segmentation mask into a three-dimensional model of key tissue structures and perform model optimization.

3. The method according to claim 2, characterized in that The S3 specifically includes: Calculate the volume, surface area and center of mass position of key tissue structures, extract and evaluate tissue compactness, and calculate the distance and relative position between tissues; Calculate the incision position and operation force reference value, and set the operation time reference value; Generate key indicators for surgical training: allowable incision position error, operating force range and surgical time limit.

4. The method according to claim 1, characterized in that The S4 specifically includes: Converting the constructed three-dimensional model from the original format to the target format required by the mixed reality system, and converting the coordinate system of the three-dimensional model to the coordinate system used by the MR system; Build a virtual operating room scene in the MR system, including the following elements: operating table: set the size and position of the operating table; medical instruments: surgical instrument models; lighting settings: configure ambient light and spotlight sources; simulate the lighting conditions of the operating room; other scene elements: walls and floors; Importing the converted 3D model into the virtual scene of the MR system, and adjusting the position, rotation, and scaling parameters of the model to match the virtual operating room environment; Designing a user interaction interface in the MR system, including operation instructions, information display and feedback prompts. The operation instructions are used to display surgical steps and precautions; the information display is used to display the trainee's operation data and evaluation results in real time; the feedback prompts are used to provide visual, auditory or tactile feedback to indicate operation errors or successes; An interaction mode is set in the MR system, including gesture recognition, voice control and handle operation, wherein the gesture recognition is used to support the movement and rotation of the gesture control model, the voice control is used to interact with the system through voice commands, and the handle operation is used to interact using a handle or controller; Setting rendering effects in the MR system, including material properties, lighting effects, shadows and reflections, and texture maps, wherein the material properties are used to set the reflectivity and transparency material parameters of the model, the lighting effects are used to adjust the intensity, color, and position of the light source, the shadows and reflections are used to enable shadow and reflection effects, and the texture maps are used to apply high-resolution texture maps; Define the real space coordinate system: a coordinate system based on the real world, in which all positions and postures in the real environment are expressed; define the virtual space coordinate system: a coordinate system of the virtual model in the MR system; Select a set of corresponding landmarks in the real space and virtual space respectively to establish a real space landmark set Establishing a virtual space landmark set The rigid body transformation matrix T = {R, Τ} is calculated using the least squares method, so that Calculate the centroid of virtual and real landmarks Decentralized Coordinates Calculate the covariance matrix Perform singular value decomposition of the covariance matrix H H = UΣV Τ , calculate the rotation matrix R = VU Τ , calculate the translation vector V a =R·V v +Τ, calculate the root mean square error Among them, P i r represents the coordinates of the i-th landmark point in the real space, P i v represents the coordinates of the corresponding i-th marker point in the virtual space, N represents the number of marker points, U, Σ and V represent the SVD decomposition results of the covariance matrix, R represents the rotation matrix, Τ represents the translation vector, V v Represents the vertex coordinate matrix of the virtual model, V a represents the vertex coordinate matrix after registration, ||·|| represents the Euclidean distance, and sets the registration error threshold. If the root mean square error is less than the registration error threshold, the registration is considered successful; otherwise, re-registration is required.

5. The method according to claim 2, characterized in that: The S5 specifically includes: Configure sensor equipment to collect the trainee's position and force operation data during the surgical training in real time. The sensor equipment includes a position sensor and a force sensor. The position sensor is used to capture the position and posture of the trainee's hand and surgical instrument in real time. The force sensor is used to measure the force applied by the trainee during the operation. Receive the training start signal, pre-process the collected operation data, and extract key features; The trainee's real-time operation data is compared with the reference evaluation value, and a real-time feedback prompt is generated. The position deviation warning threshold is set and set to 80% of the maximum allowable value. When the position deviation is judged to exceed the position deviation warning threshold, a warning is triggered, a warning message is displayed on the interface, a prompt sound is played, and a warning event is recorded; the force deviation warning threshold is set and set to 80% of the maximum allowable value. When the force deviation is judged to exceed the force deviation warning threshold, a warning is triggered, a warning message is displayed on the interface, a prompt sound is played, and a warning event is recorded; Receive the training end signal, calculate and display the trainee's total surgical training score S on the interface 总 , the total score of the surgical training is S 总 The calculation formula is: <h2 style=";text-align:left;direction:ltr">S<h2 style=";text-align:left;direction:ltr"> 总 <h2 style=";text-align:left;direction:ltr"> =w1S<h2 style=";text-align:left;direction:ltr"> p <h2 style=";text-align:left;direction:ltr"> +w2S<h2 style=";text-align:left;direction:ltr"> f <h2 style=";text-align:left;direction:ltr"> +w3S<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> ,w1+w2+w3=1, Among them, w1, w2 and w3 represent weight coefficients, S p Indicates the incision location score, S f Indicates the operation strength score, S t represents the surgical time score, △P represents the Euclidean distance between the trainee's actual incision position and the reference incision position, δF represents the force tolerance, and F a It represents the average force applied by the trainee during the operation, F r Indicates the reference strength, T u represents the actual time taken by the trainee to complete the operation, δT represents the time tolerance, and T r Indicates reference time; Set the passing score threshold S1 and the excellent score threshold S2. If S 总 > S2, it is evaluated as "excellent". If S1 ≤ S 总 < S2, it is evaluated as "qualified". If S 总 < S1, it is evaluated as "unqualified"; Generate and display evaluation results.

6. The method according to claim 2, characterized in that The S6 specifically includes: Record the trainee's hand or instrument position trajectory throughout the training process, record the change of the trainee's operating force over time, record the time of each data point, and save the data in a structured format; Receive the training end signal, compare the trainee's operation data with key indicators to analyze the cause of the error; Classify the types of operation errors made by trainees during the training process, including positioning deviation errors where the position deviation exceeds the position deviation warning threshold, force control errors where the force deviation exceeds the force deviation warning threshold, operation process errors where the surgical steps are wrong, and time errors that exceed the time threshold, and calculate the number of different types of operation errors made by trainees during the training process; Provides a positioning practice module for positioning deviation errors, including positioning tasks of different difficulty levels; Provides a strength training module to address strength control errors, including simulated tissue training tasks of different hardness; Provide video modules for operation process errors and time errors, including learning tasks of standard cesarean section operation process; According to the ID file established for the trainee, the total score of each surgical training of the corresponding trainee is used as a data point to update and generate the trainee's score curve.

7. The method according to claim 6, characterized in that The method further comprises: Set an initial warning threshold D for position deviation at the beginning of training p0 The initial warning threshold D of the operation intensity f0 ; According to the total training time T t The training process is divided into multiple time periods, each of which is T long. s ,At the end of each time period, the trainee's operation compliance rate in that time period is judged. If the operation compliance rate reaches the preset standard, the warning threshold is updated and the next time period is entered; Update the warning threshold using a dynamic warning threshold update formula; In each time period, the trainee's operation deviation is detected in real time. If the position force deviation exceeds the position deviation warning threshold in the current time period, D p (t), or when the force deviation exceeds the force deviation warning threshold in the current time period D f (t), immediately give a feedback tone, highlight the specific value of the deviation on the user interface, and suggest trainees to adjust their operations. If the deviation exceeds the threshold more than 3 times within a certain time period, the system will record the time period as not meeting the standard; If the operation compliance rate for N1 consecutive time periods is higher than the first preset compliance rate Q1, the threshold attenuation rate constant in the dynamic warning threshold update formula is adjusted, that is, λ and γ are increased by 20%. If the operation compliance rate for N2 consecutive time periods is lower than Q2, the threshold adjustment rate is relaxed, that is, λ and γ are reduced by 10% respectively, where N1>N2 and Q1>Q2; In the total training time T t When λ and γ are 80% of their initial values, they are reduced to 50% of their initial values ​​respectively; At the end of the training, the system generates an operational stability report based on the deviation records of the entire training process, and statistics the average, standard deviation and compliance rate of position and force deviations.

8. The method according to claim 7, characterized in that The method further includes: the dynamic warning threshold update formula includes a dynamic update formula for a position deviation threshold and a dynamic update formula for a force deviation threshold, wherein: The dynamic update formula of the position deviation threshold is: The dynamic update formula of the force deviation threshold is: Among them, D p (t) represents the position deviation warning threshold at time t, D f (t) represents the force deviation warning threshold at time t, D p0 Indicates the initial position deviation warning threshold, D f0 represents the force deviation warning threshold, k and m represent the dynamic adjustment coefficients of position and force feedback, T t Represents the total length of training time, T s represents the duration of a single time segment, and λ and γ represent the decay rate constants of position and strength.

9. A cesarean section surgery training system based on mixed reality using the method of claim 1, characterized in that: include: An image data acquisition module, used to acquire medical image data of the pelvis and abdomen of a pregnant woman; A preprocessing module, used for preprocessing the acquired medical image data; A key organizational structure extraction module is used to extract the characteristic values ​​of the key organizational structure, identify and mark the boundaries and regions of the key organizational structure; A 3D model building module is used to initially segment key tissue structures from image data, then perform fine segmentation, and use a reconstruction algorithm to build a 3D model of the key tissue structure; An indicator generation module, used for calculating a reference evaluation value of cesarean section training based on characteristic values ​​of key tissue structures, and generating key indicators of cesarean section training; A mixed reality conversion module is used to convert the constructed 3D model into a format recognizable by the mixed reality system so as to import it into the MR system, and perform scene setting and spatial registration; The training analysis module is used to receive a training start signal, detect the trainee's surgical operation to collect action data, compare the trainee's operation data with key indicators, generate real-time feedback prompts, receive a training end signal, and generate evaluation results; The evaluation feedback module is used to record operations that do not meet the evaluation standards, provide targeted improvement suggestions to trainees, and update and generate the trainees' score curve.

Images